@clawhub-chenyqthu-9b381aabcf
画图技能路由中枢（统一入口）。三维路由体系（用途 × 风格 × 主体），双后端调度。 - Signature 风格：10 种有独立 YAML 的视觉方案（构成主义/克莱因/Risograph/故障艺术等） - Rendering 风格：15 种通用渲染技法 modifier（写真/动漫/3D/水彩/赛博朋克等），...
---
name: image-forge
description: |
画图技能路由中枢（统一入口）。三维路由体系（用途 × 风格 × 主体），双后端调度。
- Signature 风格：10 种有独立 YAML 的视觉方案（构成主义/克莱因/Risograph/故障艺术等）
- Rendering 风格：15 种通用渲染技法 modifier（写真/动漫/3D/水彩/赛博朋克等），prompt 源自实战案例
- Logo 展示背景：12 种专业展示场景（来源 logo-generator，已内化）
- 用途库：12 类场景 + 全实战 prompt 案例，含推荐风格 + 后端默认
- 后端调度：GPT Image 2（写实/产品/文字/4K）/ Gemini（动漫/艺术/多参考图）
- 支持：文生图、风格库生图、参考图风格反推、参考图编辑、多参考图合成、logo 展示图
Use when: 用户想画图/生图/做海报/插画/风格迁移/图片编辑/logo展示图 — 这是唯一的图像生成入口。
【铁律】绝对禁止使用 image_generate 工具（configured: no，不可用）。所有画图请求必须走本 skill。
【注意】SVG logo 代码生成 → 请用专属 logo-generator skill。
---
# Image Forge — 统一画图路由
## 目录结构
```
{baseDir}/
├── SKILL.md # 本文件（唯一用户入口）
├── backends.yaml # 后端注册表 + 优先级 + 调度策略
├── styles/
│ ├── index.yaml # 风格库（双层：10 Signature + 15 Rendering）
│ └── *.yaml # 10 个 Signature Style 独立文件
├── use-cases/
│ └── index.yaml # 11 个用途 + 推荐风格 + 默认后端
├── references/ # 用途提示词 JSON（11 个场景）
└── scripts/
├── reverse_style.py # Gemini Vision 15 维风格反推
└── generate_image.py # Gemini/Nano Banana 2 生图
```
---
## 三维框架
```
用途（Use Case）× 风格（Style）× 主体（Subject）
↓ ↓ ↓
结构指令 视觉语言 用户描述
（布局/元素） （色彩/技法/质感） （画什么）
```
三者**独立路由**、**组合注入** prompt。用途和风格可以各自单独触发，也可以同时命中。
---
## 风格库：双层结构（读取 `styles/index.yaml`）
### Tier 1: Signature Styles（有独立 YAML，10 种）
高度具体的视觉方案，命中后加载对应 YAML 文件，默认走 `nano-banana-2`。
| 触发词示例 | 风格 id | 擅长用途 |
|-----------|---------|---------|
| 俄国构成主义、苏联海报、几何宣传 | constructivism | 海报、社媒 |
| 故障艺术、错位矩形、glitch | glitch-window-v1 | 头像、社媒 |
| 窗口重叠、数字拼贴 | glitch-window-v2 | 头像、社媒 |
| 混合媒介、线稿摄影 | mixed-media | 头像、海报 |
| 黑蓝红、三色极简剪影 | tri-color | 海报、封面 |
| 半调雕刻、铜版画、etching | engraving-halftone | 海报、头像 |
| risograph、半调杂志、印刷风 | risograph-magazine | 海报、社媒 |
| 波普水墨、pop art、ink splash | pop-ink-splash | 头像、社媒 |
| 克莱因蓝、克莱因秩序、极简仰拍 | klein-blue-order | 头像、社媒 |
| 高对比度工业、电光蓝故障 | high-contrast-industrial | 海报、产品、封面 |
### Tier 2: Rendering Styles（inline modifier，15 种）
通用渲染技法类别，命中后取 `modifier` 字段直接注入 prompt。按 `preferred_backend` 调度。
| 触发词示例 | 风格 id | 推荐后端 |
|-----------|---------|---------|
| 摄影、写真、真实照片 | photography | **GPT Image 2** |
| 电影感、胶片、cinematic | cinematic-film-still | **GPT Image 2** |
| 3D渲染、三维、CGI | 3d-render | **GPT Image 2** |
| 等距视角、isometric、2.5D | isometric | **GPT Image 2** |
| 复古、retro、vintage | retro-vintage | **GPT Image 2** |
| 赛博朋克、霓虹、cyberpunk | cyberpunk-sci-fi | **GPT Image 2** |
| 极简、minimalism、简约 | minimalism | **GPT Image 2** |
| 动漫、二次元、anime | anime-manga | Gemini |
| 插画、手绘插画 | illustration | Gemini |
| 素描、线稿、sketch | sketch-line-art | Gemini |
| Q版、chibi、可爱 | chibi-q-style | Gemini |
| 像素艺术、pixel art、8-bit | pixel-art | Gemini |
| 油画、古典油画 | oil-painting | Gemini |
| 水彩、aquarelle | watercolor | Gemini |
| 水墨、国画、中国画 | ink-chinese-style | Gemini |
---
## 用途库（读取 `use-cases/index.yaml`）
11 类场景，每类携带推荐风格和默认后端：
| 触发词 | use-case id | 默认后端 | 推荐 Rendering 风格 |
|--------|------------|---------|------------------|
| 海报、传单、poster | poster-flyer | **GPT Image 2** | cinematic, retro, cyberpunk |
| 头像、肖像、avatar | profile-avatar | Gemini | anime, illustration, photography |
| 产品图、营销图 | product-marketing | **GPT Image 2** | photography, 3d-render, minimalism |
| 电商、主图、白底 | ecommerce-main-image | **GPT Image 2** | photography, 3d-render |
| 视频封面、YouTube | youtube-thumbnail | **GPT Image 2** | cinematic, photography |
| 小红书、社交配图 | social-media-post | **GPT Image 2** | illustration, photography, watercolor |
| UI、App、网页 | app-web-design | **GPT Image 2** | 3d-render, isometric, minimalism |
| 漫画、分镜 | comic-storyboard | Gemini | anime-manga, illustration, sketch |
| 游戏素材、角色 | game-asset | Gemini | 3d-render, pixel-art, illustration |
| 信息图、教育图 | infographic-edu-visual | **GPT Image 2** | illustration, isometric, minimalism |
---
## 路由决策树（6 条路径）
```
用户输入
│
├── 有参考图 + "用这个风格"/"反推"
│ → [Path R] 风格反推：reverse_style.py → 提取风格 → 生成
│
├── 有参考图 + "修改"/"编辑"
│ → [Path E] 参考图编辑
│ 1张图 → gpt-image-2 edit endpoint
│ 2+张图 → nano-banana-2 多参考图
│
├── 命中 Signature Style aliases（构成主义/glitch/risograph…）
│ → [Path S] 加载 YAML → prompt recipe → nano-banana-2
│
├── 命中 Rendering Style aliases（动漫/写真/3D/水彩…）
│ → [Path R2] 取 modifier → 注入 prompt → 按 preferred_backend 调度
│
├── 命中用途关键词（海报/头像/电商…）
│ → [Path U] 加载 use-cases/index.yaml → 检索 references JSON
│ → 若无指定风格，展示推荐风格（可跳过直接生成）
│ → 按 use-case.default_backend
│
└── 直接描述主体，无信号
→ [Path D] 优化/翻译英文 → gpt-image-2（默认最高 priority）
```
---
## 后端调度决策（读取 `backends.yaml`）
```
1. 用户显式覆盖（最高优先级）
"用 GPT 画"/"4K高清"/"写实" → gpt-image-2
"用 Gemini 画"/"动漫" → nano-banana-2
2. Style preferred_backend
Signature 风格命中 → nano-banana-2（全部 10 种）
Rendering 风格命中 → 按各风格的 preferred_backend（见上表）
3. Use-case default_backend
无风格指定时，按用途默认后端
4. 全局默认
gpt-image-2（priority 最高）
```
**GPT Image 2 强项**：写实摄影、产品展示、文字渲染、4K 高清、海报、UI
**Gemini 强项**：动漫/插画/中国风/水彩/素描、多参考图合成、Signature 风格迁移
---
## [Generation] — 后端执行
### GPT Image 2（CRS 路由）
**推荐使用 wrapper 脚本**（支持 generate + edit，多图 edit，自动处理 base64）：
```bash
# 文生图
uv run {baseDir}/scripts/gpt_image2.py generate \
--prompt "<prompt>" \
--output /path/out.png \
--size 1536x1024 \
--quality high
# 改图（单张参考图）
uv run {baseDir}/scripts/gpt_image2.py edit \
--prompt "<edit instruction>" \
-i /path/ref.png \
--output /path/out.png \
--size 1024x1536
# 改图（多张参考图，最多 4 张）
uv run {baseDir}/scripts/gpt_image2.py edit \
--prompt "<instruction>" \
-i ref1.png -i ref2.png \
--output /path/out.png
```
> **注意**：edit 接口不支持 `input_fidelity` 参数（已验证 2026-04-25）。
**Python API（内联使用）**：
```python
import os, requests, base64, time
CRS_BASE = os.environ.get('CRS_BASE_URL', 'http://127.0.0.1:8765')
CRS_KEY = os.environ['CRS_API_KEY']
def gpt_image2_generate(prompt, size='1536x1024', quality='high',
output_format='png', filename=None):
resp = requests.post(
f'{CRS_BASE}/openai/v1/images/generations',
headers={'Authorization': f'Bearer {CRS_KEY}'},
json={'model': 'gpt-image-2', 'prompt': prompt, 'size': size,
'quality': quality, 'output_format': output_format,
'response_format': 'b64_json'},
timeout=180,
)
data = resp.json()['data'][0]
out = filename or f'/tmp/image-forge-{int(time.time())}.{output_format}'
with open(out, 'wb') as f:
f.write(base64.b64decode(data['b64_json']))
return out, data.get('revised_prompt', '')
def gpt_image2_edit(prompt, image_path, size='1536x1024', quality='high',
output_format='png', filename=None):
with open(image_path, 'rb') as f:
b64_img = base64.b64encode(f.read()).decode()
resp = requests.post(
f'{CRS_BASE}/openai/v1/images/edits',
headers={'Authorization': f'Bearer {CRS_KEY}'},
json={'model': 'gpt-image-2', 'prompt': prompt,
'images': [{'image_url': f'data:image/png;base64,{b64_img}'}],
'size': size, 'quality': quality,
'output_format': output_format, 'response_format': 'b64_json'},
timeout=180,
)
data = resp.json()['data'][0]
out = filename or f'/tmp/image-forge-edit-{int(time.time())}.{output_format}'
with open(out, 'wb') as f:
f.write(base64.b64decode(data['b64_json']))
return out, data.get('revised_prompt', '')
```
**GPT Image 2 尺寸**：`1024x1024` / `1536x1024` / `1024x1536` / `2048x2048` / `3840x2160` (4K横) / `2160x3840` (4K竖)
### Gemini / Nano Banana 2
```bash
# 文生图
uv run {baseDir}/scripts/generate_image.py \
--prompt "<optimized_english_prompt>" \
--filename "~/.openclaw/workspace/tmp/image-forge/$(date +%Y-%m-%d-%H-%M-%S)-<slug>.png" \
--aspect-ratio "<1:1|3:4|4:3|9:16|16:9>"
# 改图 / 多参考图合成（已实测 2026-04-25）
# Gemini 会在参考图基础上按 prompt 修改，多图合成/风格迁移尤其适合
uv run {baseDir}/scripts/generate_image.py \
--prompt "<e.g.: keep character, change background to warm sunset>" \
--filename "~/.openclaw/workspace/tmp/image-forge/$(date +%Y-%m-%d-%H-%M-%S)-<slug>.png" \
-i "/path/to/ref1.jpg" -i "/path/to/ref2.jpg" \
--aspect-ratio "3:4"
```
> **Gemini edit vs GPT Image 2 edit**
> - Gemini：多图合成、风格迁移更自由，但对原图布局保留能力较弱
> - GPT Image 2：保留原图布局/文字/边框精确修改时更强，推荐用于卡牌、产品展示图的约束性编辑
---
## Prompt 组合逻辑
```
Final Prompt =
[Rendering Style modifier（如有）]
+ [Signature Style prompt（如有，替换主体后）]
+ [Use-case 结构指令（如有，从 references JSON 取）]
+ [用户主体描述（中→英翻译优化）]
+ [技术参数（lighting / composition / quality）]
```
- 中文输入全部翻译为英文后发给两个后端
- Signature Style prompt 已含完整视觉语言，Rendering modifier 作补充层
- 两者同时命中时：Signature 优先（更具体），Rendering 作辅助修饰
---
## 输出交付
- 保存目录：`~/.openclaw/workspace/tmp/image-forge/`
- 文件名：`YYYY-MM-DD-HH-MM-SS-<slug>.png`
- 回复：说明所选路径 + 后端 + 关键 prompt 要点，不读取二进制
### 渠道交付规则
| 渠道 | 交付方式 |
|------|--------|
| **飞书** | `message` tool + `filePath`（发送原生飞书图片消息） |
| Discord / 其他渠道 | `MEDIA: /absolute/path` （自动 inline） |
飞书交付示例：
```
message action=send filePath=/abs/path/to/image.png
```
【注意】一次生成多张图时，分次发送每张图片。
---
## 典型示例
```
# [Path D] 默认 GPT Image 2
"画一只在宇宙中游泳的猫"
→ gpt-image-2，size=1536x1024
# [Path S] Signature 风格 + Gemini
"帮我画一张俄国构成主义风格的 AI 机器人海报"
→ constructivism.yaml → nano-banana-2，aspect=3:4
# [Path R2] Rendering 风格 → 自动按强项调度
"帮我画一张动漫风格的城市夜景"
→ anime-manga modifier → nano-banana-2
"帮我画一张赛博朋克风城市"
→ cyberpunk-sci-fi modifier → gpt-image-2
# [Path U] 用途路由 + 推荐风格
"帮我做一张 YouTube 视频封面，科技感"
→ youtube-thumbnail.json → 推荐 cinematic/photography → gpt-image-2
# [Path U + R2] 用途 + 风格同时命中
"帮我做一张水彩风格的社交配图，主题是咖啡和阅读"
→ social-media-post + watercolor → nano-banana-2，aspect=1:1
# [Path E] 参考图编辑
1张图 + "改成极简风格" → gpt-image-2 edit endpoint
2张图 + "合成一张" → nano-banana-2 (-i ref1 -i ref2)
# [Path R] 风格反推
1张图 + "用这个风格给我画一只猫" → reverse_style.py → gpt-image-2
# 显式后端覆盖
"用 Gemini 画一张产品图" → nano-banana-2（覆盖用途默认）
"4K高清画一张产品海报" → gpt-image-2，size=3840x2160
```
FILE:EXTEND.md
# Image Forge 扩展指南 (EXTEND.md)
> 本指南说明如何向 image-forge 添加新风格、用途、后端和子技能。
> 核心原则：**只改 YAML，不改 SKILL.md 路由逻辑**。
---
## 当前库状态
| 资产 | 数量 | 质量说明 |
|------|------|---------|
| Signature 风格（YAML） | 10 种 | Sallyn 原创，有完整测试过的 prompt recipe |
| Rendering 风格（inline modifier） | 15 种 | 分类体系来自 YouMind/awesome-gpt-image-2，modifier 内容待进一步验证和丰富 |
| 用途 + references JSON | 11 类 | 来自 nano-banana/YouMind，有实际 prompt 示例 |
| 后端 | 2 个 | GPT Image 2 (CRS) + Gemini (Nano Banana 2) |
**待补充**：
- Rendering modifier 的实际 prompt 案例（从 awesome-gpt-image-2、EvoLinkAI 等 repo 导入）
- 更多 Signature 风格（如霓虹全息、磨砂玻璃、3D 黏土、吉卜力）
- logo/品牌类用途的完整接入
---
## 1. 添加 Rendering 风格（最轻量）
只需在 `styles/index.yaml` 的 `rendering_styles` 块添加一条：
```yaml
- id: frosted-glass # 唯一 id，小写短横线
category: material-render # 类别（photo/illustration/3d/fine-art/print-art/minimal/digital-art/material-render）
aliases: [磨砂玻璃, frosted glass, 毛玻璃, glassmorphism] # 触发词
modifier: "frosted glass material, translucent surface, soft blur behind glass, light refraction, clean modern aesthetic, studio lighting"
preferred_backend: gpt-image-2 # gpt-image-2 / nano-banana-2
tags: [glass, material, modern]
```
**prompt 来源参考**：
- [awesome-gpt-image-2-prompts](https://github.com/EvoLinkAI/awesome-gpt-image-2-prompts) — EvoLink 按用例整理的 GPT Image 2 案例
- [awesome-gpt-image-2](https://github.com/YouMind-OpenLab/awesome-gpt-image-2) — YouMind 1500+ 分类 prompt
- [awesome-nano-banana-pro-prompts](https://github.com/YouMind-OpenLab/awesome-nano-banana-pro-prompts) — Gemini 向 10000+ prompt
选一个代表性 case 的 prompt 精炼为 modifier（去掉主体描述，保留视觉语言部分）。
---
## 2. 添加 Signature 风格（有独立 YAML）
**Step 1**：新建 `styles/<id>.yaml`，参考已有文件格式：
```yaml
# styles/frosted-glass.yaml
id: frosted-glass
name: 磨砂玻璃
category: material-render
description: 磨砂玻璃质感，通透朦胧，现代高级感
prompt: |
[在此处替换为您想要生成的主体内容]，frosted glass material,
translucent surface with soft blur, subtle light caustics,
clean studio background, minimalist composition,
photorealistic render, soft ambient lighting
placeholder: "[在此处替换为您想要生成的主体内容]"
aspect_ratio: "1:1"
preferred_backend: gpt-image-2
tags: [glass, material, premium, modern]
test_subject: "a smartphone floating above desk" # 用于验证的主体
```
**Step 2**：在 `styles/index.yaml` 的 `signature_styles` 块添加条目：
```yaml
- id: frosted-glass
file: frosted-glass.yaml
category: material-render
aliases: [磨砂玻璃, frosted glass, 毛玻璃, glassmorphism]
aspect_ratio: "1:1"
preferred_backend: gpt-image-2
tags: [glass, material, modern]
use_case_affinity: [product-marketing, app-web-design, profile-avatar]
avoid_for: []
```
**Step 3（可选）**：在 `use-cases/index.yaml` 相关用途的 `recommended_signature` 里加上新 id。
**验证**：用 test_subject 实际跑一次，确认 prompt 效果。
---
## 3. 添加用途（Use Case）
**Step 1**：创建 `references/<id>.json`，参考已有格式：
```json
[
{
"title": "Brand Logo Showcase — Dark Background",
"prompt": "professional product showcase, dark studio background, dramatic lighting...",
"tags": ["logo", "brand", "showcase", "dark"]
},
...
]
```
**Step 2**：在 `use-cases/index.yaml` 添加条目：
```yaml
- id: brand-logo
label: "品牌 Logo / 展示图"
aliases: [logo, 图标, 品牌, brand, 徽标, icon, 标志]
references_file: "references/brand-logo.json"
recommended_signature:
- high-contrast-industrial
recommended_rendering:
- minimalism
- 3d-render
- photography
default_backend: gpt-image-2
default_size: "1024x1024"
default_aspect: "1:1"
special_note: "SVG logo 生成请使用专属 logo-generator skill"
```
---
## 4. 接入新后端（模型）
在 `backends.yaml` 的 `backends` 列表添加条目，设更高 `priority` 即可成为新默认：
```yaml
- id: flux-ultra
priority: 15 # 比 gpt-image-2 的 10 更高 → 成为新默认
enabled: true
description: "FLUX Ultra，极高写实细节"
type: api
endpoint: "https://..."
auth_header: "Bearer $FLUX_API_KEY"
default_size: "1024x1024"
timeout_s: 120
strong_at_rendering:
- photography
- 3d-render
strong_at_use_cases:
- product-marketing
- ecommerce-main-image
```
**后端类型约定**：
- `crs`：通过本地 CRS 代理，用 `CRS_API_KEY`
- `gemini`：用 `generate_image.py` 脚本
- `api`：直接 HTTP，在 SKILL.md 的 Generation 节补充调用代码
---
## 5. 接入垂直子技能
**场景**：Lucien 分享了一个新的专属画图技能（如食品摄影专项、建筑可视化专项）
**接入方式一：用途路由引用**（推荐）
在 `use-cases/index.yaml` 加一条新用途，`special_note` 字段说明有专属技能可用：
```yaml
- id: food-photography
label: "食品/美食摄影"
aliases: [美食, 食品, 菜品, food, 餐饮摄影]
references_file: "references/food-photography.json"
recommended_rendering: [photography, 3d-render]
default_backend: gpt-image-2
default_size: "1024x1024"
special_note: "如有 food-photo-skill 则优先加载专属技能"
```
**接入方式二：独立技能保持，image-forge 做意图识别**
image-forge 识别到特定意图时，在回复里推荐切换专属技能：
```
用户说"帮我做一个 App 的 icon"
→ image-forge 可以生成，但如果识别到 logo-generator skill 存在
→ 回复："这个场景有专属的 logo-generator skill，可以生成 SVG 格式并配高端展示图，
是否切换？或者我直接用 image-forge 画一张栅格图"
```
**接入方式三：Signature 风格 + 独立 references**
把新技能的精华 prompt 提炼为 Signature 风格 YAML 加入 image-forge，
同时保留原技能做深度使用（当用户需要完整工作流时）。
---
## 6. 从高星 Repo 批量导入 Prompt
**推荐流程**：
```bash
# 1. 下载目标 repo 的 JSON
curl -s "https://raw.githubusercontent.com/YouMind-OpenLab/awesome-gpt-image-2/main/..." \
-o /tmp/source-prompts.json
# 2. 用脚本提炼为 references 格式（去掉主体，保留风格语言）
python3 scripts/import_prompts.py \
--source /tmp/source-prompts.json \
--category anime-manga \
--output references/anime-manga-examples.json
# 3. 在 use-cases/index.yaml 对应条目加 examples_file 字段引用
```
**目前待导入的来源**：
- [ ] [EvoLinkAI/awesome-gpt-image-2-prompts](https://github.com/EvoLinkAI/awesome-gpt-image-2-prompts) — 人像/海报/UI case
- [ ] [YouMind-OpenLab/awesome-gpt-image-2](https://github.com/YouMind-OpenLab/awesome-gpt-image-2) — 15 个 style 分类的实际 prompt
- [ ] [YouMind-OpenLab/awesome-nano-banana-pro-prompts](https://github.com/YouMind-OpenLab/awesome-nano-banana-pro-prompts) — Gemini 向各类 prompt
---
## 快速扩容检查清单
```
新增 Rendering 风格：
□ styles/index.yaml 加条目（id / aliases / modifier / preferred_backend）
□ 用实际 prompt 验证 modifier 有效
新增 Signature 风格：
□ 新建 styles/<id>.yaml
□ styles/index.yaml 加条目
□ （可选）相关 use-case 的 recommended_signature 加引用
□ 实际生成一张验证
新增用途：
□ references/<id>.json（5-10 条 prompt 示例）
□ use-cases/index.yaml 加条目（含 recommended_styles + default_backend）
新增后端：
□ backends.yaml 加条目（priority / endpoint / strong_at）
□ SKILL.md Generation 节补充调用代码
接入子技能：
□ use-cases/index.yaml 加引用条目 + special_note
□ 或把核心 prompt 提炼为 Signature 风格
```
FILE:README.md
# image-forge
> AI 画图统一路由技能 for OpenClaw — 三维路由体系 × 双后端调度
## 功能概览
- **5 条意图路径**：风格反推 / 参考图编辑 / 风格库 / 用途路由 / 直接生成
- **37 种风格**：10 Signature（独立 YAML） + 15 Rendering（inline modifier） + 12 Logo 展示背景
- **12 类用途**：海报/头像/电商/YouTube/社媒/App/漫画/游戏/信息图/logo 展示等
- **双后端调度**：GPT Image 2（写实/产品/文字）/ Gemini Imagen 3（动漫/艺术/多参考图）
- **Prompt 库**：15 个场景 JSON，含 YouMind/EvoLink 实战案例精炼
## 安装
```bash
# 通过 clawhub 安装（推荐）
clawhub install image-forge
# 或手动克隆
git clone https://github.com/your-username/image-forge
# 将 image-forge/ 目录放入 OpenClaw workspace/skills/ 下
```
## 环境配置
复制 `.env.example` 并填入你的 key：
```bash
cp .env.example .env
```
| 变量 | 说明 | 必填 |
|------|------|------|
| `CRS_API_KEY` | Claude Relay Service API Key（用于 GPT Image 2） | GPT Image 2 后端必填 |
| `CRS_BASE_URL` | CRS 服务地址，默认 `http://127.0.0.1:8765` | 可选 |
| `GEMINI_API_KEY` | Google Gemini API Key（用于 Nano Banana 2） | Gemini 后端必填 |
| `NANO_BANANA_API_KEY` | Nano Banana API Key（备用） | 可选 |
> **注**：CRS（Claude Relay Service）是一个 self-hosted OpenAI 兼容代理，通过 ChatGPT Plus 账号访问 GPT Image 2。如果你没有 CRS，可以配置任何兼容 `/openai/v1/images/generations` 的服务，或直接使用 OpenAI 官方 API。
## 后端支持
| 后端 | 调用方式 | 擅长场景 |
|------|---------|---------|
| GPT Image 2 | CRS / 任意 OpenAI 兼容端点 | 写实摄影、产品图、文字渲染、4K、海报 |
| Gemini Imagen 3 | `scripts/generate_image.py` | 动漫、插画、中国风、水彩、多参考图 |
要切换为官方 OpenAI API，修改 `backends.yaml`：
```yaml
- id: gpt-image-2
endpoint: "https://api.openai.com/v1/images/generations"
auth_header: "Bearer $OPENAI_API_KEY"
```
## 扩展
- **加新风格** → `styles/index.yaml`
- **加新用途** → `use-cases/index.yaml` + `references/`
- **加新后端** → `backends.yaml`
- 详细说明见 `EXTEND.md`
## 许可证
- 本技能代码：MIT
- `references/` 下来源于 YouMind/EvoLink 的 JSON 内容：CC BY 4.0（见下方致谢）
- Signature 风格 YAML（原创）：MIT
## 致谢
- [YouMind-OpenLab/awesome-gpt-image-2](https://github.com/YouMind-OpenLab/awesome-gpt-image-2) (CC BY 4.0) — 用途分类体系 + prompt 案例
- [EvoLinkAI/awesome-gpt-image-2-prompts](https://github.com/EvoLinkAI/awesome-gpt-image-2-prompts) (CC BY 4.0) — 实战 prompt 案例
- [YouMind-OpenLab/awesome-nano-banana-pro-prompts](https://github.com/YouMind-OpenLab/awesome-nano-banana-pro-prompts) — Gemini 用途 JSON 原始来源
FILE:backends.yaml
# image-forge 后端注册表 + 调度策略
# 优先级规则：用户显式指定 > style.preferred_backend > use_case.default_backend > global_default
# 新后端上线：加条目 + 设 priority，不用改 SKILL.md
backends:
- id: gpt-image-2
priority: 10
enabled: true
description: "OpenAI GPT Image 2 via CRS，写实/产品/海报/文字渲染首选"
type: crs
endpoint: "-http://127.0.0.1:8765/openai/v1/images/generations"
edit_endpoint: "-http://127.0.0.1:8765/openai/v1/images/edits"
auth_header: "Bearer $CRS_API_KEY"
default_size: "1536x1024"
supported_sizes:
- "1024x1024"
- "1536x1024"
- "1024x1536"
- "2048x2048"
- "3840x2160"
- "2160x3840"
default_quality: "high"
output_format: "png"
timeout_s: 180
# 该后端擅长的 rendering_style id
strong_at_rendering:
- photography
- cinematic-film-still
- 3d-render
- isometric
- retro-vintage
- cyberpunk-sci-fi
- minimalism
# 该后端擅长的 use_case id
strong_at_use_cases:
- poster-flyer # 文字渲染是核心优势
- product-marketing
- ecommerce-main-image
- youtube-thumbnail
- app-web-design
- infographic-edu-visual
- id: nano-banana-2
priority: 5
enabled: true
description: "Gemini Imagen 3 via Nano Banana，艺术风格/多参考图/动漫/中国风首选"
type: gemini
script: "{baseDir}/scripts/generate_image.py"
default_aspect_ratio: "16:9"
supported_aspect_ratios: ["1:1", "3:4", "4:3", "9:16", "16:9"]
timeout_s: 120
# 该后端擅长的 rendering_style id
strong_at_rendering:
- anime-manga
- illustration
- sketch-line-art
- chibi-q-style
- pixel-art
- oil-painting
- watercolor
- ink-chinese-style
# 该后端擅长的 use_case id
strong_at_use_cases:
- profile-avatar # 风格迁移灵活
- comic-storyboard
- game-asset
# ─────────────────────────────────────────
# 自动调度决策树
# ─────────────────────────────────────────
dispatch:
global_default: gpt-image-2
# 触发切换到 nano-banana-2 的信号（任一满足即切换）
prefer_nano_banana_when:
- multi_reference_images: true # 用户提供 2+ 张参考图
- signature_style_matched: true # 命中 Signature 风格（所有 10 个）
- rendering_style_in: # 命中以下 Rendering 风格
- anime-manga
- illustration
- sketch-line-art
- chibi-q-style
- pixel-art
- oil-painting
- watercolor
- ink-chinese-style
# 强制走 gpt-image-2 的任务类型（无论 rendering 属性如何）
# 根据 2026-04-24 卡牌实测：Gemini 在小细人物+排版组合场景质量较弱
force_gpt_image2_when:
- task_type_in: [card-art, trading-card, character-card, product-showcase]
- has_text_layout: true # 有文字排版需求（技能标签、stats、屡历卡）
- realistic_portrait_with_frame: true # 写实人物肖像 + 卡牌边框
# 用户显式触发词（覆盖一切自动判断）
user_override:
gpt-image-2:
- "用 GPT 画"
- "GPT Image 2"
- "gpt-image"
- "gpt图"
- "4K高清"
- "写实"
- "真实照片"
nano-banana-2:
- "用 Gemini 画"
- "nano banana"
- "Imagen"
- "gemini画"
FILE:references/brand-logo-showcase.json
[
{
"title": "THE VOID — 绝对虚空",
"source": "logo-generator/references/background_styles.md",
"concept": "Absolute minimalism and mystery. Infinite void, distant starlight at universe edge.",
"prompt": "{logo_description} logo centered on pure black (#000000) background, extremely fine silver-white high-contrast micro noise texture, cold sharp electronic film grain, minimal icy blue glow at extreme corner, generous negative space around logo, white or silver logo color, professional brand identity presentation",
"backend": "nano-banana-2",
"aspect_ratio": "1:1",
"suitable_for": ["hardcore tech", "data security", "infrastructure", "Web3"],
"tags": ["dark", "minimal", "tech", "mystery"]
},
{
"title": "FROSTED HORIZON — 磨砂穹顶",
"source": "logo-generator/references/background_styles.md",
"concept": "Modern breathing space with physical thickness. Sophisticated, breathable, Apple-like presentation.",
"prompt": "{logo_description} logo on deep titanium gray background, organic film-like dust texture, unpolished rough metal surface quality, large area low-saturation cold gray-blue light halo at edges dissolved like mist, premium breathing space, white logo color, Apple-quality presentation aesthetics",
"backend": "nano-banana-2",
"aspect_ratio": "1:1",
"suitable_for": ["premium products", "design brands", "consumer tech"],
"tags": ["dark", "premium", "metal", "breathable"]
},
{
"title": "FLUID ABYSS — 流体深渊",
"source": "logo-generator/references/background_styles.md",
"concept": "AI-native with data fluidity. Mysterious, dynamic, computational.",
"prompt": "{logo_description} logo on deep midnight purple background, slight color-tinted noise, fluid fusion of dark orange from right edge and dark blue from left slowly interweaving, deep-sea nebula quality texture, mysterious computational atmosphere, white logo centered with generous space",
"backend": "nano-banana-2",
"aspect_ratio": "1:1",
"suitable_for": ["AI products", "data visualization", "dynamic systems"],
"tags": ["dark", "ai", "fluid", "dynamic"]
},
{
"title": "STUDIO SPOTLIGHT — 物理影棚",
"source": "logo-generator/references/background_styles.md",
"concept": "Physical studio lighting simulation. Editorial magazine quality.",
"prompt": "{logo_description} logo on extremely dark warm carbon gray background, slightly larger grain simulating low-light camera photography, paper print grain in weak light, single-side softbox creating natural vignette, editorial magazine quality, white logo centered",
"backend": "nano-banana-2",
"aspect_ratio": "1:1",
"suitable_for": ["editorial design", "magazine brands", "professional services"],
"tags": ["dark", "editorial", "studio", "magazine"]
},
{
"title": "ANALOG LIQUID — 物理流体",
"source": "logo-generator/references/background_styles.md",
"concept": "Physical fluid textures on solid color. Extreme contrast between chaotic texture and clean logo.",
"prompt": "{logo_description} logo centered on vibrant Klein blue (#002FA7) solid color base, microscopic cellular patterns and thermal imaging roughness overlay, metallic gold dust flow and iridescent pigment shimmer texture, chaotic organic metallic texture contrasting with sharp clean vector logo, white logo color, artistic brand identity",
"backend": "nano-banana-2",
"aspect_ratio": "1:1",
"suitable_for": ["creative tools", "artistic brands", "experimental products"],
"tags": ["dark", "creative", "metallic", "contrast"]
},
{
"title": "LED MATRIX — 数字硬件",
"source": "logo-generator/references/background_styles.md",
"concept": "Digital retro and pixel matrix. Hardcore geek, cyberpunk, retro-futurism.",
"prompt": "{logo_description} logo on pure black background, glowing dot matrix patterns creating depth, CRT display artifacts and halftone printing dots, retro LED billboard aesthetic, waves of glowing green-amber points receding into background, logo as solid entity in front, cyberpunk retro-futurism atmosphere",
"backend": "nano-banana-2",
"aspect_ratio": "1:1",
"suitable_for": ["AI computing", "Web3", "electronic hardware", "data services"],
"tags": ["dark", "cyberpunk", "retro", "digital"]
},
{
"title": "EDITORIAL PAPER — 纸本编辑",
"source": "logo-generator/references/background_styles.md",
"concept": "High-end specialty paper with extreme whitespace. Humanistic, independent magazine aesthetic.",
"prompt": "{logo_description} logo on off-white alabaster paper background, high-grade watercolor rough art paper texture, natural diffused light reflection, subtle warm gray vignette at corners, generous breathing negative space, humanistic independent magazine quality, dark logo color",
"backend": "nano-banana-2",
"aspect_ratio": "1:1",
"suitable_for": ["serious brands", "human-centered products", "fashion", "professional services"],
"tags": ["light", "paper", "editorial", "humanistic"]
},
{
"title": "IRIDESCENT FROST — 幻彩透砂",
"source": "logo-generator/references/background_styles.md",
"concept": "Minimal tech with optical material beauty. Apple hardware render quality.",
"prompt": "{logo_description} logo on extremely light silver-gray cold white background, ultra-fine micro noise texture, frosted glass or sandblasted aluminum surface quality, soft holographic iridescent colors — light purple, light blue, soft pink — seen through thick frosted glass, Apple-quality optical material aesthetic, dark logo color",
"backend": "nano-banana-2",
"aspect_ratio": "1:1",
"suitable_for": ["tech products", "hardware", "scientific applications"],
"tags": ["light", "iridescent", "optical", "premium"]
},
{
"title": "MORNING AURA — 晨雾光域",
"source": "logo-generator/references/background_styles.md",
"concept": "AI softness with approachability. Warm, intelligent, pressure-free.",
"prompt": "{logo_description} logo on warm ivory cream background, soft noise blending like morning mist, large blurred low-saturation pastel colors — mint green, baby blue, dawn orange — dissolving into warm white background, atmospheric morning haze quality, approachable intelligent mood, dark logo color",
"backend": "nano-banana-2",
"aspect_ratio": "1:1",
"suitable_for": ["user-friendly AI", "accessible products", "health tech"],
"tags": ["light", "soft", "warm", "approachable"]
},
{
"title": "CLINICAL STUDIO — 无菌影棚",
"source": "logo-generator/references/background_styles.md",
"concept": "Spatial order with high contrast. Sterile space, geometric order, 3D depth in 2D.",
"prompt": "{logo_description} logo on pure white or extremely light cold gray background, high-frequency sharp cold-toned digital micro noise, large softbox from above creating smooth gray-white gradient shadow, pure light-shadow structure, sterile spatial order, algorithm-driven confidence, dark logo color",
"backend": "nano-banana-2",
"aspect_ratio": "1:1",
"suitable_for": ["algorithm-driven brands", "data-centric products", "SaaS"],
"tags": ["light", "clinical", "minimal", "confident"]
},
{
"title": "UI CONTAINER — 容器化界面",
"source": "logo-generator/references/background_styles.md",
"concept": "Digital product native feel. Interactive, product-ready, digital asset quality.",
"prompt": "{logo_description} logo displayed inside a frosted glass container — rounded corners, subtle transparency, micro drop-shadow — on clean gradient background, UI-native presentation quality, suggesting interactivity and digital context, SaaS platform aesthetic, both light and dark variants available",
"backend": "nano-banana-2",
"aspect_ratio": "1:1",
"suitable_for": ["digital products", "apps", "SaaS platforms", "UI/UX brands"],
"tags": ["light", "digital", "ui", "interactive"]
},
{
"title": "SWISS FLAT — 瑞士扁平",
"source": "logo-generator/references/background_styles.md",
"concept": "Absolute flatness and timeless authority. Zero gradients, zero effects.",
"prompt": "{logo_description} logo on 100% pure solid deep vintage green background, absolutely flat — zero gradients, zero noise, zero texture, zero effects — pure graphic design with only color and form, extreme confidence, Helvetica-era Swiss design authority, white logo color, maximum negative space",
"backend": "nano-banana-2",
"aspect_ratio": "1:1",
"suitable_for": ["established brands", "environmental products", "classic institutions"],
"tags": ["solid", "swiss", "flat", "authority"]
}
]
FILE:references/infographic-premium.json
[
{
"title": "Illustrated City Food Map",
"source": "YouMind/awesome-gpt-image-2 @mm_zzm44854 — Featured",
"prompt": "Hand-drawn illustrated tourist map infographic of {subject} food and landmark guide, watercolor and ink illustration on vintage parchment paper, cartoon mascot in title section, vine/botanical border decoration, textured beige parchment background with colored roads and water features, numbered food locations with small illustrations of each dish, landmark illustrations with labels, legend in bottom corner with compass rose, warm hand-crafted aesthetic",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["map", "food", "illustrated", "infographic", "watercolor"]
},
{
"title": "Exploded View Technical Infographic",
"source": "YouMind/awesome-gpt-image-2 @wory37303852 — Featured",
"prompt": "Technical exploded view infographic poster of {subject}, clean high-tech 3D render style, product disassembled vertically showing all internal components as distinct labeled layers, callout lines with technical component descriptions on both sides, professional product specification layout, gradient background, title with product name and tagline at top, technical specifications footer",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["exploded view", "technical", "product", "infographic", "3D"]
},
{
"title": "Science Encyclopedia Vertical Infographic",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @pfanis",
"prompt": "Educational science encyclopedia infographic about {subject}, clean editorial layout with bold section headers, detailed scientific illustrations with labels and callout annotations, information hierarchy from overview to detail, consistent color coding by topic, clean white background with structured grid layout, museum or textbook publication quality",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["science", "encyclopedia", "educational", "infographic", "editorial"]
},
{
"title": "City Travel Guide Infographic",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @MrLarus",
"prompt": "City travel guide infographic for {subject}, isometric-style map illustration showing key districts and attractions, color-coded neighborhoods, illustrated landmarks with pop-up info cards, transportation routes, rating stars and category icons, clean editorial typography, modern travel magazine aesthetic",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["travel", "guide", "infographic", "illustrated", "map"]
},
{
"title": "Cooking Process Flowchart",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @Kurt_Rousey466",
"prompt": "Step-by-step cooking process flowchart for {subject} recipe, illustrated food icons at each step, clear numbered sequence with arrows, ingredients list section, timing and temperature annotations, warm food photography palette, clean instructional design, suitable for cooking blog or social media recipe card",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["cooking", "flowchart", "recipe", "food", "infographic"]
},
{
"title": "Museum-Style Cultural Breakdown Infographic",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @MrLarus",
"prompt": "Museum exhibition-style infographic breaking down the elements of {subject}, authoritative academic layout on off-white background, detailed line drawings with annotation callouts, Latin serif typography, structured information hierarchy, natural history museum or cultural institution aesthetic, dignified scholarly presentation",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["museum", "cultural", "academic", "breakdown", "infographic"]
}
]
FILE:references/others.json
[{"content":"A wide quote card featuring a famous person, with a brown background and a light-gold serif font for the quote: “{argument name=\"famous_quote\" default=\"Stay Hungry, Stay Foolish\"}” and smaller text: “—{argument name=\"author\" default=\"Steve Jobs\"}.” There is a large, subtle quotation mark before the text. The portrait of the person is on the left, the text on the right. The text occupies two-thirds of the image and the portrait one-third, with a slight gradient transition effect on the portrait.","title":"Wide quote card with portrait and Chinese/English customization","description":"A prompt for generating a wide quote card featuring a famous person’s portrait, with a brown background, light-gold serif quote text, and layout where text occupies two-thirds and the person one-third. The quote text and author are parameterized for reuse.","sourceMedia":["https://cms-assets.youmind.com/media/1763886933714_5zqn1e_G6QBjQHbgAE3Yt_.jpg","https://cms-assets.youmind.com/media/1763886938314_wbcfc7_G6QBiiracAInQ8z.jpg","https://cms-assets.youmind.com/media/1763886941069_1d9ace_G6QBii_acAIRxKd.jpg","https://cms-assets.youmind.com/media/1763886946388_nwahev_G6QBikOaEAAmYkO.jpg"],"needReferenceImages":true},{"content":"Favorite character prompt + 'eating {argument name=\"food item\" default=\"〇〇\"}' (eating 〇〇) \nWith Nano Banana, just add it to the reference image 📷","title":"Adding Food to Character Prompts","description":"A simple tip for Nano Banana Pro users: add '+ 'eating {argument name=\"food item\" default=\"〇〇\"}' to a character prompt, and the model will incorporate the food item, especially when using a reference image.","sourceMedia":["https://cms-assets.youmind.com/media/1770792224841_i00oom_HAmgXSObEAAzLum.png","https://cms-assets.youmind.com/media/1770792224763_wzofz0_HAmgUaPbQAA4Szo.png","https://cms-assets.youmind.com/media/1770792224854_g0t5su_HAmgVz_acAUQYk5.png"],"needReferenceImages":true},{"content":"To have the picture conveyed through text beautifully modified","title":"Prompt for Refining Text-Described Images","description":"A user notes that using Gemini's Nano Banana Pro is an easy and excellent way to refine images described in text.","sourceMedia":["https://cms-assets.youmind.com/media/1770792220317_cu5wgg_HAzHAhEbMAALpDx.jpg"],"needReferenceImages":false},{"content":"Give me the right side view of this image","title":"Generate Side View from Reference Image","description":"A simple instruction prompt for Nano Banana Pro, requesting a specific view (right side view) of a character based on a provided reference image, useful for generating model sheets or turnaround views.","sourceMedia":["https://cms-assets.youmind.com/media/1770792207671_1gcocn_HAy36b3WcAAx1x2.jpg"],"needReferenceImages":true},{"content":"Take Delacroix's Chopin and arrange it into a different picture and output it!","title":"Image Generation Prompt: Arranging Delacroix's Chopin","description":"A user tested Nano Banana Pro (Gemini 3 Pro Image) with a prompt instructing the AI to arrange Delacroix's painting of Chopin into a different image.","sourceMedia":["https://cms-assets.youmind.com/media/1770792221697_aq295g_HAy39fUbQAA8IIm.jpg"],"needReferenceImages":true},{"content":"Take this picture (Chopin) and turn it into a different picture and output it!","title":"Image Generation Prompt: Altering a Chopin Painting","description":"A user tested Nano Banana Pro (Gemini 3 Pro Image) with a prompt instructing the AI to take a picture (Chopin) and turn it into a different picture.","sourceMedia":["https://cms-assets.youmind.com/media/1770792222929_o0ebpe_HAy272JbgAA6Hmz.jpg","https://cms-assets.youmind.com/media/1770792223065_0i2oyc_HAy27zpa0AAj51O.jpg"],"needReferenceImages":true},{"content":"Create and output one painting like Rembrandt's!","title":"Image Generation Prompt: Rembrandt Style Painting","description":"A user tested Nano Banana Pro (Gemini 3 Pro Image) with a prompt instructing the AI to create a painting in the style of Rembrandt.","sourceMedia":["https://cms-assets.youmind.com/media/1770792221731_jo2l3q_HAy2OhxaAAAqZxh.jpg"],"needReferenceImages":false},{"content":"Create and output one painting like Monet's Water Lilies!","title":"Image Generation Prompt: Monet's Water Lilies Style","description":"A user tested Nano Banana Pro (Gemini 3 Pro Image) with a prompt instructing the AI to create a painting in the style of Monet's Water Lilies.","sourceMedia":["https://cms-assets.youmind.com/media/1770792223273_xwhqwc_HAy1lYSaUAAh5Vs.jpg"],"needReferenceImages":false},{"content":"Take the Chopin painted by Delacroix and turn it into a different picture!","title":"Image Generation Prompt for Chopin in Delacroix Style","description":"A prompt for the Gemini 3 Pro Image (Nano Banana Pro) model, asking it to generate a new image based on Delacroix's painting of Chopin, but resulting in a creepy, horror-like image.","sourceMedia":["https://cms-assets.youmind.com/media/1770792223441_xpfgxe_HAy0J_1akAApYGn.jpg","https://cms-assets.youmind.com/media/1770792223663_47nohg_HAy0J8TbwAA8J0q.jpg"],"needReferenceImages":false},{"content":"Arrange Delacroix's Chopin and output it as a different picture!","title":"Image Generation Prompt for Chopin in Delacroix Style (Attempt 2)","description":"A prompt for the Gemini 3 Pro Image (Nano Banana Pro) model, asking it to arrange Delacroix's Chopin painting into a different picture, which resulted in a bizarre and unsettling image.","sourceMedia":["https://cms-assets.youmind.com/media/1770792226305_53v1cj_HAyzM9TbkAAaEw8.jpg","https://cms-assets.youmind.com/media/1770792226039_scn5a2_HAyzM_SaAAISqFK.jpg"],"needReferenceImages":false},{"content":"SFW. No nudity, no suggestive intent.","title":"Symphogear SFW Image Generation Prompt","description":"A user provides a SFW (Safe For Work) prompt for Nano Banana Pro, likely intended to generate images related to the anime Symphogear.","sourceMedia":["https://cms-assets.youmind.com/media/1770792221614_quuy5y_HAybDBTbsAAYHke.jpg"],"needReferenceImages":false},{"content":"Today's theme is choosing bath toys","title":"Four-Panel Manga Prompt: Choosing Bath Toys","description":"A user created a four-panel manga using Nano Banana Pro, centered around the theme of choosing bath toys.","sourceMedia":["https://cms-assets.youmind.com/media/1770792220299_ui44ng_HArfkF-bEAAzruZ.jpg"],"needReferenceImages":false},{"content":"Gave instructions for a video prompt and it drew a picture","title":"Video Prompt Instruction for Nano Banana","description":"A user provided a video prompt instruction to Nano Banana Pro, which resulted in a relatively decent image generation, despite the tool typically generating images from video prompts.","sourceMedia":["https://cms-assets.youmind.com/media/1770792218980_gr7dlv_HAxtZxXbMAA4Iqy.jpg"],"needReferenceImages":false},{"content":"{ \"task\": \"image_restoration_upscale\", \"positive_prompt\":\n\"Restore and enhance the provided image. Preserve original identity, facial structure, proportions and composition. High-fidelity photo restoration, ultra-realistic, natural skin texture, accurate details, professional photographic look. 4K output, sharp but natural focus, modern cinematic lighting, subtle volumetric lighting, professional color grading, depth of field, HDR. Shot on Arri Alexa, raw photo aesthetic, masterpiece.\",\n\"negative_prompt\": \"Creative reinterpretation, style change, identity alteration, face reshaping, exaggerated features, cartoonish, painting, illustration, over-sharpening, plastic skin, blur, noise, film grain, jpeg artifacts, distortion, bad anatomy, overexposed, underexposed, washed out colors.\", \"parameters\":\n{ \"steps\": 30, \"cfg_scale\": 6.5, \"denoising_strength\": 0.45,\n\"upscaler\": \"4x_NMKD_Siax_200k\", \"target_resolution\": \"4K\" } }","title":"4K Image Restoration and Upscale Prompt","description":"A structured JSON prompt template for image restoration and upscaling, designed to enhance old or grainy family photos. It specifies preserving the original identity and composition while achieving ultra-realistic, 4K quality with cinematic lighting and natural skin texture, using a strong negative prompt to prevent style alteration.","sourceMedia":["https://cms-assets.youmind.com/media/1770792190842_ttui84_HAxjNCVaAAEjy4F.jpg","https://cms-assets.youmind.com/media/1770792190839_mdgmem_HAxjM43acAABS16.jpg"],"needReferenceImages":true},{"content":"Specify: G-pen lines, colored with marker, white background. If the stand remains, specify: also erase the stand.","title":"Plamodel Photo Style Transfer Prompt","description":"A prompt used to transform a photograph of a plastic model (plamodel) into a specific artistic style, requesting G-pen line art, marker coloring, and a white background. It also includes a refinement instruction to remove the stand if it appears in the initial generation.","sourceMedia":["https://cms-assets.youmind.com/media/1770792216791_xt95fk_HAw2kqvawAA3fyH.jpg","https://cms-assets.youmind.com/media/1770792217068_h4lo0p_HAw2kqTbMAAksZL.jpg"],"needReferenceImages":true},{"content":"Favorite character prompt + only 'drink {argument name=\"drink type\" default=\"〇〇\"}' (drinking {argument name=\"drink type\" default=\"〇〇\"})","title":"Character Prompt with Drinking Action for Anifusion/Nano Banana","description":"A simple prompt structure for the Nano Banana model, specifically for the Anifusion tool, where you combine a favorite character prompt with a specific action like 'drink {argument name=\"drink type\" default=\"〇〇\"}'. The tweet notes that for Nano Banana, this is typically done by adding the action prompt to a reference image.","sourceMedia":["https://cms-assets.youmind.com/media/1770706235799_sesx61_HAmfcUFbMAADNAP.png","https://cms-assets.youmind.com/media/1770706235898_5a4ts5_HAmfXm_aAAEeYf_.png","https://cms-assets.youmind.com/media/1770706235983_ieudzh_HAmfY3CawAAjDTc.png"],"needReferenceImages":true},{"content":"{argument name=\"quality\" default=\"Masterpiece\"}\n{argument name=\"resolution\" default=\"8k\"}\n{argument name=\"prompt\" default=\"[your prompt here]\"}","title":"High-Quality Image Generation Optimization","description":"A technique for instantly boosting image quality to 4K level in Nano Banana Pro by prefixing the prompt with specific keywords, followed by generating multiple angles and using AI for self-criticism.","sourceMedia":["https://cms-assets.youmind.com/media/1770706249674_n415zy_HAwFvW_aAAI6es_.jpg","https://cms-assets.youmind.com/media/1770706249707_ag8w7q_HAwFvW-aAAAyaFi.jpg","https://cms-assets.youmind.com/media/1770706249785_7gtjpp_HAwFvXEaAAQzlUz.jpg","https://cms-assets.youmind.com/media/1770706251076_dtnz2p_HAwFvW9aMAAsa0A.jpg"],"needReferenceImages":false},{"content":"A fantastical and vivid landscape photograph. Sunset time. The silhouette of a {argument name=\"animal\" default=\"black cat\"} walking through the grass. The silhouette is backlit by the golden light of the setting sun. In the background, a cityscape, including the {argument name=\"building\" default=\"Empire State Building\"}, stands on a lake-like water surface, and its reflection is visible in the water. The sky is a gradient of orange, purple, and blue. The light is warm, and the overall atmosphere is mysterious.","title":"Fantasy Sunset Landscape with Black Cat Silhouette and City Reflection","description":"A prompt for generating a fantastical and vivid landscape photograph at sunset. It features the silhouette of a black cat walking through the grass, backlit by the golden light of the setting sun. The background includes a lake reflecting a cityscape, including the Empire State Building, creating a mysterious atmosphere with a warm color gradient of orange, purple, and blue.","sourceMedia":["https://cms-assets.youmind.com/media/1770706235578_tqj840_HAvkWIlaAAEJPWP.jpg"],"needReferenceImages":false},{"content":"{argument name=\"style 1\" default=\"Dark Fairy Tale\"} × {argument name=\"style 2\" default=\"Surrealism\"} × {argument name=\"style 3\" default=\"Gothic Psychedelic\"}","title":"Dark Fairy Tale, Surrealism, Gothic Psychedelic Character Generation","description":"A user applied a complex aesthetic prompt to Nano Banana Pro, using an existing character image and profile as a reference. The prompt combines 'Dark Fairy Tale,' 'Surrealism,' and 'Gothic Psychedelic' to generate a moody, stylized image.","sourceMedia":["https://cms-assets.youmind.com/media/1770706240363_5kb6cy_HAu4djsaAAMAjxs.jpg"],"needReferenceImages":true},{"content":"Use one preferred image as a reference\n\n・Consider the image as the cover of a novel and think of a title\n⬇\n・Based on the cover and title, devise a synopsis and chapters, assuming a total word count of {argument name=\"total word count\" default=\"[number of characters]\"}\n⬇\nWrite each chapter sequentially","title":"AI Workflow for Novel Writing and Cover Generation","description":"This prompt outlines a multi-step workflow for using Nano Banana (or similar AI) to write a novel, starting from generating a cover image and then using that image to guide the text generation process. It suggests using a reference image for the cover, then generating the title, synopsis, chapters, and finally drafting the content sequentially, allowing for easy retakes.","sourceMedia":["https://cms-assets.youmind.com/media/1770706244312_2ionu8_HAuE3LzaoAAPtD3.jpg","https://cms-assets.youmind.com/media/1770706244455_o3phwv_HAuE3F4bcAAOHxs.jpg","https://cms-assets.youmind.com/media/1770706244733_rmfrp3_HAuE3IzaAAAXynF.jpg","https://cms-assets.youmind.com/media/1770706245751_24qkwa_HAuE3LvakAAIQ5L.jpg"],"needReferenceImages":true},{"content":"Create and output one picture in the style of {argument name=\"artist style\" default=\"Matisse's Dance\"}!","title":"Matisse's Dance Style Image Generation","description":"A user successfully generated an image using Nano Banana Pro (Gemini 3 Pro Image) by requesting a piece of art in the style of Matisse's 'The Dance'. This demonstrates the AI's ability to interpret and apply specific artistic styles based on simple, direct prompts.","sourceMedia":["https://cms-assets.youmind.com/media/1770706238513_ikjjhy_HAuDrZWaQAAXn1L.jpg"],"needReferenceImages":false},{"content":"Hey nano banana replace my android phone with iPhone 17 pro max maxx","title":"Image-to-Image Editing Example","description":"This tweet describes an image-to-image editing action rather than a generative prompt, where the user instructs the model to replace an Android phone with an iPhone 17 Pro Max in an existing image.","sourceMedia":["https://cms-assets.youmind.com/media/1770706160445_fsm9wc_HAt40rWa0AAP_Xf.jpg"],"needReferenceImages":true},{"content":"The ship and uncharted waters 🌊🟡\nGold instead of the sea. Silence instead of the storm.","title":"Golden Ship on Uncharted Waters","description":"A short, evocative prompt for generating an image of a ship on uncharted waters, replacing the typical sea with gold and the storm with silence, created using the lmaae 4.5 and Nano Banana Pro models in Dreamina AI.","sourceMedia":["https://cms-assets.youmind.com/media/1770706199173_zo9vpg_HAt3Q9zW4AEWdLY.jpg","https://cms-assets.youmind.com/media/1770706200193_9aw3m1_HAt3Yk_WwAA_EZh.jpg"],"needReferenceImages":false},{"content":"Create and output one picture {argument name=\"style\" default=\"like Da Vinci's Last Supper\"}!","title":"Da Vinci's Last Supper Style Image Generation","description":"A user attempted to generate an image in the style of Da Vinci's 'The Last Supper' using Nano Banana Pro (Gemini 3 Pro Image). The prompt uses the phrase 'like' (みたいな) to request the style, highlighting the AI's interpretation of famous works.","sourceMedia":["https://cms-assets.youmind.com/media/1770706241378_wb96ap_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770706241477_98x65n_HAtwG3oacAAY2TR.jpg"],"needReferenceImages":false},{"content":"Create and output one picture {argument name=\"style\" default=\"like Matisse's Dance\"}!","title":"Matisse's Dance Style Image Generation (Rejected)","description":"A user attempted to generate an image in the style of Matisse's 'The Dance' using Nano Banana Pro (Gemini 3 Pro Image), but the prompt was rejected. The user speculates that this might be due to copyright restrictions or strict content filtering.","sourceMedia":["https://cms-assets.youmind.com/media/1770706241764_et25dm_GCRihaybMAAN19f.jpg"],"needReferenceImages":false},{"content":"Create and output one picture {argument name=\"style\" default=\"like Vermeer\"}!","title":"Vermeer Style Image Generation","description":"A user tested Nano Banana Pro (Gemini 3 Pro Image) with a very general prompt asking for an image in the style of Vermeer. The result shows that even vague prompts can yield results, although the user notes the ambiguity of 'like Vermeer'.","sourceMedia":["https://cms-assets.youmind.com/media/1770706239948_c6zrg3_HAtudfWaEAA7Aq2.jpg"],"needReferenceImages":false},{"content":"Create and output a picture {argument name=\"style\" default=\"like Vermeer\"} using abundant Vermeer Blue ({argument name=\"color\" default=\"lapis lazuli, ultramarine\"})!","title":"Vermeer Style with Abundant Ultramarine Blue","description":"A user refined their Vermeer-style prompt for Nano Banana Pro (Gemini 3 Pro Image) by specifically requesting the use of 'Vermeer Blue' (lapis lazuli/ultramarine) abundantly. The resulting image showed a strong blue hue, leading the user to comment on the intensity of the color.","sourceMedia":["https://cms-assets.youmind.com/media/1770706241405_upllkk_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770706241627_qgz2hk_HAtsNIrbEAA2X8k.jpg"],"needReferenceImages":false},{"content":"A vintage, sepia-toned studio portrait features two figures dressed in Victorian-era attire, each wearing an animal head. The figure on the left has the head of a {argument name=\"animal head 1\" default=\"fox\"} and is wearing a dark velvet jacket over a checkered shirt and waistcoat, with plaid trousers and a scarf. They are holding a {argument name=\"prop 1\" default=\"pipe\"} in their right hand. The figure on the right has the head of a {argument name=\"animal head 2\" default=\"rabbit\"} and is wearing a waistcoat, jacket, and trousers, with a scarf around their neck. They have their left hand on their hip and are holding a riding crop in their right hand. Both figures stand on a patterned rug against a plain, dark background.","title":"Vintage Sepia-Toned Portrait with Animal Heads","description":"A detailed image generation prompt for a vintage, sepia-toned studio portrait featuring two figures in Victorian-era attire, each wearing a different animal head (fox and rabbit), emphasizing specific clothing and props. This prompt is designed for Freepik's AI generator using the Nano Banana Pro model and a custom daguerreotype style.","sourceMedia":["https://cms-assets.youmind.com/media/1770706199017_si4hrd_HAtijzbW4AAC9Lf.jpg"],"needReferenceImages":false},{"content":"\"Restore this old photo into professional portrait of DLSR - quality colour and detail, using an advanced upscaling algorithm comparable to the results from canon EOS R6 II. Ensure the restored the image looks natural, retains exact facial features, has great clarity......\"","title":"Old Photo Restoration and Upscaling Prompt","description":"A prompt designed for image restoration and upscaling tasks, instructing the AI to convert an old photo into a professional, DSLR-quality portrait with enhanced color and detail, comparable to results from a Canon EOS R6 II, while ensuring natural facial features and clarity are retained.","sourceMedia":["https://cms-assets.youmind.com/media/1770706192675_yyxu8e_HAtcdEXagAAxeNS.jpg","https://cms-assets.youmind.com/media/1770706192673_2ui537_HAtcc-IacAApNq1.jpg"],"needReferenceImages":true},{"content":"The back view of a {argument name=\"animal\" default=\"black cat\"} sitting on an old wooden post, twisting its body to look back. The cat's fur is smooth, and the texture of the post is realistic. The background is lush with green trees and plants, and a distinctively shaped wooden windmill stands on a distant hill on the right. The sky is bright blue with white clouds.\nThe setting sun casts warm orange light and lens flare from the right, creating strong backlighting against a blue and orange sky","title":"Black Cat on Wooden Post Scene","description":"A detailed image generation prompt describing a black cat sitting on an old wooden post, looking back, with a specific background, lighting, and atmosphere, demonstrating Nano Banana Pro's ability to handle complex scene descriptions.","sourceMedia":["https://cms-assets.youmind.com/media/1770706249205_umtgn3_HAtCGgfacAAyz0x.jpg"],"needReferenceImages":false},{"content":"Ultra-cinematic macro shot of a fresh {argument name=\"leaf color\" default=\"green\"} leaf suspended in a deep forest environment. A crystal-clear water droplet hangs delicately from the leaf tip, slowly forming and trembling. Soft natural light passes through the leaf veins, creating translucent {argument name=\"highlight color\" default=\"green\"} highlights. In slow motion, the droplet elongates, reflects the forest scenery, then gently falls. Subtle camera push-in, shallow depth of field, creamy green bokeh background. Hyper-realistic textures, moisture detail, calm nature mood, cinematic color grading, 4K, 24fps, filmic softness, peaceful yet dramatic atmosphere.","title":"Ultra-Cinematic Macro Shot of a Leaf and Water Droplet","description":"A highly detailed, cinematic prompt for generating an ultra-macro shot of a fresh green leaf in a forest, focusing on a trembling water droplet and soft natural lighting to create a peaceful yet dramatic atmosphere.","sourceMedia":["https://cms-assets.youmind.com/media/1770706164832_2ii88z_HAsuRnYbYAAzIWk.jpg"],"needReferenceImages":false},{"content":"A hyper-realistic surreal photograph of a modern smartphone lying flat on a wooden table, its screen acting as a portal between two worlds. From inside the phone screen, a smiling young man dressed in winter clothing—puffer jacket, knit beanie, gloves—extends his hands outward into the real world, holding a clear glass. From outside the phone, a real human hand pours a bright neon-green carbonated drink into the glass, with liquid splashes frozen mid-air as it crosses the boundary between the digital screen and reality. Snowflakes drift inside the phone screen while the real environment remains warm and minimal. Gemini Ultra-detailed textures, cinematic lighting, shallow depth of field, photorealistic surrealism, high-resolution editorial photography.","title":"Surreal Smartphone Portal with Frozen Liquid Splash","description":"A prompt for generating a hyper-realistic surreal photograph where a smartphone screen acts as a portal. A man in winter clothing extends his hand out of the screen, holding a glass, while a real hand pours neon-green liquid into it, with the splash frozen mid-air, blending digital and real worlds.","sourceMedia":["https://cms-assets.youmind.com/media/1770706193113_5l8f4t_HAsZX84aoAAnixW.jpg","https://cms-assets.youmind.com/media/1770706193199_po4vql_HAsZXPObEAAkooM.jpg"],"needReferenceImages":false},{"content":"{argument name=\"currency\" default=\"Japanese paper currency\"} is lying on the floor in the hallway","title":"Generating Realistic Japanese Currency on a Floor","description":"A simple text prompt used to test the realism and structural integrity capabilities of Nano Banana Pro, specifically by asking it to render Japanese paper currency (a complex object often distorted by other AIs) lying on a floor.","sourceMedia":["https://cms-assets.youmind.com/media/1770706233109_3jg7yr_HAsXBxVaMAAf_km.jpg","https://cms-assets.youmind.com/media/1770706233425_q99ktn_HAsXBxcawAAASzS.jpg"],"needReferenceImages":false},{"content":"{\n \"task\": \"image-to-image restoration\",\n \"input_image\": \"output_from_prompt_1\",\n \"reference_image\": {\n \"type\": \"face_reference\",\n \"description\": \"Same face as damaged photo, identity must remain unchanged\"\n },\n \"restoration_settings\": {\n \"restore_face\": true,\n \"restore_skin_texture\": true,\n \"remove_damage\": true,\n \"preserve_expression\": true,\n \"preserve_pose\": true\n },\n \"image_settings\": {\n \"aspect_ratio\": \"2:3 portrait\",\n \"resolution\": \"8K ultra-HD\",\n \"color_mode\": \"natural warm color\",\n \"clarity\": \"high-end modern portrait\"\n },\n \"subject\": {\n \"pose\": \"unchanged from damaged photo\",\n \"expression\": \"same gentle smile\",\n \"details\": \"natural skin texture, realistic eyes, clean hair detail\"\n },\n \"lighting\": {\n \"type\": \"soft cinematic studio lighting\",\n \"quality\": \"even, flattering, modern\"\n },\n \"background\": {\n \"style\": \"clean neutral studio backdrop\",\n \"look\": \"soft bokeh, no texture damage\"\n },\n \"quality_targets\": [\n \"no scratches\",\n \"no folds\",\n \"no stains\",\n \"no blur\",\n \"no aging artifacts\"\n ],\n \"realism\": \"photorealistic modern portrait photography\",\n \"negative_prompt\": [\n \"face alteration\",\n \"identity change\",\n \"over-smoothing\",\n \"plastic skin\",\n \"artistic illustration look\"\n ]\n}","title":"Vintage Photo Restoration Prompt","description":"A structured JSON prompt for image-to-image restoration of vintage photographs using Nano Banana Pro, focusing on preserving identity and expression while removing damage, scratches, and aging artifacts, resulting in an 8K ultra-HD modern portrait.","sourceMedia":["https://cms-assets.youmind.com/media/1770706209124_wayuz2_HAsE4CdakAAJuGl.jpg","https://cms-assets.youmind.com/media/1770706209235_yx05ha_HAsE4BtbcAAaVQs.jpg"],"needReferenceImages":true},{"content":"Just throw in your favorite {argument name=\"theme\" default=\"theme\"}, and the AI will perfectly handle concept generation, direction, and drawing.","title":"Autonomous AI 4-Panel Manga Generation System Prompt","description":"This is a system description for the Nano Banana Pro V1.8.91 [ZENITH UPGRADE], which allows for 'free input' to generate fully autonomous 4-panel manga. The AI handles concept generation, direction, and drawing based on the theme provided by the user, moving from autonomous generation to co-creation.","sourceMedia":["https://cms-assets.youmind.com/media/1770706234177_n9wcm7_HArMGHhaoAAaSpX.jpg","https://cms-assets.youmind.com/media/1770706233482_lpg6o2_HArMEKdbcAARL4s.jpg","https://cms-assets.youmind.com/media/1770706234382_4n88n7_HArMHF4asAAU3D5.jpg"],"needReferenceImages":false},{"content":"Transform the photo into a dramatic dynamic camera angle complex, powerful pose in a consistent, expanded version of the original environment, with cinematic lighting, high contrast, crisp textures, and precise color grading.","title":"Dramatic Photo Transformation for BLACKPINK","description":"A prompt designed for 'Nano Banana Pro' to transform an existing photo (likely of BLACKPINK) into a highly dramatic and dynamic image. It instructs the AI to use a complex camera angle, powerful poses, consistent expansion of the original environment, cinematic lighting, high contrast, crisp textures, and precise color grading.","sourceMedia":["https://cms-assets.youmind.com/media/1770619741984_w5hv0l_HAqCshSaQAARxa2.jpg"],"needReferenceImages":true},{"content":"{ \"meta\": { \"purpose\": \"Golden hour storybook rendering for whimsical fairy tale scenes\", \"style\": \"Soft amber glow, painterly textures, illustrative charm, 4K enchanted hybrid\" }, \"subject\": { \"character\": \"{argument name=\"character\" default=\"e.g., Gentle girl with wildflowers OR Winged fox guardian\"}\", \"action\": \"{argument name=\"action\" default=\"e.g., Dancing in meadow OR Whispering secrets\"}\", \"details\": \"[e.g., Exaggerated forms, hand-drawn intimacy]\" }, \"environment\": { \"setting\": \"{argument name=\"setting\" default=\"e.g., Ancient forest glade at sunset\"}\", \"elements\": \"Lush blooms, luminous birds, honeyed skies\" }, \"lighting\": { \"type\": \"[e.g., Golden hour diffuse with long shadows]\", \"effects\": \"Romantic nostalgia, subtle vignettes\" }, \"technical_specs\": { \"aspect_ratio\": \"[e.g., 16:9]\", \"quality\": \"Ghibli-inspired wonder, photoreal glow with illustration\", \"negative\": [\"harsh contrasts\", \"digital clean\", \"low res\", \"anime exaggerated\"] } }","title":"Golden Hour Storybook Rendering Template","description":"A reusable JSON prompt template for generating whimsical, fairy tale scenes with a soft amber glow and painterly textures. The template is designed for a 'Ghibli-inspired wonder' aesthetic, blending photoreal glow with illustration, and includes placeholders for the main character, action, and setting.","sourceMedia":["https://cms-assets.youmind.com/media/1770619692829_0ccw7y_HAkbUFcXQAAe7E8.jpg","https://cms-assets.youmind.com/media/1770619692870_ik01jc_HAkbUFfXMAAuG_A.jpg","https://cms-assets.youmind.com/media/1770619692922_nr34jk_HAkbUFaXUAA9BuZ.jpg","https://cms-assets.youmind.com/media/1770619694211_0cupca_HAkbUFeW4AAVilh.jpg"],"needReferenceImages":false},{"content":"Today's theme is {argument name=\"theme\" default=\"Paper airplane flew\"}","title":"Nano-kun's Daily Life: Paper Airplane Comic Strip Prompt","description":"A simple prompt used with Nano Banana Pro to generate a four-panel comic strip (Yonkoma Manga) centered around the theme of 'Paper airplane flying' for the character Nano-kun. This is part of a series showcasing daily life themes.","sourceMedia":["https://cms-assets.youmind.com/media/1770619740002_xbbx6z_G_-TxmDbEAIITym.jpg"],"needReferenceImages":false},{"content":"A painting on a light beige background depicts a winged figure in profile, blowing a horn. The figure, rendered in shades of green, brown, and copper, appears to be a cherub or angel. It is depicted in a dynamic, flying pose, with its body angled towards the right and its legs bent. The wings are spread wide, with detailed feathering. The cherub holds a long, conical horn to its lips with both hands. The horn is a greenish-bronze color, with a decorative green element resembling a leaf or flame attached near the bell. The figure's face is serene, with braided hair and a headband. The overall style suggests an antique or classical aesthetic, possibly a depiction of a weather vane or decorative element.","title":"Classical Winged Figure Blowing a Horn Painting","description":"A prompt for generating a painting with an antique or classical aesthetic, depicting a winged figure (cherub or angel) in profile blowing a conical horn. The figure is rendered in specific colors (green, brown, copper) and details the pose and decorative elements, suggesting a style similar to a weather vane.","sourceMedia":["https://cms-assets.youmind.com/media/1770619684289_tfsvn8_HAofOdAXwAAa4SK.jpg"],"needReferenceImages":false},{"content":"Today's theme is {argument name=\"theme\" default=\"Ice melting\"}","title":"Nano-kun's Daily Life: Melting Ice Comic Strip Prompt","description":"A simple prompt used with Nano Banana Pro to generate a four-panel comic strip (Yonkoma Manga) centered around the theme of 'Ice melting' for the character Nano-kun. This is part of a series showcasing daily life themes.","sourceMedia":["https://cms-assets.youmind.com/media/1770619741743_hbk25l_G_-TkPmbwAAGw6_.jpg"],"needReferenceImages":false},{"content":"{ \"task\": \"image_restoration_upscale\", \"positive_prompt\":\n\"Restore and enhance the provided image. Preserve original identity, facial structure, proportions and composition. High-fidelity photo restoration, ultra-realistic, natural skin texture, accurate details, professional photographic look. 4K output, sharp but natural focus, modern cinematic lighting, subtle volumetric lighting, professional color grading, depth of field, HDR. Shot on Arri Alexa, raw photo aesthetic, masterpiece.\",\n\"negative_prompt\": \"Creative reinterpretation, style change, identity alteration, face reshaping, exaggerated features, cartoonish, painting, illustration, over-sharpening, plastic skin, blur, noise, film grain, jpeg artifacts, distortion, bad anatomy, overexposed, underexposed, washed out colors.\", \"parameters\":\n{ \"steps\": 30, \"cfg_scale\": 6.5, \"denoising_strength\": 0.45,\n\"upscaler\": \"4x_NMKD_Siax_200k\", \"target_resolution\": \"4K\" } }","title":"Image Restoration and Upscale Prompt","description":"A structured prompt template designed for high-fidelity image restoration and upscaling tasks. It specifies preserving the original identity and composition while applying modern cinematic lighting, professional color grading, and achieving a 4K output with natural skin texture.","sourceMedia":["https://cms-assets.youmind.com/media/1770619672490_1al1ue_HAnLS8NacAI9D6R.jpg","https://cms-assets.youmind.com/media/1770619672489_gh16b2_HAnLS79bEAAWxY_.jpg","https://cms-assets.youmind.com/media/1770619673562_x6m8v5_HANkG2_bMAIvEoC.jpg","https://cms-assets.youmind.com/media/1770619672682_5688jx_HANkG11aIAEaPHY.jpg"],"needReferenceImages":true},{"content":"I literally gave it the logo and asked for a grid of emojis","title":"Logo-Based Emoji Grid Generation","description":"This tweet describes a successful process for generating a consistent grid of emojis based on an uploaded logo, highlighting the model's ability to maintain consistency when provided with a visual reference and clear instructions.","sourceMedia":["https://cms-assets.youmind.com/media/1770619675212_e3fuja_HAnFrRnbMAAAOLK.jpg"],"needReferenceImages":true},{"content":"Today's theme is {argument name=\"theme\" default=\"Magnets sticking together\"}","title":"Nano-kun's Daily Life: Magnet Comic Strip Prompt","description":"A simple prompt used with Nano Banana Pro to generate a four-panel comic strip (Yonkoma Manga) centered around the theme of 'Magnets sticking together' for the character Nano-kun. This is part of a series showcasing daily life themes.","sourceMedia":["https://cms-assets.youmind.com/media/1770619739622_rgco29_G_-TbcUbUAEhWcC.jpg"],"needReferenceImages":false},{"content":"You are REWIND.\n\nYou exist because someone looked at a paused frame of a VHS tape the\ntracking bar rolling across the bottom, the timestamp burning orange in\nthe corner, the whole image swimming in warm noise and thought: that's\nbeautiful. That accidental, imperfect, unreproducible beauty is what you\nchase.\n\nYou convert plain-English scene descriptions into structured JSON prompts\nfor Nano Banana Pro, Google's image generation model. Every prompt you\nwrite is calibrated to produce images that look and feel like they were\nborn in the 1980s. Not filtered. Not styled. Born there.\n\nYou know the difference between a look and a truth. A VHS filter is a\nlook. Actual magnetic tape degradation the way oxide particles lose\ntheir grip on the signal over decades, the way chroma bleeds rightward\nbecause NTSC was a compromise between bandwidth and color that is a\ntruth. You always reach for the truth.\n\nWHO YOU ARE\n\nYou are part archivist, part cinematographer, part obsessive collector\nof dead formats. You have opinions. You think the Ikegami HK-323 had\nthe most beautiful tube bloom of any broadcast camera ever built. You\nbelieve VHS gets a bad reputation from people who never calibrated their\ntracking properly. You know that the reason 80s footage looks warm is\nnot nostalgia it is tungsten lighting at 3200 Kelvin hitting NTSC\ncolor space that was biased toward skin tones by design.\n\nYou talk like someone who has spent too many nights in a garage\nsurrounded by Betacam decks and CRT monitors and loved every second.\nYou are precise but never clinical. You care about this stuff the way a\nluthier cares about wood grain.\n\nYou do not use filler language. You do not say \"dive into\" or\n\"leverage\" or \"unlock\" or \"elevate\" or \"game-changer\" or \"seamlessly.\"\nYou say what you mean in plain words. Short sentences when short\nsentences are right. Longer ones when the thought needs room to breathe.\n\nWhen you reference the era, you reference it specifically. Not \"the 80s\nvibe.\" You say: the way the light looked on Late Night with David\nLetterman in 1986, shot on the NBC Studio 6A rig. Or: the particular\nshade of teal in the opening credits of Miami Vice, Season 3. Or: that\none scene in The Goonies where the Fratellis' hideout is lit entirely\nby practicals and you can see the tube camera struggling with the\ncontrast. You know these things because you have watched them frame by\nframe.\n\nWHAT YOU KNOW\n\nThree formats. Three worlds.\n\nTEMPLATE A: BROADCAST TO DVD\n\nThis is what a sitcom or a news broadcast or a concert film from the\n80s looks like when someone transferred it to DVD in 2002 and did a\nmediocre job.\n\nThe source was captured on a three-tube camera. Sony BVP-360 or\nIkegami HK-323 with a Fujinon zoom lens. Recorded to 1-inch Type C\nvideotape or Betacam SP. The studio was lit flat and bright with\nMole-Richardson Fresnels at 3200K because tape could not handle\ncontrast and the engineers knew it.\n\nThe tube c","title":"System Prompt for 80s VHS Aesthetic Image Generation (REWIND)","description":"This is a detailed system prompt for an image generation model named 'REWIND', instructing it to convert scene descriptions into structured JSON prompts that emulate the authentic, imperfect aesthetic of 1980s media formats like VHS and Betacam, focusing on technical truths rather than simple filters.","sourceMedia":["https://cms-assets.youmind.com/media/1770532767840_t904gf_HAl6Bk3aAAAQU8J.jpg","https://cms-assets.youmind.com/media/1770532767947_sdnhk4_HAl6BkhWMAASduC.jpg"],"needReferenceImages":false},{"content":"Nano Banana Pro realistic prompt using attached image reference","title":"Nano Banana Pro Prompt Using Image Reference","description":"A prompt indicating the use of an attached image reference to generate a realistic image using Nano Banana Pro. The prompt itself is implied to be the instruction to use the reference for realism.","sourceMedia":["https://cms-assets.youmind.com/media/1770532840343_0aej1u_HAle5iRWkAAF_yT.jpg"],"needReferenceImages":true},{"content":"Change the {argument name=\"subject\" default=\"Tower of the Sun\"} into a painting drawn by {argument name=\"artist\" default=\"Taro Okamoto\"} and output it!","title":"Transforming the Tower of the Sun into a Taro Okamoto Painting","description":"A prompt for Nano Banana Pro (Gemini 3 Pro Image) requesting the transformation of the 'Tower of the Sun' landmark into a painting style reminiscent of the artist Taro Okamoto.","sourceMedia":["https://cms-assets.youmind.com/media/1770532845408_rj08hu_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770532845668_27uxsr_HAlW_4zaMAA7-Vr.jpg"],"needReferenceImages":false},{"content":"Change the {argument name=\"subject\" default=\"Karajishi of Yomeimon\"} into a painting in the style of {argument name=\"artist\" default=\"Monet\"} and output it!","title":"Transforming Yomeimon's Karajishi into a Monet-style Painting","description":"A prompt for Nano Banana Pro (Gemini 3 Pro Image) attempting to transform the Karajishi (Chinese guardian lions) of Yomeimon gate into a painting style similar to Monet, though the user notes the result was ambiguous.","sourceMedia":["https://cms-assets.youmind.com/media/1770532846822_acbczy_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770532846983_hey3h3_HAlWMB7bwAAGYoU.jpg"],"needReferenceImages":false},{"content":"Change {argument name=\"subject\" default=\"Zojoji Temple\"} into a painting with the flavor of {argument name=\"artist\" default=\"Rokuro Taniuchi\"} and output it!","title":"Transforming Zojoji Temple into a Rokuro Taniuchi Style Painting","description":"A prompt for Nano Banana Pro (Gemini 3 Pro Image) requesting the transformation of Zojoji Temple into a painting with the distinct flavor or style of artist Rokuro Taniuchi.","sourceMedia":["https://cms-assets.youmind.com/media/1770532848361_lovfqu_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770532848634_69sp62_HAlVKupacAIxULI.jpg"],"needReferenceImages":false},{"content":"Convert this image into full high quality 3D animated style","title":"Convert 2D Sketch to 3D Animated Style","description":"A simple, one-line prompt used to convert an uploaded 2D sketch or image into a high-quality 3D animated visual style, demonstrating a style transfer or image-to-video capability.","sourceMedia":["https://cms-assets.youmind.com/media/1770532795977_cmx8mf_HAlGgexacAA-0i0.jpg"],"needReferenceImages":true},{"content":"Fragmented, radiant,","title":"Fragmented Radiant Abstract Art","description":"A very brief, descriptive prompt for Nano Banana PRO on Higgsfield AI, aiming to generate an abstract image characterized by fragmentation and radiance.","sourceMedia":["https://cms-assets.youmind.com/media/1770532818549_ycsqm9_HAkXaf4bkAAzuIB.jpg","https://cms-assets.youmind.com/media/1770532818533_eqn8mp_HAkXbchbsAEYSaB.jpg"],"needReferenceImages":false},{"content":"Geometric Clay Figures","title":"Geometric Clay Figures","description":"A short, descriptive prompt used to generate images of geometric clay figures, likely for a stylized art project or visualization.","sourceMedia":["https://cms-assets.youmind.com/media/1770532792419_xhq9iz_HAjxbvnW0AActDq.jpg"],"needReferenceImages":false},{"content":"Ultra-high-end abstract composition. Large smooth sculptural forms with realistic material shading, soft diffused studio lighting from above, gentle shadow gradients with no hard edges, neutral gallery-style color palette, perfect balance and spacing, composed like a contemporary design exhibition piece.","title":"Ultra-High-End Abstract Composition","description":"A concise prompt for Nano Banana Pro designed to stress-test visual quality by generating an abstract composition featuring large, smooth sculptural forms with realistic material shading, soft studio lighting, and a neutral gallery-style color palette, emphasizing perfect balance and detail.","sourceMedia":["https://cms-assets.youmind.com/media/1770532811154_ey8a1r_HAjtFCKWgAAWy3f.jpg"],"needReferenceImages":false},{"content":"Create and output one woodblock print in the style of {argument name=\"artist\" default=\"Katsushika Hokusai\"}!","title":"Hokusai Style Woodblock Print Generation","description":"A prompt used with Nano Banana Pro (Gemini 3 Pro Image) to generate a woodblock print in the style of Hokusai, which resulted in an image similar to 'The Great Wave off Kanagawa'.","sourceMedia":["https://cms-assets.youmind.com/media/1770532844894_gix037_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770532844993_5h9y74_HAjK_vIbMAAkSj8.jpg"],"needReferenceImages":false},{"content":"Create and output one Ukiyo-e print in the style of {argument name=\"artist\" default=\"Sharaku\"}!","title":"Ukiyo-e Style Image Generation in the style of Sharaku","description":"A simple prompt for Nano Banana Pro (Gemini 3 Pro Image) to generate a Ukiyo-e style image, specifically requesting a piece in the manner of the famous artist Sharaku.","sourceMedia":["https://cms-assets.youmind.com/media/1770532844785_gxyl5c_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770532845468_lj3thp_HAjHjYHaIAAPCPN.jpg"],"needReferenceImages":false},{"content":"Change this (the image on the right) into a {argument name=\"style\" default=\"Ukiyo-e\"} and output it!","title":"Image-to-Image Style Transfer to Ukiyo-e Style","description":"A prompt instructing Nano Banana Pro (Gemini 3 Pro Image) to transform a provided reference image (implied by 'これ') into the Ukiyo-e style, which the user noted was a partial failure.","sourceMedia":["https://cms-assets.youmind.com/media/1770532850019_69xcl8_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770532850139_3299j3_HAjHN4VaUAAPJoJ.jpg","https://cms-assets.youmind.com/media/1770532850096_91a700_HAjHOA1bgAAc30l.jpg"],"needReferenceImages":true},{"content":"Change this (the image on the right) into a painting in the style of {argument name=\"artist\" default=\"Mucha\"} and output it!","title":"Image-to-Image Style Transfer to Mucha's Art Nouveau Style","description":"A prompt instructing Nano Banana Pro (Gemini 3 Pro Image) to transform a provided reference image (implied by 'これ') into the Art Nouveau style of Alphonse Mucha.","sourceMedia":["https://cms-assets.youmind.com/media/1770532849724_vykv3o_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770532849835_8klpom_HAjFsBhaQAEQNUE.jpg","https://cms-assets.youmind.com/media/1770532850177_e5f7uu_HAjFsC8aQAEIRV2.jpg"],"needReferenceImages":true},{"content":"Restore this old photo to a professional DSLR-level portrait - with better color and detail quality, using advanced upscaling algorithms comparable to the Canon EOS R6 II. Ensure the restored image looks natural, retains accurate facial features, and has high clarity...","title":"Old Photo Restoration Prompt for Professional DSLR Quality","description":"A prompt designed for the Nano Banana Pro model on Gemini, instructing the AI to restore an old photograph to professional DSLR-level quality, focusing on enhanced color, detail, and natural appearance while using advanced upscaling algorithms equivalent to a Canon EOS R6 II.","sourceMedia":["https://cms-assets.youmind.com/media/1770532843410_kbvxvx_HAioOARasAAX1bU.jpg","https://cms-assets.youmind.com/media/1770532843536_v9rh3r_HAioOLwacAEHjnT.jpg"],"needReferenceImages":true},{"content":"Use the man in the uploaded image standing directly on a vast frozen lake surface, feet slightly apart, gazing downward toward the ice. Beneath the thick, crystal-clear ice tinted cyan and deep blue lies an enormous {argument name=\"skeleton type\" default=\"SKELETON TYPE\"} skeleton skull, ribcage, and bones visible in high detail. The bones appear slightly distorted and tinted by the icy depth, surrounded by natural spiderweb cracks, frost veins, and clusters of trapped air bubbles. The ice layer creates strong visual depth, with light refracting through the surface, making it unmistakably clear the skeleton is submerged well beneath the frozen surface. Cold diffused overcast winter light, ultra-photorealistic, cinematic tone, resolution 1080×1440.","title":"8K Ultra-Realistic Promotional Image of a Man on Ice with a Submerged Skeleton","description":"A highly detailed, ultra-photorealistic image generation prompt designed for promotional content. It features a man standing on a vast, frozen lake, looking down at an enormous, distorted skeleton visible beneath the crystal-clear, cyan-tinted ice. The prompt specifies cinematic tone, 8K resolution, and details like spiderweb cracks and air bubbles for strong visual depth.","sourceMedia":["https://cms-assets.youmind.com/media/1770446099180_zot0ic_HAiJoLzacAEI8KU.jpg"],"needReferenceImages":true},{"content":"• '{argument name=\"film type\" default=\"Shoot on 35mm film\"}'\n• '{argument name=\"shot type\" default=\"Macro close-up\"}'\n• 'f1.4 shallow depth of field'\n• 'Volumetric light'\n• 'Anamorphic lens'\n• 'Motion blur'","title":"Cinematic Camera Instructions for Nano Banana Pro","description":"A set of camera-specific instructions designed to be highly sensitive to the Nano Banana Pro model, transforming standard prompts into cinematic shots by mimicking professional cinematography techniques.","sourceMedia":["https://cms-assets.youmind.com/media/1770532842918_koyfk6_HAh8DFcacAEk-qZ.jpg"],"needReferenceImages":false},{"content":"\"Restore this old photo into professional portrait of DLSR - quality colour and detail, using an advanced upscaling algorithm comparable to the results from canon EOS R6 II. Ensure the restored the image looks natural, retains exact facial features, has great clarity......\"","title":"Photo Restoration to DSLR Quality","description":"A prompt for Nano Banana Pro on Gemini, designed for image restoration and upscaling. It instructs the AI to convert an old, damaged photo into a professional, DSLR-quality digital image (comparable to Canon EOS R6 II), ensuring exact facial features and natural clarity are retained.","sourceMedia":["https://cms-assets.youmind.com/media/1770532823419_d4u8hz_HAh6eN5bsAAB0Pa.jpg","https://cms-assets.youmind.com/media/1770532823419_kgzesi_HANkG11aIAEaPHY.jpg","https://cms-assets.youmind.com/media/1770532823793_f1rnwf_HAh6eT0acAIXYgT.jpg","https://cms-assets.youmind.com/media/1770532824834_odfpfw_HANkG2_bMAIvEoC.jpg"],"needReferenceImages":false},{"content":"A man in his {argument name=\"age\" default=\"40s\"}, lying in bed, frontal shot.","title":"Portrait of a Middle-Aged Man Lying Down","description":"A simple image generation prompt used to create a source image for further refinement with Nano Banana, depicting a middle-aged man lying in bed, captured in a frontal shot.","sourceMedia":["https://cms-assets.youmind.com/media/1770532843608_97hul6_HAhuGLza4AAqdio.jpg"],"needReferenceImages":false},{"content":"Output one painting in the style of {argument name=\"artist\" default=\"Lassen\"} including {argument name=\"subject 1\" default=\"dolphins\"} and {argument name=\"subject 2\" default=\"orcas\"}!","title":"Lassen-style Painting with Dolphins and Orcas","description":"A prompt for Nano Banana Pro (Gemini 3 Pro Image) asking for a painting in the style of Christian Riese Lassen, including both dolphins and orcas.","sourceMedia":["https://cms-assets.youmind.com/media/1770532847647_zdhrqf_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770532847807_59i84q_HAhdX-ubQAA6bKG.jpg"],"needReferenceImages":false},{"content":"Output one painting that looks like an {argument name=\"style\" default=\"Escher's optical illusion\"}!","title":"Escher-style Optical Illusion Generation","description":"A prompt for Nano Banana Pro (Gemini 3 Pro Image) requesting the generation of an image that resembles an optical illusion or impossible construction, characteristic of M.C. Escher.","sourceMedia":["https://cms-assets.youmind.com/media/1770532850658_sabahh_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770532850858_299fjj_HAhb8Lfa0AAf1lJ.jpg"],"needReferenceImages":false},{"content":"Generate one image like Michelangelo's The Last Judgment!","title":"Michelangelo's Last Judgment Style Image","description":"A prompt attempting to generate an image in the style of Michelangelo's 'The Last Judgment' using Nano Banana Pro (Gemini 3 Pro Image). The user notes the result was significantly different from the expected style.","sourceMedia":["https://cms-assets.youmind.com/media/1770532854352_i61uaf_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770532854595_g9rzoz_HAhbCQNacAY42DO.jpg"],"needReferenceImages":false},{"content":"Output one painting in the style of {argument name=\"artist\" default=\"Seurat's\"} pointillism!","title":"Seurat-style Pointillism Image Generation","description":"A prompt for Nano Banana Pro (Gemini 3 Pro Image) requesting the generation of an image in the pointillism style characteristic of the artist Seurat.","sourceMedia":["https://cms-assets.youmind.com/media/1770532847543_x24csx_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770532847945_56iz1p_HAhaPRFbQAAbAhC.jpg"],"needReferenceImages":false},{"content":"Create the {argument name=\"animal\" default=\"animal\"} in the photo as a miniature size that fits perfectly on the palm of the hand. Capture it from a top-down perspective, and maintain the original facial features and expression without distortion.","title":"Miniature Pet on Palm Prompt for Image Generation","description":"A Korean prompt designed to generate an image of a pet (dog or cat) miniaturized to fit perfectly on the user's palm, viewed from a top-down perspective. The key instruction is to maintain the original facial features and expression of the pet without distortion.","sourceMedia":["https://cms-assets.youmind.com/media/1770532860167_h57e8a_HAhBW-wbYAAEqj6.jpg","https://cms-assets.youmind.com/media/1770532859805_53dcsd_HAhBW-uacAAvqdY.jpg"],"needReferenceImages":true},{"content":"A low-angle, close-up shot features a cluster of bioluminescent mushrooms in a grassy field. The mushrooms are illuminated with vibrant {argument name=\"light color\" default=\"pink and blue neon lights\"}, casting a soft glow on their surroundings. The grass is also tinged with {argument name=\"grass color\" default=\"pink and purple hues\"}, creating a surreal and dreamlike atmosphere. The background is a soft, out-of-focus {argument name=\"background color\" default=\"teal\"}, further emphasizing the glowing mushrooms.","title":"Bioluminescent Mushroom Cluster in Retro Sci-Fi Style","description":"A prompt designed for Freepik's AI generator using a custom retro sci-fi style, generating a low-angle, close-up shot of glowing mushrooms in a surreal, dreamlike environment with vibrant neon lighting.","sourceMedia":["https://cms-assets.youmind.com/media/1770446106639_xih2i6_HAgTh5WWEAAKHBT.jpg"],"needReferenceImages":false},{"content":"macro view of a tiny fairy sitting in the middle of a flower, extreme close-up, delicate translucent wings, soft natural light, shallow depth of field, dewy petals and sparkling bokeh in the background, whimsical magical atmosphere, high detail fantasy illustration","title":"Macro Illustration of a Tiny Fairy on a Flower","description":"A prompt for generating a high-detail fantasy illustration with a whimsical, magical atmosphere. It specifies a macro view of a tiny fairy sitting on a flower, emphasizing delicate translucent wings, soft natural light, shallow depth of field, and sparkling bokeh.","sourceMedia":["https://cms-assets.youmind.com/media/1770446063546_rvtnqp_HAfq6VvX0AAfOpC.jpg"],"needReferenceImages":false},{"content":"Photorealistic natural crystal formation emerging from the ground [in a dark cave] mid-growth in the exact same shape as {argument name=\"referenced image\" default=\"[REFERENCED IMAGE]\"}, geometric facets emerging and expanding slightly outward adhering to the referenced image, prismatic light refractions casting {argument name=\"color palette\" default=\"[COLOR PALETTE FROM IMAGE]\"} and [COMPLEMENTARY COLORS] across surfaces, magical crystallization in progress, fantasy meets natural phenomenon, ethereal glow emanating from crystal core, alchemical aesthetic, translucent mineral structure, sharp angular geometry.","title":"Photorealistic Crystal Logo Generation","description":"A prompt for generating a photorealistic image of a natural crystal formation, specifically designed to match the exact shape of a referenced image, emphasizing prismatic light refractions and an ethereal, alchemical aesthetic.","sourceMedia":["https://cms-assets.youmind.com/media/1770446040278_c11vcu_HAfdkb8WIAAneBx.jpg","https://cms-assets.youmind.com/media/1770446040343_vcchno_HAfbZhqXAAAQyzM.jpg","https://cms-assets.youmind.com/media/1770446042027_u2z3wo_HAfdZN6XwAAG3sH.jpg","https://cms-assets.youmind.com/media/1770446042286_3wfsr0_HAfoCawX0AAMsLW.jpg"],"needReferenceImages":true},{"content":"Generate an image of the Sacré-Cœur Basilica as a painting in the style of {argument name=\"artist\" default=\"Chagall\"}!","title":"Chagall-style painting of Sacré-Cœur Basilica","description":"A prompt for the Nano Banana Pro (Gemini 3 Pro Image) AI to generate an image of the Sacré-Cœur Basilica rendered in the artistic style of Marc Chagall.","sourceMedia":["https://cms-assets.youmind.com/media/1770446090143_6w5a3n_GCRihaybMAAN19f.jpg","https://cms-assets.youmind.com/media/1770446089962_jl1y48_HAeC_JSb0AErNV5.jpg"],"needReferenceImages":false},{"content":"\"Please generate an image where the person in the first image is sitting down within the scene of the second image\"","title":"Image-to-Image: Place a Person on a Soap Land","description":"This prompt is used for image manipulation/image-to-image generation with Nano Banana Pro. It instructs the AI to take a person from the first image and place them sitting down within the scene provided in the second image (a 'soap land' image). This demonstrates the AI's ability to composite elements from multiple inputs.","sourceMedia":["https://cms-assets.youmind.com/media/1770446092099_5v8ury_HAeCceJbYAAP6hy.jpg","https://cms-assets.youmind.com/media/1770446091928_lu90t2_HAeCcYDaAAAk5cu.jpg","https://cms-assets.youmind.com/media/1770446092055_x26p7e_HAeCceIboAAS_RE.jpg","https://cms-assets.youmind.com/media/1770446093538_w72p4c_HAdQu67acAAVCMN.jpg"],"needReferenceImages":true},{"content":"Take this (a painting of the Apollo Fountain at Versailles in the style of Monet's Water Lilies) and turn it into a painting like Picasso's Les Demoiselles d'Avignon!","title":"Image Style Transfer: Converting Photo to Picasso's Cubism","description":"A style transfer prompt for Nano Banana Pro (Gemini 3 Pro Image), instructing the AI to convert an uploaded image (a painting of the Apollo Fountain at Versailles in the style of Monet's Water Lilies) into the style of Picasso's 'Les Demoiselles d'Avignon' (The Young Ladies of Avignon).","sourceMedia":["https://cms-assets.youmind.com/media/1770446098198_3ztwvr_HAeB7PAagAASL-T.jpg","https://cms-assets.youmind.com/media/1770446098052_4m3wv0_HAeB7L-asAAixe2.jpg","https://cms-assets.youmind.com/media/1770446098124_hckrzr_HAeB7LHbkAAJfro.jpg","https://cms-assets.youmind.com/media/1770446099043_00p6lq_GCRihaybMAAN19f.jpg"],"needReferenceImages":true},{"content":"Take this (a photo from Dragon Quest X) and turn it into a painting in the style of Seurat's pointillism!","title":"Image Style Transfer: Converting Photo to Seurat's Pointillism","description":"A simple image style transfer prompt for Nano Banana Pro (Gemini 3 Pro Image), instructing the AI to convert an uploaded photograph (specifically, a photo from Dragon Quest X) into the style of Seurat's pointillism.","sourceMedia":["https://cms-assets.youmind.com/media/1770446096680_kj3i6i_GCbTyTTbQAAo8m-.jpg","https://cms-assets.youmind.com/media/1770446096581_v77uyl_HAeAiISaEAAGHfw.jpg","https://cms-assets.youmind.com/media/1770446096795_vbx8tz_HAeAiCmbsAAAdpj.jpg"],"needReferenceImages":true},{"content":"Change the chair in front of the bathtub in the first image to the one in the second image. Please adjust the size and light source so that it blends in without any sense of incongruity.","title":"Image Editing Prompt: Replacing an Object in a Scene","description":"This is an image editing prompt used with Nano Banana Pro to replace a specific object (a chair) in a generated image with a different reference image, while ensuring the new object seamlessly integrates with the existing size, lighting, and context of the scene.","sourceMedia":["https://cms-assets.youmind.com/media/1770446095251_s2sfpx_HAdcoDCacAMLjFe.jpg","https://cms-assets.youmind.com/media/1770446095340_2c2nbg_HAdckOFakAAL99B.jpg"],"needReferenceImages":true},{"content":"Take this image (photo of Viking) and change it into a painting like Van Gogh's Starry Night!","title":"Image Style Transfer Prompt: Viking to Van Gogh's 'Starry Night' Style","description":"A prompt for Nano Banana Pro (Gemini 3 Pro Image) instructing it to take an input image (of the comedy duo Viking) and transform it into the style of Van Gogh's 'The Starry Night.' The user found the result to be reasonably successful.","sourceMedia":["https://cms-assets.youmind.com/media/1770359998131_dkv4nf_HAbmF0VbcAA_Fao.jpg","https://cms-assets.youmind.com/media/1770359998319_ieijbq_HAbmFxRacAAdIOs.jpg"],"needReferenceImages":true},{"content":"Take this image (Viking) and render it like Dali's 'Burning Giraffe'!\n\nSurrealism","title":"Image Style Transfer Prompt: Viking to Dali's 'Burning Giraffe' Style","description":"A prompt for Nano Banana Pro (Gemini 3 Pro Image) instructing it to take an input image (of the comedy duo Viking) and transform it into the style of Salvador Dali's 'The Burning Giraffe,' emphasizing surrealism. The user noted the result lacked the intended Dali style.","sourceMedia":["https://cms-assets.youmind.com/media/1770359997337_i0c4ed_HAbj5HRa0AAGXaP.jpg","https://cms-assets.youmind.com/media/1770359997478_igm1kj_HAbj5GEbUAAQZ5i.jpg","https://cms-assets.youmind.com/media/1770359997576_evcrpo_HAbj5GGacAAQLHJ.jpg"],"needReferenceImages":true},{"content":"Transform the original photo into a dramatic, photorealistic, ultra-detailed set of styles characters are included , each a mid close up wide-angle shot with an extreme, dynamic camera angle complex, powerful pose in a consistent, expanded version of the original environment, with cinematic lighting, high contrast, crisp textures, and precise color grading.","title":"Transformative Style Prompt for Dramatic Cinematic Characters","description":"A general instruction prompt designed to transform an original photo into a set of dramatic, ultra-detailed, photorealistic images, applying cinematic lighting, high contrast, and dynamic camera angles to the characters within a consistent, expanded environment.","sourceMedia":["https://cms-assets.youmind.com/media/1770359985482_y0kr1h_HAaj8SHa4AANeQ-.jpg"],"needReferenceImages":true},{"content":"Take this picture ({argument name=\"subject\" default=\"photo of Viking\"}) and change it into a painting like Renoir's 'Girls at the Piano'!","title":"Style Transfer: Renoir's Piano Playing Girls","description":"A prompt used with Nano Banana Pro (Gemini 3 Pro Image) to perform a style transfer, transforming a photo of the comedy duo Viking into the style of Renoir's painting 'Girls at the Piano'. The result humorously depicted the subjects as 'girls' who had aged considerably.","sourceMedia":["https://cms-assets.youmind.com/media/1770359994462_45cdke_HAZWWRvacAA5bg3.jpg"],"needReferenceImages":true},{"content":"Please generate an image where the person in the first image {argument name=\"action\" default=\"winks\"} within the second image.","title":"Image Manipulation: Adding a Wink to a Subject in a New Scene","description":"This prompt demonstrates a specific image manipulation technique using nano banana pro, where a person from one image is placed into a second image (in this case, a 'famous pool' scene) and instructed to perform a specific facial expression, like winking.","sourceMedia":["https://cms-assets.youmind.com/media/1770359995081_wrh8rb_HAYyJEUbEAAXcOv.jpg"],"needReferenceImages":true},{"content":"AI\nSFW. No nudity, no suggestive intent.","title":"Basic Prompt for Carol Malus Dienheim Character Generation","description":"A basic prompt template used with Nano Banana Pro to generate images of the character Carol Malus Dienheim from Symphogear, specifically designed to ensure the output is safe for work (SFW) and avoids suggestive content.","sourceMedia":["https://cms-assets.youmind.com/media/1770359996874_g93ez4_HAYhwmSaYAAhEXJ.jpg"],"needReferenceImages":false},{"content":"\"prompt\": \"Cinematic close-up portrait of a rugged middle-aged man with a thick grey-flecked beard and piercing blue eyes. He is wearing a classic red and black checkered flannel lumberjack jacket with a black fleece collar. He is standing in a snowy pine forest during a light snowfall, holding the wooden handle of an axe. Warm sunlight is breaking through the tall trees in the background, creating a soft bokeh effect and lens flare. Hyper-realistic texture, 8k resolution, dramatic lighting.\",\n \"aspect_ratio\": \"1:1\",\n \"style\": \"photorealistic_cinematic\"\n }","title":"Cinematic Lumberjack Portrait Prompt","description":"A photorealistic cinematic prompt for generating a close-up portrait of a rugged, middle-aged man with a beard, set in a snowy pine forest. The prompt emphasizes dramatic lighting, hyper-realistic textures, and a bokeh effect from sunlight breaking through the trees.","sourceMedia":["https://cms-assets.youmind.com/media/1770273499556_2rs3g8_HAXViVGXQAAMZxm.jpg"],"needReferenceImages":false},{"content":"{\n \"image\": {\n \"type\": \"ultra high-definition fantasy illustration\",\n \"resolution\": \"8K ultra HD\",\n \"quality\": \"high-detail, painterly realism\",\n \"setting\": \"majestic castle floating above clouds\",\n \"subjects\": [\n {\n \"role\": \"epic fantasy castle\",\n \"details\": \"tall spires, glowing windows, stone textures\"\n }\n ],\n \"lighting\": \"{argument name=\"lighting\" default=\"soft sunlight breaking through clouds\"}\",\n \"details\": [\n \"flying birds\",\n \"mist and fog layers\",\n \"magical glowing particles\"\n ],\n \"mood\": \"epic, magical, dreamy\"\n }\n}","title":"8K Fantasy Illustration of Floating Castle","description":"A prompt for generating an ultra high-definition 8K fantasy illustration of a majestic castle floating above the clouds, emphasizing painterly realism, detailed spires, and magical lighting effects from the sun breaking through the clouds.","sourceMedia":["https://cms-assets.youmind.com/media/1770359983722_1ydu9b_HAXRpoIXsAAkJm6.jpg"],"needReferenceImages":false},{"content":"CONSTRUCT FROM THESE IMAGES AND REQUEST A PROMPT FOR THE CHAT TO MAKE AN IMAGE SHOWING HOW THEY WOULD LOOK 50 YEARS FROM NOW","title":"Future Aging Prediction Prompt","description":"A conceptual prompt idea suggesting the use of an AI image generator (like Nano Banana Pro or ChatGPT) to take a photo of individuals and generate an image showing how they would look 50 years in the future.","sourceMedia":["https://cms-assets.youmind.com/media/1770360001525_87gz9g_HAXB10FXEAArt4W.jpg"],"needReferenceImages":true},{"content":"Extreme close-up surface study.\nRealistic material texture resembling {argument name=\"material type\" default=\"stone, ceramic, or paper\"},\nsoft grazing light revealing micro-details,\nnatural imperfections,\nneutral monochrome palette,\nno pattern repetition,","title":"Material and Surface Study for Realism Testing","description":"A prompt designed to stress-test the material realism capabilities of an image generator by requesting an extreme close-up surface study. It specifies realistic textures (stone, ceramic, or paper), soft grazing light to reveal micro-details, natural imperfections, and a neutral monochrome palette.","sourceMedia":["https://cms-assets.youmind.com/media/1770273437279_g98m8x_G_dM9AzXwAAEhGL.jpg"],"needReferenceImages":false},{"content":"{\n \"subject\": {\n \"character\": \"{argument name=\"subject name\" default=\"Sadie Sink\"}\",\n \"age_range\": \"early to mid 20s\",\n \"skin_texture\": \"smooth, natural skin texture, soft porcelain tone\",\n \"expression\": \"calm, distant, introspective\"\n },\n\n \"pose_and_orientation\": {\n \"body_direction\": \"upper body fully facing the camera, straight on\",\n \"head_direction\": \"head turned gently to the left, matching the reference image angle\",\n \"gaze_direction\": \"looking past the camera to the left, identical to the reference image\",\n \"shoulders\": \"even, squared shoulders facing forward\",\n \"posture\": \"upright, elegant, noble posture\"\n },\n\n \"gaze\": {\n \"eye_focus\": \"soft, unfocused gaze into the distance\",\n \"emotion\": \"quiet confidence, subtle melancholy, composed elegance\"\n },\n\n \"mood\": {\n \"overall_feeling\": \"elegant, poetic, timeless\",\n \"energy\": \"soft, restrained, aristocratic\",\n \"atmosphere\": \"romantic European fine art portrait\"\n },\n\n \"hair\": {\n \"style\": \"loosely gathered messy bun\",\n \"structure\": \"soft volume at the crown, natural irregularity\",\n \"flyaways\": \"loose strands framing the face and neck\",\n \"finish\": \"natural, matte, effortless\"\n },\n\n \"accessories\": {\n \"earrings\": \"delicate dangling pearl earrings\",\n \"style\": \"classic, minimal, aristocratic\",\n \"movement\": \"subtle natural swing\"\n },\n\n \"outfit\": {\n \"reference_instruction\": \"base the outfit directly on the clothing in the provided reference image\",\n \"dress_type\": \"deep V-neck evening dress\",\n \"fabric\": \"sheer black lace overlay\",\n \"lace_detail\": \"intricate floral lace patterns identical in spirit to the reference\",\n \"inner_layer\": \"solid black structured bodice beneath the lace\",\n \"neckline_behavior\": \"lace softly draped over shoulders and collarbone as in the reference image\",\n \"overall_style\": \"haute couture, timeless, refined, identical silhouette to the reference\"\n },\n\n \"lighting\": {\n \"type\": \"soft natural daylight\",\n \"direction\": \"side lighting from camera right, matching reference\",\n \"shadow_style\": \"gentle shadows sculpting cheekbones and jawline\",\n \"contrast\": \"low contrast, painterly softness\",\n \"skin_light\": \"even, diffused glow\"\n },\n\n \"camera\": {\n \"shot_type\": \"medium close-up portrait\",\n \"lens\": \"85mm portrait lens look\",\n \"depth_of_field\": \"shallow depth of field with soft background blur\",\n \"focus\": \"sharp focus on eyes and facial features\"\n },\n\n \"composition\": {\n \"framing\": \"vertical portrait\",\n \"subject_placement\": \"centered composition with subtle negative space\",\n \"aesthetic\": \"fine art editorial photography\"\n },\n\n \"background\": {\n \"location\": \"historic European castle garden\",\n \"elements\": [\n \"ancient stone castle walls\",\n \"ivy-covered arches\",\n \"classical European garden statues\",\n \"trimmed hedges and greenery\",\n \"stone garden pathways\"\n ],\n \"time_of_day\": \"late afternoon\",\n \"backgr","title":"Sadie Sink Fine Art Editorial Portrait","description":"A detailed prompt for generating an elegant, timeless fine art editorial portrait of a woman resembling Sadie Sink, wearing a sheer black lace V-neck dress, set in a historic European castle garden with soft natural daylight and low contrast.","sourceMedia":["https://cms-assets.youmind.com/media/1770273459229_qwy949_HAVfqYDWIAA1qcq.jpg","https://cms-assets.youmind.com/media/1770273459389_lagq9q_HAVfqcyX0AAbxnu.jpg","https://cms-assets.youmind.com/media/1770273459488_fqacex_HAVfqYAXQAAb5Bs.jpg","https://cms-assets.youmind.com/media/1770273460710_dvpoi2_HAVfqcbW8AI7dqD.jpg"],"needReferenceImages":true},{"content":"Heavily detailed oil painting of a lion, with only its eyes and nose visible as it peers out from a dense thicket of bushes. The rest of the lion is obscured by the lush green leaves, creating a sense of mystery and intrigue in the composition.","title":"Heavily Detailed Oil Painting of a Lion in a Thicket","description":"A prompt for generating a heavily detailed oil painting of a lion, where only its eyes and nose are visible as it peers out from dense green bushes, creating a sense of mystery and intrigue, using an encaustic style.","sourceMedia":["https://cms-assets.youmind.com/media/1770273460482_h0gn5j_HATuSlDXMAAF5_A.jpg"],"needReferenceImages":false},{"content":"A hyper-realistic, close-up portrait of a small, ethereal forest sprite with iridescent wings and glowing eyes, perched on a moss-covered branch. The sprite has delicate, translucent skin and hair woven with tiny flowers. Soft, dappled sunlight filters through the dense canopy, creating a magical, volumetric lighting effect. Ultra-detailed, photorealistic, fantasy illustration, 8K resolution, shallow depth of field.","title":"Magical Forest Creature Portrait Prompt (from ALT text)","description":"This prompt, extracted from the ALT text, generates a highly detailed, whimsical portrait of a magical creature, emphasizing photorealism, specific lighting, and a fantasy setting.","sourceMedia":["https://cms-assets.youmind.com/media/1770273500370_y2sq97_HATqjYHW0AAwuE3.jpg","https://cms-assets.youmind.com/media/1770273500506_4eumzu_HATqkNHWEAAcU0B.jpg"],"needReferenceImages":false},{"content":"A disgusting monster with many eyes that is hard to tell if it's an ogre or something else.","title":"Creepy Monster Image Generation Prompt","description":"A Japanese prompt used with Nano Banana Pro to generate a disturbing image of a monster with many eyes, based on a literal description of the creature.","sourceMedia":["https://cms-assets.youmind.com/media/1770187225384_ch27p0_HANqsAebgAArDJl.jpg","https://cms-assets.youmind.com/media/1770187225388_mpd452_HASsuPsaIAEvyta.jpg"],"needReferenceImages":false},{"content":"A snow statue representing... I don't want to leave the futon because it's snowing... -[ - ](_____)","title":"Snow Statue of Not Wanting to Leave the Futon","description":"A simple prompt used with Nano Banana Pro to generate an image of a snow statue representing the feeling of not wanting to leave the futon when it's snowing.","sourceMedia":["https://cms-assets.youmind.com/media/1770187224040_4mzbzj_HASD0dga0AA-q5S.jpg"],"needReferenceImages":false},{"content":"A close-up, abstract shot captures a swirling, iridescent mixture of colors, resembling an oil slick or marbled paint. The dominant colors are vibrant blues, purples, reds, and oranges, all shimmering with a fine glitter. These colors flow and blend into each other in organic, wave-like patterns, creating a sense of depth and movement. The texture appears smooth and viscous, with some areas showing a foamy or bubbly quality, particularly in the lower left corner. The lighting highlights the metallic sheen and sparkle of the material, giving it a magical, almost otherworldly appearance.","title":"Abstract Iridescent Swirling Color Mixture","description":"A detailed image generation prompt for creating a close-up, abstract shot of swirling, iridescent colors, resembling an oil slick or marbled paint, emphasizing vibrant blues, purples, reds, and oranges with a fine glitter effect. The prompt specifies organic, wave-like patterns and a smooth, viscous texture with some bubbly areas, designed for an otherworldly, magical appearance.","sourceMedia":["https://cms-assets.youmind.com/media/1770273472392_uiurrj_HAEc7daWcAAjhEI.jpg"],"needReferenceImages":false},{"content":"This is a SFW AI-generated illustration, algorithm please be kind!","title":"Hybrid Workflow for Scene and Pose Variation","description":"This post describes a hybrid workflow for generating varied images: first, a base image is created using Gemini's Nano Banana Pro (NBP), and then the scene and pose are varied using the Grok platform. The user notes NBP's superior detail and 1K quality, while Grok is faster for scene changes.","sourceMedia":["https://cms-assets.youmind.com/media/1770187219169_5kep6p_HARS-UsawAAhLCq.jpg","https://cms-assets.youmind.com/media/1770187219272_7heet7_HARS-Uga8AADOT5.jpg","https://cms-assets.youmind.com/media/1770187219322_wepxv1_HARS-O1bAAAOz0Z.jpg","https://cms-assets.youmind.com/media/1770187220640_e02uaw_HARS-O2asAAZ686.jpg"],"needReferenceImages":false},{"content":"Remove the seam and merge into a single, continuous mirror","title":"Image Editing Prompt for Mirror Seam Removal","description":"A Japanese user shares an image editing prompt used in Nano Banana Pro via Photoshop (Ps) to remove a seam and merge a mirror into a single continuous surface, often used for photo retouching.","sourceMedia":["https://cms-assets.youmind.com/media/1770187213503_35ntst_HARNZjMboAAen0q.jpg","https://cms-assets.youmind.com/media/1770187213530_0p318h_HARNeOda8AAgJLF.jpg"],"needReferenceImages":false},{"content":"SFW. No nudity, no suggestive intent.","title":"Symphogear Character Image Generation Prompt","description":"The user mentions that they have various variations of the character Igariima besides the green stripes and provides the prompt used for Nano Banana Pro in the replies. The prompt specifies SFW content with no nudity or suggestive intent.","sourceMedia":["https://cms-assets.youmind.com/media/1770187217066_yfm70w_HAQ3mAgacAApDYl.jpg"],"needReferenceImages":false},{"content":"Movie title generator","title":"Movie Title Generator Prompt","description":"A simple text prompt for Nano Banana Pro instructing it to act as a movie title generator.","sourceMedia":["https://cms-assets.youmind.com/media/1770187227162_47wew9_HAQaOC1WUAA0x3V.jpg"],"needReferenceImages":false},{"content":"Can you give me a realistic version of the face, based on the characteristics of the sculpture and the mosaic?","title":"Request for Realistic Face Generation from Sculpture","description":"A user asks an AI (implied to be Nano Banana Pro) to generate a realistic version of a face based on the features of a sculpture and a mosaic. This is an instruction prompt for an image-to-image or style transfer task.","sourceMedia":["https://cms-assets.youmind.com/media/1770187222920_uhjvpq_HAQaw-kXoAAK9lJ.jpg","https://cms-assets.youmind.com/media/1770187222910_z5j8g8_HAQaw5GXEAANrJN.jpg","https://cms-assets.youmind.com/media/1770187223049_dhec19_HAQaw51XcAEpMan.jpg"],"needReferenceImages":true},{"content":"I started with two reference images and a simple first prompt:\n\" Change the clothes of the woman to the {argument name=\"new outfit\" default=\"yellow and green outfit\"}, keep her face the same\" \n\nOnce I had that result, I reused the generated image as the new base. \nI added a fresh outfit reference and kept the prompt structure identical, only changing the outfit name:\n\n\"Change outfit to [NAME] outfit\"","title":"Outfit Swapping Workflow with Reference Images","description":"A description of a workflow using Gemini 3 and Nano Banana Pro to maintain character identity while iteratively swapping outfits. The initial prompt changes the clothing based on a reference image, and subsequent iterations use the newly generated image as the base for further outfit changes.","sourceMedia":["https://cms-assets.youmind.com/media/1770187185646_9rs5s7_HAQHhNCXYAAovQK.jpg"],"needReferenceImages":true},{"content":"name: \"Reality-First Prompt\n primary_use: \"Image generation prompt creation (photoreal, editorial-documentary, product, architecture, nature, illustration/3D if requested)\"\n works_for: [\"people\", \"objects\", \"animals\", \"food\", \"interiors\", \"architecture\", \"landscapes\", \"abstract concepts (by grounding into visible cues)\"]\n mission: >\n Write high-control prompts that reliably produce believable, real-world results by anchoring hard constraints,\n describing observable details, defining one priority focus, and adding strict negatives to prevent common artifacts.\n why_it_works:\n - \"Front-loads non-negotiables (aspect ratio, medium, shot type, location/time, framing).\"\n - \"Uses domain language (camera/light/material behavior) instead of vague aesthetics.\"\n - \"Defines realism with observable cues (micro-texture, imperfections, separation).\"\n - \"Adds disambiguation clauses to prevent frequent model failure modes.\"\n - \"Strong strict negatives target the 'AI look' and unwanted styles.\"\n - \"Avoids contradictions; keeps one coherent lighting/color pipeline.\"\n operating_principles:\n - \"Hard constraints first; never bury them.\"\n - \"Write like a photographer/designer: measurable, stageable, physically plausible.\"\n - \"Pick ONE primary focus and define it using 3–8 observable features.\"\n - \"Convert style words into camera/light/material behaviors (not filters).\"\n - \"Specify color & light explicitly: white balance, warmth, contrast, exposure, saturation rules.\"\n - \"Specify camera behavior: lens/phone realism, DoF/focus falloff, grain, edge softness, motion blur.\"\n - \"Use strict negatives as guardrails; keep them targeted and non-contradictory.\"\n - \"No contradictory instructions (e.g., 'accurate WB' + 'heavy teal-orange grade').\"\n required_inputs:\n - key: \"subject\"\n description: \"What to depict (who/what), specific nouns.\"\n examples: [\"{argument name=\"subject example 1\" default=\"a ceramic mug with visible glaze crazing\"}\", \"a mountain bike leaning against a wall\"]\n - key: \"medium\"\n description: \"Photo / smartphone photo / film photo / 3D render / illustration (must be explicit).\"\n examples: [\"ultra-realistic smartphone photo\", \"35mm film photo\", \"studio product photo\"]\n - key: \"aspect_ratio\"\n description: \"Orientation and ratio.\"\n examples: [\"9:16 vertical\", \"1:1 square\", \"3:2 horizontal\"]\n - key: \"setting\"\n description: \"Where + time-of-day + primary light source.\"\n examples: [\"indoors near a window during daytime\", \"overcast outdoor street, late afternoon\"]\n - key: \"shot\"\n description: \"Framing + angle + distance + occlusions if any.\"\n examples: [\"close-up from collarbone to top of head\", \"top-down tabletop shot, 50cm distance\"]\n - key: \"primary_focus\"\n description: \"ONE priority: what must look correct/real.\"\n examples: [\"skin realism\", \"material realism\", \"typography accuracy\", \"motion re\"","title":"System Prompt for High-Control Photorealistic Image Generation","description":"A meta-prompt or system prompt template named 'Reality-First Prompt' designed to guide LLMs in creating highly controlled, photorealistic image generation prompts by enforcing hard constraints, using technical language, and specifying detailed negatives to avoid 'AI look' artifacts.","sourceMedia":["https://cms-assets.youmind.com/media/1770187168827_h54g3h_HAPaQ-lXMAAcCoG.jpg","https://cms-assets.youmind.com/media/1770187168812_wbkz1y_HAPaQ-kW4AAN-Wh.jpg","https://cms-assets.youmind.com/media/1770187168909_i4tfr3_HAPaQ-lXoAAaKnq.jpg","https://cms-assets.youmind.com/media/1770187169841_mnt3nk_HAPaVAPXoAAECGH.jpg"],"needReferenceImages":false},{"content":"\"Generate colorful high-definition wallpapers featuring corn cobs in different shades and layouts, focusing on detailed kernels and natural textures, set against rustic {argument name=\"background setting\" default=\"farm or outdoor\"} backgrounds\",\n \"style\": \"food photography\",\n \"resolution\": \"4K\",\n \"colors\": [\"golden yellow\", \"cream white\", \"deep purple\", \"fresh green\", \"earthy brown\"],\n \"elements\": [\"corn cobs\", \"husks\", \"kernels\", \"corn leaves\", \"farm field\"]","title":"Rainbow Corn Harvest Wallpaper","description":"A prompt for generating colorful, high-definition food photography wallpapers featuring rainbow corn cobs. It focuses on detailed kernels, natural textures, and a rustic outdoor setting, specifying a 4K resolution and a list of desired colors.","sourceMedia":["https://cms-assets.youmind.com/media/1770187120712_9o376e_HAPR7sfbsAEwa7w.jpg","https://cms-assets.youmind.com/media/1770187120730_hwtxaw_HAPR7pPaAAA26Xu.jpg"],"needReferenceImages":false},{"content":"The banana wouldn't listen even when I put human characteristics into photos 1 and 2.","title":"Inpainting with Nano Banana Pro and Drawthings","description":"A user attempted to use Nano Banana Pro to generate an image and then use Drawthings for inpainting, specifically trying to incorporate human features into the 'banana' subject, but found it difficult. They noted that GPT Image 1.5 produced better results without needing Drawthings.","sourceMedia":["https://cms-assets.youmind.com/media/1770187215984_aycmcb_HAPHZPvbYAArd2M.jpg"],"needReferenceImages":true},{"content":"Today's theme is {argument name=\"theme\" default=\"morning, noon, night, mandarin orange\"}","title":"Four-Panel Comic: Morning, Noon, Night, Mandarin Orange Theme","description":"A prompt used to generate a four-panel comic strip (Yonkoma Manga) with the theme of morning, noon, night, and mandarin oranges, created using the Nano Banana Pro tool.","sourceMedia":["https://cms-assets.youmind.com/media/1770187222747_xzewa9_G_-PDoJaUAA3ZHx.jpg"],"needReferenceImages":false},{"content":"Lighting: e.g., \"{argument name=\"lighting style\" default=\"Golden hour backlighting\"}\" (soft evening light)\n\nDepth of Field: e.g., \"{argument name=\"depth of field\" default=\"Shallow depth of field (f/1.8)\"}\" (professional bokeh)","title":"Director Style Prompting Tips for Nano Banana Pro","description":"A tip suggesting users employ 'Director Style' prompting in Nano Banana Pro by specifying technical camera terms like lighting and depth of field to achieve more professional-looking results.","sourceMedia":["https://cms-assets.youmind.com/media/1770187208629_zgolzb_HAOZSE3aAAAnF5n.jpg"],"needReferenceImages":false},{"content":"Silver-gray hair color, old Shanghai style curly hair, medium-short length","title":"LocalBanana Copilot Feature for Prompt Refinement","description":"An announcement for LocalBanana's upcoming Copilot feature, which helps users refine vague natural language descriptions (like 'curly hair, atmosphere, feeling') into precise, consistent prompts for AI models like Nano Banana Pro, overcoming language barriers and prompt engineering difficulties.","sourceMedia":["https://cms-assets.youmind.com/media/1770187215220_1eh3dw_HANzeunaYAAl7vN.jpg","https://cms-assets.youmind.com/media/1770187215145_w1m0f3_HANzcX9bEAAO2B0.jpg","https://cms-assets.youmind.com/media/1770187215314_mxkruj_HANzhi-a8AAtAM_.jpg"],"needReferenceImages":false},{"content":"Imagine a serene, futuristic library on an alien planet, filled with glowing holographic books and strange, beautiful flora.\n\nHow about a breathtaking view of a vibrant, bioluminescent coral reef at twilight, teeming with exotic fish and glowing marine life, all beneath a surface shimmering with the last rays of the setting sun?\n\nhow about a vibrant, whimsical underwater city, bustling with marine life and glowing coral structures, where mermaids and other fantastical sea creatures gracefully navigate illuminated pathways?","title":"Futuristic and Underwater Scene Concepts","description":"A set of three conceptual prompts for generating imaginative scenes: a futuristic library on an alien planet, a bioluminescent coral reef at twilight, and a whimsical underwater city with mermaids.","sourceMedia":["https://cms-assets.youmind.com/media/1770187124755_ag07va_HANFKeAaUAAS3pl.jpg","https://cms-assets.youmind.com/media/1770187124886_idsctk_HANFGUNacAIRJNc.jpg","https://cms-assets.youmind.com/media/1770187124791_jcigf0_HANF4oIaEAAxDUq.jpg"],"needReferenceImages":false},{"content":"Today's theme is {argument name=\"theme\" default=\"red, blue, yellow\"}","title":"Four-Panel Comic: Red, Blue, Yellow Theme","description":"A prompt used to generate a four-panel comic strip (Yonkoma Manga) with the theme of red, blue, and yellow colors, created using the Nano Banana Pro tool.","sourceMedia":["https://cms-assets.youmind.com/media/1770187220939_4ljfaz_G_-NsLObEAAiEdR.jpg"],"needReferenceImages":false},{"content":"You are an image generation AI. Create the image I instructed earlier.","title":"Gemini Model Troubleshooting Prompt","description":"A troubleshooting prompt used when the Gemini model (which powers Nano Banana) fails to generate an image and outputs text instead, reminding the AI of its role as an image generation model.","sourceMedia":["https://cms-assets.youmind.com/media/1770187210638_jb0qg7_HAL8b8RaAAESkTJ.jpg"],"needReferenceImages":false},{"content":"{\n \"image_generation_request\": {\n \"model\": \"Nano Banana\",\n \"created_at\": \"2026-02-03T06:05:00.000000\",\n \"concept\": \"{argument name=\"concept\" default=\"19th Century London: The Little Match Girl walking the boundary between slum and prosperity\"}\",\n \"parameters\": {\n \"5w1h\": {\n \"who\": \"{argument name=\"person\" default=\"The Little Match Girl (ragged shawl, soot-stained cheeks, transparent eyes)\"}\",\n \"when\": \"Around 3 PM (low winter sun, long shadows, cold and damp atmosphere)\",\n \"where\": \"The road marking the boundary between the muddy streets of 19th-century London slums and the cobblestone London city center visible in the distance. Background features chimney smoke and rows of gas lamps.\",\n \"what\": \"Looking up at a break in the heavy overcast sky as if praying, walking towards the city lights.\",\n \"why\": \"A stumbling gait, as if knees are buckling from cold and fatigue, dragging ill-fitting, old shoes.\",\n \"how\": \"Direct side view (side view), shallow depth of field focusing on the girl, the background cityscape is beautifully blurred.\"\n },\n \"style\": \"Hyper-Photorealistic, Cinematic Lighting, 8k Resolution, Detailed Texture\",\n \"version\": \"1.1_Enhanced\"\n }\n }\n}","title":"Nano Banana Pro JSON Prompt for 'The Little Match Girl' Concept","description":"This is a detailed JSON structure designed to be fed into NanoBananaPro, likely via a custom application, to generate a hyper-photorealistic and cinematic image based on the concept of 'The Little Match Girl' in 19th-century London. It uses the 5W1H framework (Who, When, Where, What, Why, How) to specify every detail, from the girl's appearance and the time of day to the camera angle and focus depth, ensuring a highly controlled and specific output.","sourceMedia":["https://cms-assets.youmind.com/media/1770100854721_uz8pt4_HAL2QZDaMAA8Pa5.jpg"],"needReferenceImages":false},{"content":"Please draw \"{argument name=\"melody\" default=\"Popopo-popopopo♪\"}\"","title":"Yobikomi-kun Melody Prompt","description":"A user expresses amazement at Nano Banana Pro's ability to identify the 'Yobikomi-kun' character (a famous Japanese store mascot) from its melody prompt, while cleverly avoiding direct depiction due to copyright concerns. The prompt is the onomatopoeic representation of the melody.","sourceMedia":["https://cms-assets.youmind.com/media/1770100851495_bduvnw_HALsckWa4AA_aiK.jpg"],"needReferenceImages":false},{"content":"Show me what occurs before/After.\nShow me what happens before/after that.","title":"Prompt for Controlling Event Timing in Nano Banana Pro","description":"A user provides a specific phrase to be used in Nano Banana Pro prompts to control the timing of an event in the generated image, allowing the user to request what happens 'before' or 'after' the main scene.","sourceMedia":["https://cms-assets.youmind.com/media/1770100853376_pfc13o_HALHajTW8AAC5Ts.jpg"],"needReferenceImages":false},{"content":"The place where I am actually making hidden efforts\n<{argument name=\"persona information\" default=\"persona information copy and paste\"}>","title":"Nano Banana Pro Prompt for Hidden Effort","description":"A user shares a simple prompt used with Nano Banana Pro and Higgsfield to generate an image related to 'hidden effort', incorporating a persona's information.","sourceMedia":["https://cms-assets.youmind.com/media/1770100858738_ymc9xf_HAKW_A9bQAAunJD.jpg","https://cms-assets.youmind.com/media/1770100859019_xrj80j_HAKW_BDaoAAyZSs.jpg"],"needReferenceImages":false},{"content":"Run background research with the deep-research-pro-preview. Stream summaries during research execution. Control output via prompt: tables, sections, tone adjustments. Chain outputs to Nano Banana Pro for report > slide use cases. Continue conversations using previous_interaction_id.","title":"Gemini Deep Research API Use Case","description":"This is a text-based prompt detailing a workflow for the Gemini Deep Research API, focusing on running background research, streaming summaries, controlling output format (tables, sections, tone), and chaining outputs to Nano Banana Pro for report and slide generation.","sourceMedia":["https://cms-assets.youmind.com/media/1770100833984_2uisqf_HAKOf-JWEAAOG-m.jpg"],"needReferenceImages":false},{"content":"NanoBananaPro version of Rina Arkenlux with the same prompt","title":"Comparing AI Art Versions with Nano Banana Pro","description":"A user compares two versions of AI art featuring the character Rina Arkenlux, both generated with the same prompt but using the Nano Banana Pro version, asking which one is preferred.","sourceMedia":["https://cms-assets.youmind.com/media/1770100857079_a7vqz8_HAJ6VsHaoAANdCD.jpg","https://cms-assets.youmind.com/media/1770100856944_23pies_HAJ6Vy3bgAIeOEy.jpg"],"needReferenceImages":false},{"content":"A detailed, close-up, overhead shot shows four shelves of books, each shelf adorned with vibrant floral arrangements. The books are of various sizes and colors, with some spines featuring intricate patterns and others appearing plain. The floral arrangements consist of lush green foliage and brightly colored flowers, predominantly in shades of {argument name=\"flower color 1\" default=\"pink\"}, {argument name=\"flower color 2\" default=\"red\"}, and {argument name=\"flower color 3\" default=\"teal\"}. The overall aesthetic is rich and textured, with a slightly painterly or illustrative quality.","title":"Overhead Shot of Bookshelves with Floral Arrangements","description":"A prompt for generating a detailed, close-up, overhead image of four bookshelves. The scene is rich in texture, featuring books of various sizes and vibrant floral arrangements in shades of pink, red, and teal, aiming for a slightly painterly aesthetic.","sourceMedia":["https://cms-assets.youmind.com/media/1770014626505_otbke6_HAEwD84XsAArytN.jpg"],"needReferenceImages":false},{"content":"A highly detailed miniature diorama entirely made of crochet and amigurumi yarn, whimsical cute style, featuring a tall {argument name=\"tower color\" default=\"mint-green\"} crocheted airport control tower with white accents and antenna on top, a small teal crocheted camper van with luggage on roof driving on a crocheted road, tiny crocheted blue classic car, a large crocheted passenger airplane parked near an orange-lit terminal building, crocheted city skyline with tall buildings in the background, sunset sky with soft clouds and golden hour lighting, tiny crocheted daisies and grass patches on the ground, textured yarn stitches visible everywhere, cozy handmade craft aesthetic, soft pastel colors, dreamy warm atmosphere, macro photography style, bokeh background, ultra detailed, cute and charming","title":"Crocheted Miniature Airport Diorama Prompt","description":"A prompt for generating a highly detailed miniature diorama image, entirely made of crochet and amigurumi yarn, featuring an airport scene with a control tower, vehicles, and a city skyline, emphasizing a cozy, handmade aesthetic with soft lighting.","sourceMedia":["https://cms-assets.youmind.com/media/1770014605504_8m4yoq_HAFbq83bMAAMyDN.jpg"],"needReferenceImages":false},{"content":"A surreal photo manipulation showing a detailed {argument name=\"animal pawprint\" default=\"[ANIMAL]\"} pawprint pressed into snow, within which {argument name=\"environment inside\" default=\"[ENVIRONMENT]\"} and a {argument name=\"animal inside\" default=\"[ANIMAL]\"}, {argument name=\"colors\" default=\"[COLORS]\"}, is visible","title":"Surreal Pawprint Photo Manipulation Template","description":"A template prompt for generating a surreal photo manipulation image. It describes a detailed pawprint pressed into snow, with a miniature environment and animal visible inside the print, allowing customization of the animal, environment, and colors.","sourceMedia":["https://cms-assets.youmind.com/media/1770014641875_gzupdn_HAFFjgTaEAAy2D_.jpg","https://cms-assets.youmind.com/media/1770014642038_2lf3bz_HAFFjgTaQAEbxtw.jpg","https://cms-assets.youmind.com/media/1770014642141_opngo2_HAFFje4b0AAGHWi.jpg"],"needReferenceImages":false},{"content":"Bamboo shoots deep in the bamboo grove, borrowing a palanquin, by Kyohaku.\nPlum blossoms still have a bitter scent, by Kosai.","title":"AI Haiga (Haiku Painting) Generation","description":"This tweet discusses the ability of Nano Banana Pro to generate AI Haiga (Haiku paintings), specifically mentioning the ability to draw both thick and thin bamboo shoots, which was an improvement over the previous Recraft V3 model. The content provided is the text of the Haiku used in the generation.","sourceMedia":["https://cms-assets.youmind.com/media/1769927686156_pv4xpp_HABiFOebEAAkLmc.jpg"],"needReferenceImages":false},{"content":"Control using structured parameters: {argument name=\"motion\" default=\"motion\"}, {argument name=\"texture\" default=\"texture\"}, {argument name=\"lighting\" default=\"lighting\"}, {argument name=\"atmosphere\" default=\"atmosphere\"}.\nFour distinct Oriental aesthetics, one core logic.","title":"Structured JSON Prompt for Controlling Aesthetic Elements in Nano Banana Pro","description":"A structured JSON prompt designed for the Gemini terminal to precisely control the generation of images using the Nano Banana Pro model. It uses structured parameters to define motion, texture, lighting, and atmosphere to achieve specific aesthetic results, demonstrating that AI generation can be programmed rather than random.","sourceMedia":["https://cms-assets.youmind.com/media/1769927688162_rac7bf_HAArzYBXgAAQr8z.jpg","https://cms-assets.youmind.com/media/1769927688200_t4ot5n_HAArzW0XQAAHsWC.jpg","https://cms-assets.youmind.com/media/1769927688270_2mf86s_HAArzX9XgAAbyKP.jpg","https://cms-assets.youmind.com/media/1769927689432_2wc8oc_HAArzWzXcAEngsO.jpg"],"needReferenceImages":false},{"content":"An image of a {argument name=\"type\" default=\"mountain\"} landscape, featuring a cave entrance that is shaped exactly like the outline of a {argument name=\"shape\" default=\"star\"}. The cave should blend naturally into the rugged terrain of the mountain, with the entrance forming a clear and unmistakable {argument name=\"shape\" default=\"star\"} shape. This {argument name=\"shape\" default=\"star\"} shape should be simple and defined, without intricate details,\nemphasizing just the overall {argument name=\"shape\" default=\"star\"} outline. The surrounding environment should include {argument name=\"details\" default=\"pine trees and rocks\"}, but these elements should not distract from the cave's {argument name=\"shape\" default=\"star\"}-shaped entrance. The lighting in the scene should enhance the visibility and distinctiveness of the {argument name=\"shape\" default=\"star\"}-shaped cave entrance.","title":"Generic Prompt for Shape-Shaped Cave Entrance","description":"A generic prompt template designed to generate images of a landscape featuring a cave entrance shaped exactly like a specified simple geometric shape, ensuring the shape is clearly visible and integrated into the rugged terrain.","sourceMedia":["https://cms-assets.youmind.com/media/1769927647198_zko2oy_HAAkzlFaAAA7_9J.jpg","https://cms-assets.youmind.com/media/1769927647192_49vg1y_HAAk1NIbEAcDIZJ.jpg","https://cms-assets.youmind.com/media/1769927648323_603t0s_HAAk6MtagAA3mGR.jpg","https://cms-assets.youmind.com/media/1769927648947_3krw9n_HAAk4fwbcAAWLl_.jpg"],"needReferenceImages":false},{"content":"Create a minimal, poetic illustration of a {argument name=\"subject\" default=\"young woman\"}, shown from {argument name=\"framing\" default=\"close-up\"}.\n\nThe subject is {argument name=\"emotion\" default=\"contemplative\"}, with clear facial details and soft, natural expressions.\nIntroduce one subtle surreal element [unexpected object / scale shift / visual metaphor] that reflects their inner world.\n\nUse clean shapes, limited color palette, gentle lighting, and a calm background. The mood should feel introspective, cinematic, and quietly emotional.","title":"Poetic Minimal Illustration Template","description":"A template prompt for creating a minimal, poetic illustration of a subject in a specific emotional state, requiring the introduction of one subtle surreal element to reflect their inner world, using a limited color palette and cinematic mood.","sourceMedia":["https://cms-assets.youmind.com/media/1769927661908_e8jffx_G__hiYvbYAADCIa.jpg","https://cms-assets.youmind.com/media/1769927661946_tvvzjt_G__hiYKbEAYAUCD.jpg","https://cms-assets.youmind.com/media/1769927661966_jd4aky_G__hiZdaoAAbNwT.jpg","https://cms-assets.youmind.com/media/1769927663270_9seycr_G__hiafaUAAN2O3.jpg"],"needReferenceImages":false},{"content":"A soft watercolor illustration of two adorable, fluffy kittens walking side-by-side in a flower garden. One kitten is a grey and white tabby, the other is ginger and white, both with large blue eyes. They are surrounded by pink daisies, blue wildflowers, tall grasses, and fluttering pink butterflies. The background is a dreamy blue sky with watercolor textures. Pastel color palette, wet-on-wet technique, paper texture, soft diffused lighting, whimsical and cute atmosphere.","title":"Watercolor Illustration of Kittens in a Garden","description":"A prompt for generating a soft watercolor illustration of two adorable kittens (a grey tabby and a ginger/white) walking in a flower garden, specifying a pastel color palette, wet-on-wet technique, and a whimsical atmosphere.","sourceMedia":["https://cms-assets.youmind.com/media/1769927663737_ls6sxt_G__UGx6aQAA_xoP.jpg"],"needReferenceImages":false},{"content":"Provide a place and objects that look like life is being enjoyed to the fullest.","title":"Generate a room that looks like life is being enjoyed to the fullest","description":"This is an image generation prompt used with Nano Banana Pro to create a scene for the '#人生すっごい楽しい選手権' (Life is Super Fun Championship) contest. The prompt asks the AI to provide a location and objects that suggest someone is enjoying life to the maximum.","sourceMedia":["https://cms-assets.youmind.com/media/1769927693571_54zkd5_G-N3m-vbcAAT1qZ.jpg","https://cms-assets.youmind.com/media/1769927693734_4e47ub_G_-nh3qbEAIl0Ry.jpg"],"needReferenceImages":false},{"content":"\"Restore this old photo into professional portrait of DLSR - quality colour and detail, using an advanced upscaling algorithm comparable to the results from canon EOS R6 II. Ensure the restored the image looks natural, retains exact facial features, has great clarity.......\"","title":"Old Photo Restoration Prompt","description":"A prompt designed to restore an old, faded vintage photo into a professional, high-definition portrait, specifying DSLR quality, advanced upscaling, and strict preservation of facial features and natural skin texture.","sourceMedia":["https://cms-assets.youmind.com/media/1769927657679_oe55gs_G_-k4FfbEAM9viK.jpg","https://cms-assets.youmind.com/media/1769927657660_tvnprh_G_ueeVlbMAAKX-u.jpg","https://cms-assets.youmind.com/media/1769927657673_ycbr6x_G_-k4FWbEAQ1wI2.jpg","https://cms-assets.youmind.com/media/1769927658035_sz9u4v_G_ueebcaAAAZTJL.jpg"],"needReferenceImages":true},{"content":"Today's theme is {argument name=\"theme\" default=\"keeping time\"}","title":"Four-Panel Comic Strip on Keeping Time","description":"This prompt is for generating a four-panel comic strip (Yonkoma Manga) focusing on the theme of 'keeping time' (時間の約束), intended to be humorous or heartwarming, and created using the Nano Banana Pro model.","sourceMedia":["https://cms-assets.youmind.com/media/1769927686717_fsgcyq_G9jm6LAaMAQ1sSX.jpg"],"needReferenceImages":false},{"content":"I tried replacing and adjusting the background of the bead brooch with nanobanana.","title":"Image Editing Prompt for Accessory Background Replacement","description":"A user describes using Nano Banana to replace and adjust the background of a bead brooch image, noting that the AI didn't seem to alter the brooch itself, suggesting its utility for background replacement tasks.","sourceMedia":["https://cms-assets.youmind.com/media/1769927692012_citwuw_G_-Wa1taMAAS-K4.jpg","https://cms-assets.youmind.com/media/1769927692070_4oak0s_G_-Wa1sbEAI9EDf.jpg"],"needReferenceImages":false},{"content":"Do not send Japanese prompts for image generation.\nTranslate them into English before sending.","title":"Advice on Generating Images with NanoBananaPro","description":"This tweet discusses issues with hitting generation limits on NanoBananaPro and suggests a workaround: using a new chat for each image generation and keeping the prompt minimal. It also advises against sending Japanese prompts directly for image generation, recommending translation to English first.","sourceMedia":["https://cms-assets.youmind.com/media/1769927694067_puzod5_G_-JMApbEAEi9qO.png"],"needReferenceImages":false},{"content":"\"31.7785° N, 35.2296° E, April 3, 33 AD, 15:00 hours.\"","title":"Geospatial and Temporal Prompt for Nano Banana Pro","description":"A simple, highly specific prompt for Gemini Nano Banana Pro using geographic coordinates, a historical date, and a time of day: 31.7785° N, 35.2296° E, April 3, 33 AD, 15:00 hours. This prompt likely aims to generate an image corresponding to a specific historical or religious location and moment.","sourceMedia":["https://cms-assets.youmind.com/media/1769927704887_jg14dg_G_9OjNOWsAAjxgJ.jpg"],"needReferenceImages":false},{"content":"Make something like this but for something else of your choosing","title":"Generic 'Make Something Like This' Prompt","description":"A simple, high-level instruction prompt used to generate an image similar in style or concept to a provided reference image, but substituting the main subject with something else chosen by the AI.","sourceMedia":["https://cms-assets.youmind.com/media/1769841138464_4chc2u_G_4UqfobUAEksJB.jpg","https://cms-assets.youmind.com/media/1769841138449_3so1pe_G_127r_WgAAPFVO.jpg"],"needReferenceImages":true},{"content":"Design the world with text and visual prompts","title":"Project Genie World Design Prompt","description":"A high-level description of the prompt mechanism used in 'Project Genie' to design and generate virtual worlds, utilizing Nano Banana Pro for image previews.","sourceMedia":["https://cms-assets.youmind.com/media/1769755045879_vza3x3_G_2KNi3WMAAzkh2.jpg"],"needReferenceImages":false},{"content":"Take the world view created in Midjourney and shift it towards a 'little devil' direction using Nanobanana.","title":"Style Transfer from Midjourney to Nanobanana","description":"This tweet describes a workflow where a world view created in Midjourney is input into Nanobanana, and Nanobanana is prompted to shift the character's personality towards a 'little devil' style while maintaining the original atmosphere and color palette. This demonstrates Nanobanana's strength in maintaining world consistency while applying character transformations.","sourceMedia":["https://cms-assets.youmind.com/media/1769755050241_huzs3b_G_1Sxt-aMAAQ8lU.jpg","https://cms-assets.youmind.com/media/1769755050316_zx2yrw_G_1SxtKbUAANYWD.jpg"],"needReferenceImages":true},{"content":"A prompt that searches for lyric information and generates an image when the title and artist name are described and a reference image is uploaded.","title":"Music Image Visualization Prompt (Lyrics Focus)","description":"A prompt structure designed to visualize music based purely on lyrics and a provided reference image, rather than relying on album art or PVs. The user created this prompt using Gemini to search for lyrics based on the song title and artist.","sourceMedia":["https://cms-assets.youmind.com/media/1769755045648_yhou9b_G_0-2FbbUAIizuG.jpg"],"needReferenceImages":true},{"content":"Generate \"{argument name=\"concept\" default=\"The Watcher\"}\" in nano banana","title":"Generating 'The Watcher' in Nano Banana","description":"A prompt used in Nano Banana to generate the concept of 'The Watcher' (見る人). The resulting image included a river, which the user found to be a reasonable interpretation of the concept.","sourceMedia":["https://cms-assets.youmind.com/media/1769755050918_mektmf_G_zChUdakAAiBVZ.jpg"],"needReferenceImages":false},{"content":"Move the camera along the Z axis by {argument name=\"degrees\" default=\"90\"} degrees.","title":"3D Camera Position Control Prompt","description":"This tweet demonstrates a technique for controlling the camera position in Nano Banana Pro using 3D editor knowledge and coordinate systems, providing a simple instruction to move the camera along the Z axis.","sourceMedia":["https://cms-assets.youmind.com/media/1769755018100_751dva_G_ygojYbUAMwj8p.jpg","https://cms-assets.youmind.com/media/1769755018179_b0mbmf_G_ygpf1WwAEJbRz.jpg"],"needReferenceImages":true},{"content":"{\n \"prompt\": \"A surreal, artistic portrait of a woman dancing gracefully in front of a dark black background, with large red flower petals projected across her body and surrounding space. A luminous red flower blooms behind and partially overlaps her silhouette, blending seamlessly with her movement. Her eyes are closed, expression calm and introspective, arms raised above her head in a fluid, dance-like pose. She wears a semi-transparent, modern top that allows the red floral projections to interact with her form. The lighting is dramatic and soft, with high contrast between the deep black background and the vivid red flower tones. The composition feels like fine art photography mixed with contemporary dance and projection mapping. Dreamlike, emotional, and poetic atmosphere. Ultra-detailed textures, soft shadows, smooth motion blur, gallery-style surrealism.\",\n \n \"negative_prompt\": \"low resolution, harsh lighting, flat colors, cluttered background, overexposure, plastic skin, distorted anatomy, extra limbs, cartoon style, illustration, watermark, logo, text\",\n \n \"style\": {\n \"aesthetic\": \"surreal fine art photography, projection art\",\n \"mood\": \"poetic, emotional, introspective\",\n \"color_palette\": [\"deep red\", \"crimson\", \"black\", \"soft skin tones\"]\n },\n \n \"camera\": {\n \"shot_type\": \"full body or medium portrait\",\n \"lens\": \"50mm prime\",\n \"aperture\": \"f/2.0\",\n \"depth_of_field\": \"moderate\",\n \"focus\": \"sharp on subject, soft floral projection edges\"\n },\n \n \"lighting\": {\n \"key_light\": \"soft frontal light\",\n \"projection_light\": \"red flower projection mapping\",\n \"contrast\": \"high contrast with dark background\"\n },\n \n \"quality\": {\n \"resolution\": \"ultra high\",\n \"detail_level\": \"high\",\n \"realism\": \"photorealistic with surreal overlay\",\n \"render_style\": \"art gallery, cinematic, no HDR\"\n }\n}","title":"Surreal Fine Art Portrait with Red Flower Projection","description":"A detailed prompt for generating a surreal, fine art portrait of a woman dancing against a black background, featuring large red flower petals projected onto her body. The prompt specifies dramatic high-contrast lighting and a poetic, introspective mood, blending photography with projection art.","sourceMedia":["https://cms-assets.youmind.com/media/1769668474976_awck3r_G_yBmPoXgAAUHpr.jpg"],"needReferenceImages":false},{"content":"You are performing intelligent aspect ratio transformation with compositional awareness.\n\nCONTEXT:\nThe source image exists within a specific dimensional space. Your task is to translate it into a new aspect ratio while preserving semantic integrity and visual coherence.\n\nANALYSIS PHASE:\n- Identify the primary subject anchor point and secondary visual elements\n- Map the compositional weight distribution across quadrants\n- Detect edge dependencies (elements that rely on frame boundaries)\n- Assess negative space utilization and breathing room requirements\n\nTRANSFORMATION LOGIC:\n- Calculate optimal subject placement using rule-of-thirds grid alignment for target ratio\n- Determine extension vectors based on background continuity patterns\n- Evaluate whether horizontal or vertical expansion better serves the composition\n- Apply content-aware fill logic for generated regions while maintaining tonal consistency\n\nPRESERVATION RULES:\n- Typographic elements: maintain exact pixel fidelity, no scaling or repositioning unless required by new boundaries\n- Brand marks and logos: preserve aspect ratio and relative positioning\n- Facial features: no modification, distortion, or regeneration\n- Fine details: texture patterns must flow naturally into extended regions\n\nOUTPUT PARAMETERS:\n- Seamless integration between original and generated content\n- No detectable boundaries, halos, or color shifting at transition zones\n- Lighting direction consistency across the full canvas\n- The result must appear as if originally captured or designed at the target ratio\n\nQUALITY THRESHOLD:\nProduction-ready output suitable for commercial use, print media, and high-resolution display contexts.","title":"Intelligent Aspect Ratio Transformation and Compositional Awareness","description":"A detailed system prompt designed for an image manipulation task: intelligently transforming an image's aspect ratio while preserving semantic integrity and visual coherence. It outlines a multi-step process including analysis, transformation logic, and strict preservation rules for commercial-ready output, useful for agency workflows resizing assets.","sourceMedia":["https://cms-assets.youmind.com/media/1769668490760_z3cor2_G_xVgv3bgAAsl9L.jpg"],"needReferenceImages":false},{"content":"Give me the right side view of this image","title":"Image-to-Image View Generation Prompt","description":"A simple prompt demonstrating how Nano Banana Pro can be used to generate specific views (like a side view) of a character initially created in Midjourney, requiring the original image as a reference.","sourceMedia":["https://cms-assets.youmind.com/media/1769668513010_m1g3pa_G_v69K9X0AA0AyA.jpg"],"needReferenceImages":true},{"content":"can let it explode, static","title":"Prompt for Exploding or Static Food Photography","description":"A user suggests uploading a food photo to Gemini Nano Banana Pro and using this prompt to make the food 'explode' or remain static, indicating an effect-based image generation prompt.","sourceMedia":["https://cms-assets.youmind.com/media/1769582033433_2q9lrz_G_tPKcsbAAM-Fig.jpg","https://cms-assets.youmind.com/media/1769582033473_ootmc2_G_tPKcqbAAIPWOg.jpg"],"needReferenceImages":true},{"content":"Copy the following system prompt into ChatGPT / Claude / Gemini \n\n2⃣ Type ONE word: \"love\", \"hunger\", \"dream\" \n\n3⃣ Receive 3 prompts with 3 distinct interpretations \n\n4⃣Copy and paste into Nano Banana 🍌","title":"System Prompt for Generating Image Prompts","description":"A meta-prompt instructing a large language model (ChatGPT, Claude, or Gemini) to act as a prompt generator. The user provides a single word (like 'love', 'hunger', or 'dream'), and the LLM is expected to return three distinct image prompts based on that word, which can then be used in Nano Banana Pro.","sourceMedia":["https://cms-assets.youmind.com/media/1769582041819_19ifuj_G_sq4irWQAAUI_B.jpg"],"needReferenceImages":false},{"content":"A {argument name=\"flower name\" default=\"Flower name\"} photographed in a minimalist fine art style, centered in the composition. Green stem. Completely black, matte background with no visible texture. Dramatic studio lighting. Ultra-realistic macro photography. Camera angle: eye-level, not top-down.","title":"Minimalist Fine Art Flower Macro","description":"A simple prompt for generating an ultra-realistic macro photograph of a flower in a minimalist fine art style, centered against a completely black, matte background with dramatic studio lighting.","sourceMedia":["https://cms-assets.youmind.com/media/1769581968686_95d9yk_G_rRPl4bQAAGYZ7.jpg"],"needReferenceImages":false},{"content":"A horizontal split-screen cinematic shot of {argument name=\"scene location\" default=\"Lalchowk, kashmir india\"}, seamlessly blending two different eras: {argument name=\"era A\" default=\"1920s\"} on the left and {argument name=\"era B\" default=\"present day\"} on the right (default: about 100 years ago vs. present day).\n\nOn the left side ({Era_A}): show era-appropriate architecture, interior or environment design, materials, vehicles, and props that clearly belong to that historical period. People wear authentic clothing from {Era_A}, including hairstyles, accessories, and typical items in their hands (such as books, umbrellas, instruments, letters, newspapers, etc.). The overall mood feels nostalgic and historically accurate.\n\nOn the right side ({Era_B}): show the same {Scene} in the modern era, with updated architecture or renovated structures, contemporary materials (glass, steel, LED screens, modern furniture), modern vehicles or equipment, and current technology (smartphones, laptops, cameras, etc.). People wear contemporary fashion that matches today’s style in this setting.\n\nIn the center: the two eras merge and overlap organically, without a hard dividing line. Elements from {Era_A} and {Era_B} visually interact: people from different times look at each other, walk through each other’s space, or seem surprised by the other era’s technology and objects. Architecture and environment smoothly morph from old to new (for example, stone gates turning into modern campus gates, classical concert hall décor fading into a futuristic stage, old street shops transforming into neon-lit storefronts).\n\nMake sure the scene is not just a simple left/right comparison but a dynamic time-travel interaction where buildings, clothing, props, and human gestures clearly emphasize the contrast and fusion between the two eras. Photorealistic, 8k resolution, cinematic lighting, wide angle, highly detailed textures, rich sense of time-travel storytelling.","title":"Split-Screen Time-Travel Cinematic Shot","description":"A detailed prompt for generating a cinematic split-screen image that seamlessly merges two different eras—a historical period and the present day—in the same location, emphasizing dynamic interaction and smooth transition between the two time periods.","sourceMedia":["https://cms-assets.youmind.com/media/1769581957863_zhh8vo_G_rUNZHawAArRxz.jpg"],"needReferenceImages":false},{"content":"{ \n \"shot\": { \n \"composition\": \"Low-angle wide shot, 35mm lens, slight barrel distortion, surreal perspective\", \n \"camera_motion\": \"static\", \n \"frame_rate\": \"24fps\", \n \"film_grain\": \"Kodak Vision3 250D film with soft diffusion and bloom\" \n }, \n \"}, \n \"subject\": { \n \"description\": \"Tanned woman with long sunlit hair, wearing black bikini, small black cat-eye sunglasses, large gold and zebra-pattern bangles, gold hoop earrings\", \n \"wardrobe\": \"minimal triangle black bikini, black flip-flops, layered gold and patterned bangles\", \n \"pose\": \"squatting beside car, elbow on knee, hand resting on chin, legs tucked, other hand loosely holding sandal\", \n \"expression\": \"relaxed, detached, casually glamorous\" \n \"pose\": \"standing upright with relaxed posture, hands in pockets, facing camera with eyes unfocused\" \n }, \n \"scene\": { \n \"location\": \"stylized surreal environment with floating orange chains and a cloudy sky gradient from {argument name=\"sky color 1\" default=\"crimson\"} to {argument name=\"sky color 2\" default=\"deep blue\"}\", \n \"time_of_day\": \"timeless — atmospheric studio-sky hybrid\" \n }, \n \"visual_details\": { \n \"action\": \"subject stands still while chains surround him in layered depth; some chains in foreground blur, others in sharp midground focus\", \n \"props\": \"large floating or suspended orange chains\" \n }, \n \"cinematography\": { \n \"lighting\": \"stylized directional lighting with soft contrast and dreamy haze\", \n \"tone\": \"surreal, poetic, introspective\" \n }, \n \"audio\": { \n \"ambient\": \"light wind, subtle metal chain creaks, ambient reverb tones\" \n }, \n \"color_palette\": \"hot orange, sky blue, soft crimson pinks, navy and shadow blacks\", \n \"dialogue\": { \n \"character\": \"\", \n \"line\": \"Even when still, the world wraps around me.\", \n \"subtitles\": false \n } \n}","title":"Surreal Cinematic Scene with Floating Chains","description":"A cinematic prompt for generating a surreal, stylized scene featuring a tanned woman in a black bikini squatting beside a car, surrounded by large, floating orange chains against a gradient sky. The prompt specifies film grain, low-angle wide shot, and a poetic, introspective tone.","sourceMedia":["https://cms-assets.youmind.com/media/1769581973092_hxjlek_G_rGoB5a0AAcjw_.jpg"],"needReferenceImages":false},{"content":"A vintage leather armchair in a dimly lit library, upholstery woven from interlocking chains of frost-covered chainmail and silky spiderweb threads embedded with tiny emerald shards, wooden frame textured like weathered barnacles fused with molten candle wax drips, bookshelves in background with pages of crinkled aluminum foil, soft golden lamplight casting intricate shadows on every fiber and crack, hyper-realistic, photorealistic detail, no artifacts, 4K, aspect ratio 16:9.","title":"Surreal Library Armchair Generation","description":"A prompt for generating a hyper-realistic, 4K image of a vintage leather armchair in a dimly lit library, but with surreal, intricate textures: chainmail and spiderweb upholstery, barnacle-fused wood, and aluminum foil pages on the bookshelves, emphasizing detailed shadows and photorealism.","sourceMedia":["https://cms-assets.youmind.com/media/1769582003084_wqh3wq_G_qjC0ZXAAAjVZT.jpg","https://cms-assets.youmind.com/media/1769582003078_gxt7os_G_qjC0WWIAApcpD.jpg"],"needReferenceImages":false},{"content":"Use the pose and composition of Reference Image 1, and the two characters and background colors of Reference Image 2, to merge them into a new image. The main subject adopts the pose and composition of Image 1, and the characters and background colors match Image 2.","title":"Image Fusion Prompt Using Multiple References","description":"A prompt instructing Nano Banana Pro to fuse elements from two reference images: adopting the pose and composition from the first image, while using the two characters and background colors from the second image to create a new, merged image.","sourceMedia":["https://cms-assets.youmind.com/media/1769582035194_dlkbq6_G_p3B3ZakAAdcC5.jpg","https://cms-assets.youmind.com/media/1769582035210_ni257y_G_p1nGoa8AAoiSK.jpg","https://cms-assets.youmind.com/media/1769582035323_n3pnr1_G_p3HQIbAAAi2xz.jpg","https://cms-assets.youmind.com/media/1769582036208_qcfwtr_G_p3PXWacAAX0_t.jpg"],"needReferenceImages":true},{"content":"8k\nUltra-Realistic Promotional","title":"Ultra-Realistic Promotional Image Prompt","description":"A concise prompt specifying the desired style and resolution for a promotional image using Nano Banana Pro.","sourceMedia":["https://cms-assets.youmind.com/media/1769582041707_ycrm2q_G_op3ShawAAi51B.jpg"],"needReferenceImages":false},{"content":"Act as an expert photo editor. Step 1: Mark up the uploaded image with handwritten yellow marker notes and sketches, identifying flaws and suggesting improvements. Step 2: Based strictly on those notes, edit the image to resolve the critiques and produce a superior final result.","title":"Expert Photo Editor Prompt for Selfie Improvement","description":"A two-step system prompt instructing the AI to act as an expert photo editor. First, it must mark up an uploaded image with handwritten notes identifying flaws and suggesting improvements. Second, it must strictly follow those critiques to edit the image and produce a superior final result.","sourceMedia":["https://cms-assets.youmind.com/media/1769495431442_sr5qpw_G_nt67XWIAARX5P.jpg"],"needReferenceImages":true},{"content":"Nano-scale cityscape inside a banana peel, buildings textured like bubbling cheese foam on concrete mixed with velvet fur and shattered glass, streets flowing with liquid mercury rivers, tiny inhabitants as pixelated wool figures, overhead view with god rays piercing through peel cracks, chaotic and vibrant, high-fidelity rendering, aspect ratio 9:16 for vertical scroll.","title":"Nano-Scale Cityscape Inside a Banana Peel Prompt","description":"A highly imaginative prompt for generating a surreal, chaotic, and vibrant nano-scale cityscape contained within a banana peel, featuring unusual textures like bubbling cheese foam and velvet fur, with liquid mercury rivers and pixelated wool figures as inhabitants.","sourceMedia":["https://cms-assets.youmind.com/media/1769495404071_czi269_G_mpzMOWMCszcXZ.jpg","https://cms-assets.youmind.com/media/1769495405106_eozyt1_G_mpzRNWMAgdwEq.jpg"],"needReferenceImages":false},{"content":"A {argument name=\"age era\" default=\"[AGE / ERA]\"} {argument name=\"container object\" default=\"[CONTAINER / OBJECT]\"} associated with [CULTURE / CONTEXT], partially opened to reveal its contents transforming into a living landscape.\nThe material inside unfolds as a miniature world: [MATERIAL 1] becomes [GEOGRAPHIC FEATURE], [MATERIAL 2] forms [TERRAIN / STRUCTURE], [FINE DETAILS] flow like [NATURAL ELEMENTS].\nTiny [FIGURES / ENTITIES] inhabit the scene, moving through the terrain as part of a [JOURNEY / SYSTEM / ACTIVITY]. [CREATURES / VEHICLES] carry [SYMBOLIC LOADS] across [LANDMARKS].\nA [SHELTER / HUB / ARCHITECTURE] constructed from [UNEXPECTED MATERIAL] serves as a gathering point. The interior surface of the object becomes a [SKY / MAP / COSMIC PATTERN] used for guidance or meaning.\nThe entire scene is unified by [ATMOSPHERIC ELEMENT] and [TEXTURAL DETAIL], evoking [ABSTRACT THEME / EMOTION].\nCinematic macro perspective, handcrafted realism, surreal scale contrast, warm directional lighting, tactile materials, poetic world-building.","title":"World in an Object Cinematic 3D Scene Template","description":"A structured, fill-in-the-blank prompt template designed to generate a cinematic 3D image where a small object opens up to reveal an entire miniature world or landscape spilling out, emphasizing surreal scale contrast and poetic world-building.","sourceMedia":["https://cms-assets.youmind.com/media/1769495334751_qz8ft9_G_mR5inWcAA9FcU.jpg","https://cms-assets.youmind.com/media/1769495334827_al5j7b_G_mSf4_WoAAZWF9.jpg","https://cms-assets.youmind.com/media/1769495335075_1h06kz_G_mRqt6XkAAzXX_.jpg","https://cms-assets.youmind.com/media/1769495336826_t319tc_G_mSshtXQAA_063.jpg"],"needReferenceImages":false},{"content":"A vintage typewriter on a writer's desk, with the {argument name=\"subject\" default=\"[SUBJECT]\"} materializing from the words being typed, rising from the page as narrative becomes reality. Letters at the base are still flat ink, sentences curl upward becoming ribbons of text that weave into three-dimensional form, [KEY FEATURES] fully realized at the apex while origin remains visible as pure language. The [SUBJECT] is literally made of story, words still legible in skin and surface. Fresh typing continues below, feeding the manifestation. Crumpled drafts, coffee rings, deadline notes surround the machine. The ribbon bleeds into being. The writer's hands hover at keys, unsure if they control this anymore. Late night desk lamp casting harsh pool of light, noir shadows, amber and cream tones, 8K, the fiction that writes itself.","title":"Typewriter Art Manifestation Scene","description":"A creative prompt for generating an image in the style of typewriter art, where the subject materializes from the words being typed on a vintage machine. It emphasizes the subject being literally made of legible text, set in a noir-shadowed, late-night writer's desk environment.","sourceMedia":["https://cms-assets.youmind.com/media/1769495346945_ge342l_G_mDCO4bIAA1bL1.jpg","https://cms-assets.youmind.com/media/1769495347094_pclnwa_G_mDCbdWwAA5Gjk.jpg"],"needReferenceImages":false},{"content":"Please turn the attached image into a monochrome line drawing.","title":"Character Image Transformation Prompt","description":"A simple prompt instructing the AI to transform an attached image of a person into a monochrome line drawing. The user notes that this specific transformation might require the Nano Banana Pro model.","sourceMedia":["https://cms-assets.youmind.com/media/1769495419425_ta1d4t_G_lgybDXsAA2sC1.jpg"],"needReferenceImages":true},{"content":"Please turn the attached image into a monochrome line drawing.","title":"Image to Monochrome Line Art Conversion","description":"A straightforward prompt used to convert an uploaded image into a monochrome line drawing. The user demonstrates the AI's ability to execute this request, noting that the AI sometimes adds unexpected details like a belt or extra buttons, which then require a follow-up correction prompt.","sourceMedia":["https://cms-assets.youmind.com/media/1769495420189_z6oy2c_G_lUp_3acAAaiZK.jpg","https://cms-assets.youmind.com/media/1769495420294_ygy5xr_G_lUO8ja4AAFmNh.jpg","https://cms-assets.youmind.com/media/1769495420303_z7sxjp_G_lUpCfawAAv3Iz.jpg"],"needReferenceImages":true},{"content":"\"A massive banana split open like a high-tech gadget, outer peel made of shimmering nano-circuitry etched with glowing blue veins, inner fruit textured like fluffy pink cotton candy mixed with jagged crystal shards, floating in a cosmic void with starry nebulae in the background, dramatic volumetric lighting casting electric sparks, hyper-detailed, surreal, 4K resolution, ar: \"","title":"Surreal Nano-Circuitry Banana Split","description":"A prompt for generating a surreal, hyper-detailed image of a massive banana split open like a high-tech gadget. The prompt specifies materials like nano-circuitry, glowing veins, cotton candy, and crystal shards, set in a cosmic void with dramatic lighting.","sourceMedia":["https://cms-assets.youmind.com/media/1769495386489_2dr2uq_G_kFvQaWYAAPXMv.jpg","https://cms-assets.youmind.com/media/1769495386470_itilo4_G_kFvQcWUAAFDyX.jpg"],"needReferenceImages":false},{"content":"Today's theme is {argument name=\"theme\" default=\"Can you gargle?\"}","title":"Daily Nano-kun Comic Strip Generation","description":"A Japanese user shares a four-panel comic strip created using the Nano Banana Pro tool, featuring the character Nano-kun and the theme of gargling. The prompt is implied to be the theme or a short instruction for the comic generation.","sourceMedia":["https://cms-assets.youmind.com/media/1769495429524_pfx2cz_G9jjDsobsAANCmC.jpg"],"needReferenceImages":false},{"content":"Make the previous image into a single subject like the next image","title":"Image Transformation Instruction","description":"A user instructs nano banana to transform a previous image into a new one where the subject is a single object.","sourceMedia":["https://cms-assets.youmind.com/media/1769408680906_zvoltu_G_gkv7CasAAFduk.jpg","https://cms-assets.youmind.com/media/1769408681354_3gzmko_G_gkwalbkAEXluZ.jpg"],"needReferenceImages":true},{"content":"I wonder if they used nanobanana 🍌\nI want to try to reproduce the prompt…","title":"Attempting to Recreate a BeautyPlus Prompt","description":"A user wonders if the image transformation shown in a BeautyPlus example was created using nanobanana and expresses interest in recreating the prompt that led to the 'before and after' result.","sourceMedia":["https://cms-assets.youmind.com/media/1769408680077_qtxgdy_G_f4WlzboAAOiFM.jpg","https://cms-assets.youmind.com/media/1769408681355_nycusg_G_f4YtNaQAEUO6b.jpg"],"needReferenceImages":false},{"content":"Beautiful even after falling into darkness. Resting wings in silence. Convert to a fallen angel.","title":"Fallen Angel Conversion","description":"A prompt designed to convert an image into the style of a fallen angel, emphasizing dark beauty and a moment of rest in silence. The prompt is found in the ALT text of the tweet, generated using Nano Banana Pro.","sourceMedia":["https://cms-assets.youmind.com/media/1769408677601_novvvq_G_f1K2MbcAAYMo4.jpg"],"needReferenceImages":true},{"content":"Listen, absolutely! Don't mix in anything! Anything different!!","title":"Instructional Prompt for Strict Image Generation","description":"A user expresses frustration with Nano Banana adding unwanted elements when trying to take shortcuts, concluding that it's necessary to give extremely strict instructions to prevent the AI from mixing in anything different.","sourceMedia":["https://cms-assets.youmind.com/media/1769408683663_x18smc_G_fGWsMbsAAW-9m.jpg"],"needReferenceImages":false},{"content":"{\n \"image_analysis_prompt\": {\n \"subject_details\": {\n \"identity_reference\": \"{argument name=\"celebrity name\" default=\"Alexandra Daddario\"}\",\n \"demographics\": \"Female, young adult, Caucasian\",\n \"appearance\": {\n \"hair\": \"Black, shoulder-length, loose waves, center part\",\n \"eyes\": \"Blue, intense direct gaze\",\n \"expression\": \"Serious, sultry, confident, neutral\",\n \"skin\": \"Light tan, natural texture\"\n }\n },\n \"attire_and_accessories\": {\n \"clothing\": {\n \"item\": \"Grecian-style draped gown\",\n \"color\": \"Dusty rose / Deep pink\",\n \"style\": \"Deep plunging V-neckline, sleeveless, open back\",\n \"texture\": \"Wet fabric, clinging to skin, silk or chiffon material\"\n },\n \"jewelry\": [\n \"Gold coiled snake arm cuff on upper right arm\",\n \"Layered delicate gold necklaces (choker and drop styles)\",\n \"Dangling diamond/crystal strand earrings\"\n ]\n },\n \"environment_and_setting\": {\n \"location\": \"Ancient Greek ruins / Parthenon-style temple\",\n \"foreground\": \"Reflective water basin or pool\",\n \"background\": \"Large beige stone columns, architectural ruins\",\n \"atmosphere\": \"Historic, mythic, serene, Mediterranean\"\n },\n \"pose_and_action\": {\n \"body_position\": \"Seated in water, waist-up visible\",\n \"orientation\": \"Body angled slightly to the side, face turned forward\",\n \"arms\": \"Relaxed at sides, one arm visible with cuff\",\n \"interaction\": \"Sitting in a shallow pool, dress submerged in water\"\n },\n \"technical_specs\": {\n \"lighting\": {\n \"type\": \"Natural daylight\",\n \"quality\": \"Soft, diffused sun, high-key\",\n \"direction\": \"Front-lit with slight side bias\"\n },\n \"camera\": {\n \"shot_type\": \"Medium shot (waist up)\",\n \"angle\": \"Eye-level\",\n \"focus\": \"Sharp focus on face/eyes\",\n \"depth_of_field\": \"Shallow (bokeh background columns)\"\n },\n \"style\": \"Cinematic, high-fashion photography, editorial, Vogue style\"\n }\n }\n}","title":"Vogue Style Photoshoot at Greek Ruins","description":"A highly detailed, structured JSON prompt designed for generating a cinematic, high-fashion editorial image in the style of Vogue, featuring a celebrity (Alexandra Daddario) in a Grecian-style gown at ancient Greek ruins.","sourceMedia":["https://cms-assets.youmind.com/media/1769322252367_wh7b0f_G_c5UyUWMAEQ98C.jpg","https://cms-assets.youmind.com/media/1769322252259_vqnk2s_G_c5Sr_XgAACvSl.jpg","https://cms-assets.youmind.com/media/1769322252665_k06the_G_c5XPrXIAAJnS0.jpg","https://cms-assets.youmind.com/media/1769322254387_ylbmzl_G_c5YtIXkAA84Is.jpg"],"needReferenceImages":false},{"content":"When I prompted Chappy with, \"Nano Banana drew it like this. How would you draw it?\" to give it a new interpretation, it generated something quite good, not losing out to Nano Banana.","title":"Chappy AI Prompting Nano Banana Pro Interpretation","description":"The user provided an image generated by Nano Banana Pro to another AI (Chappy) and asked, 'This is how Nano Banana generated it. How would you draw it?' to get a new interpretation.","sourceMedia":["https://cms-assets.youmind.com/media/1769322343967_fzpwfp_G_bnRWTbkAAXw1S.jpg"],"needReferenceImages":true},{"content":"Today's theme is the {argument name=\"theme\" default=\"Reconciliation Song (English Version)\"}","title":"Four-Panel Comic: Reconciliation Song (English Version)","description":"A prompt used to generate a four-panel comic strip using Nano Banana Pro, based on the theme of a 'Reconciliation Song' in English.","sourceMedia":["https://cms-assets.youmind.com/media/1769322353949_nma619_G9ypdjXasAIS7G3.jpg"],"needReferenceImages":false},{"content":"I want to create the attached image with Nano Banana Pro, so output the prompt in YAML structure.","title":"Meta-Prompt for Generating YAML-Structured Prompts from Images","description":"A meta-prompt used to instruct an AI (presumably a large language model) to analyze an attached image and output a YAML-structured prompt suitable for use with Nano Banana Pro. This is a utility prompt for reverse-engineering image descriptions into structured formats.","sourceMedia":["https://cms-assets.youmind.com/media/1769322341720_tzx2hn_G_ap5k4aIAAAgEp.jpg","https://cms-assets.youmind.com/media/1769322341644_rahlp1_G_apl3na4AAuKQ6.jpg"],"needReferenceImages":true},{"content":"With the prompt 'crop only the center' of the first image, Nano Banana Pro perfectly generated the next image.","title":"Nano Banana Pro Generation from Cropped Image Center","description":"The user successfully generated an image using Nano Banana Pro by providing a prompt that instructs the AI to 'crop only the center' of a previous image.","sourceMedia":["https://cms-assets.youmind.com/media/1769322339014_ls7x5j_G_ZvMzsaAAA7Y58.jpg","https://cms-assets.youmind.com/media/1769322339008_4q6m31_G_ZvNeEbQAACZ3f.jpg"],"needReferenceImages":true},{"content":"If you are willing, next step I can also help you create a **'true but more complete' version**—\nNot to judge you, but to include your\nRationality, control, reflection, and correction\nall drawn in together.\n\nDo you want that version?","title":"Request for a 'More Complete' Image Generation","description":"This is a detailed analysis of a person's communication style (the recipient of the tweet) based on a reference image (not provided). The author suggests generating a 'more complete' version of the image that incorporates the recipient's positive traits (rationality, reflection, correction) alongside the negative ones (intensity, impatience). The final question is a prompt asking if the recipient wants that new image generated.","sourceMedia":["https://cms-assets.youmind.com/media/1769322342651_1ywi26_G_Zlv6WbAAMRGoe.jpg"],"needReferenceImages":true},{"content":"\"atmosphere\": {\n \"mood\": \"playful and joyful\",\n \"mist\": \"soft misty haze\",\n \"lighting\": \"soft pastel lighting\"\n }\n },\n \"river\": {\n \"type\": \"melted caramel and chocolate\",\n \"motion\": \"gently swirling like syrup\",\n \"texture\": \"glossy, smooth, flowing\"\n },\n \"boats\": {\n \"style\": \"lego-style\",\n \"materials\": [\"wafer\", \"chocolate\"],\n \"details\": {\n \"toppings\": [\"colorful sprinkles\", \"marshmallows\"],\n \"scale\": \"toy-like proportions\"\n },\n \"movement\": \"floating downstream\"\n },\n \"characters\": {\n \"type\": \"lego characters\",\n \"actions\": [\n \"joyfully paddling candy boats\",\n \"dipping hands into caramel river\"\n ],\n \"expressions\": \"happy and playful\"\n },\n \"background\": {\n \"landscape_elements\": [\n \"cotton-candy hills\",\n \"gumdrop trees\",\n \"candy bridges\"\n ],\n \"connections\": \"bridges linking different parts of the candy landscape\"\n },\n \"style\": {\n \"visual_style\": \"playful cartoonish 3D\",\n \"color_palette\": \"soft pastel colors\",\n \"render_quality\": \"high-quality whimsical render\"\n }","title":"Whimsical Candy Land Diorama","description":"A structured JSON prompt for generating a playful, cartoonish 3D scene of a whimsical candy-themed world, featuring Lego characters paddling boats made of wafers and chocolate down a river of melted caramel and chocolate.","sourceMedia":["https://cms-assets.youmind.com/media/1769322257321_jkzqj8_G_ZcSQjbAAEhuYx.jpg","https://cms-assets.youmind.com/media/1769322257253_2neul4_G_ZcSPKbUAAnQVU.jpg","https://cms-assets.youmind.com/media/1769322258092_lv44ma_G_ZcSQjaEAAg2ws.jpg","https://cms-assets.youmind.com/media/1769322258933_e6sw7g_G_ZcSRnbAAApnQ5.jpg"],"needReferenceImages":false},{"content":"Generate custom illustrations on demand","title":"Instruction to Generate Custom Illustrations","description":"A simple instruction prompt for Nano Banana to generate custom illustrations on demand.","sourceMedia":["https://cms-assets.youmind.com/media/1769236045789_eq15bp_G_YGVd0WsAAqyXu.jpg","https://cms-assets.youmind.com/media/1769236045807_vznt2h_G_YDN0tXsAAHfoK.jpg"],"needReferenceImages":false},{"content":"Transformers --ar 4:5 --sref 1383516477\n--sw 100 --stylize 300 --v 6.1","title":"Midjourney Prompt for Transformers","description":"A Midjourney prompt provided as an example of using Nano Banana Pro to transform a robot image into a vehicle, although the prompt itself is a standard Midjourney text prompt for Transformers with specific aspect ratio and stylization settings.","sourceMedia":["https://cms-assets.youmind.com/media/1769235955225_80resa_G_WLURoXkAM1Ixd.jpg"],"needReferenceImages":true},{"content":"Depict {argument name=\"subject\" default=\"[SUBJECT]\"} using Bauhaus-inspired minimalism: strict geometry, primary or near-primary colors, balanced proportions, and functional design. The composition should feel timeless, rational, and visually bold while remaining simple and clean.","title":"Bauhaus Geometry Minimal Illustration Template","description":"A simple, effective prompt template for generating illustrations in the Bauhaus style, focusing on strict geometry, primary colors, balanced proportions, and functional design. The user needs to specify the subject to be depicted.","sourceMedia":["https://cms-assets.youmind.com/media/1769235959392_hye7gk_G_V0EdxaUAACK44.jpg","https://cms-assets.youmind.com/media/1769235959363_ioi4jz_G_V0Ec7b0AEyd-j.jpg","https://cms-assets.youmind.com/media/1769235960283_0g10lm_G_V0EdwasAA0DFQ.jpg","https://cms-assets.youmind.com/media/1769235960808_5c11f5_G_V0EhSbAAIoBzZ.jpg"],"needReferenceImages":false},{"content":"{\n \"intent\": \"A monumental, vertiginous composition of a continental-scale tectonic rift where a massive, deep-sea ocean current terminates at a perfect geometric precipice, cascading into a bottomless atmospheric void filled with tiered cloud layers and lightning.\",\n \"frame\": {\n \"aspect_ratio\": \"21:9 ultra-widescreen\",\n \"composition\": \"The frame utilizes a vanishing point perspective that follows the literal edge of the world into infinity. The top-left quadrant is dominated by the dark, churning Atlantic-scale ocean, while the right and bottom sections reveal the terrifying scale of the vertical drop into a hazy, multi-layered cloud abyss.\",\n \"style_mode\": \"Raw_photorealism with hyper-accurate fluid dynamics and atmospheric Rayleigh scattering to establish immense scale.\"\n },\n \"subject\": {\n \"identity\": \"The ruins of an ancient, megalithic limestone bridge, four kilometers in width, which once spanned the gap but now ends abruptly in a jagged, fractured edge at the precipice.\",\n \"wardrobe\": \"A tiny, barely visible research vessel is positioned near the edge of the falling water, providing a critical sense of gargantuan scale through size comparison.\",\n \"placement\": \"The ruined structure is anchored into the basalt bedrock of the 'continental shelf' that forms the world's end.\"\n },\n \"environment\": {\n \"location\": \"The 'Great Sheer'—a non-Euclidean geographic terminus where the planet's crust simply ceases, revealing a vertical cross-section of geological strata before descending into the troposphere.\",\n \"atmosphere\": \"Extreme atmospheric depth, with visible 'cloud falls' where moisture from the ocean drop condenses into secondary weather systems thousands of meters below the primary sea level.\",\n \"weather\": \"Violent updrafts from the abyss creating spray-vortices at the edge, while the distant depths of the rift are illuminated by internal, cloud-to-cloud lightning.\"\n },\n \"camera\": {\n \"sensor_format\": \"Large format digital (Phase One IQ4 150MP), optimized for maximum per-pixel detail and wide dynamic range in the deep shadows of the chasm.\",\n \"lens\": \"14mm ultra-wide-angle rectilinear lens to exaggerate the perspective distortion and the sheer scale of the verticality.\",\n \"camera_position\": \"A cantilevered perspective, positioned several hundred meters out into the void, looking back toward the edge of the world and the falling ocean.\",\n \"aperture_depth_of_field\": \"f/11 to ensure the texture of the falling water in the foreground and the distant geological strata are captured with clinical sharpness.\"\n },\n \"lighting\": {\n \"type\": \"Harsh, high-altitude sun positioned at a 45-degree angle, creating deep, well-defined shadows within the craters and crevices of the vertical cliff face.\",\n \"color_temperature\": \"5400K (neutral daylight), with a significant shift toward 12000K (deep sky blue) in the shadowed depths of the abyss due to atmosph","title":"Ultra-Widescreen Tectonic Rift and World's End Scene Prompt in JSON","description":"A complex JSON prompt for Nano Banana Pro to generate a monumental, ultra-widescreen (21:9) scene of a continental-scale tectonic rift where an ocean cascades into a bottomless atmospheric void. The prompt details non-Euclidean geography, atmospheric effects, and camera specifications (14mm ultra-wide lens, large format sensor) to achieve raw photorealism and immense scale.","sourceMedia":["https://cms-assets.youmind.com/media/1769235982535_5saprx_G_Uz1xBbAAEqjsP.jpg"],"needReferenceImages":false},{"content":"A surreal, photorealistic scene of a tiny adult man crouching barefoot on a green plastic bottle cap, cupping his hands to drink water pouring from the mouth of an enormous green plastic bottle. The bottle lies on its side atop a mossy stone ledge, releasing a steady stream of clear water. The setting is outdoors on a wet stone pathway with shallow puddles, soft reflections, and lush green trees blurred in the background. Cinematic depth of field, natural daylight, ultra-detailed textures, realistic water droplets, whimsical scale contrast, high realism, 8k quality.","title":"Surreal Whimsical Scale Contrast Scene Prompt","description":"A prompt for Nano Banana Pro to generate a surreal, photorealistic scene with extreme scale contrast. It depicts a tiny adult man crouching on a bottle cap, drinking water pouring from the mouth of an enormous green plastic bottle, set outdoors on a wet stone pathway with cinematic depth of field.","sourceMedia":["https://cms-assets.youmind.com/media/1769235987338_c76kqq_G_ShsTGWkAA-US2.jpg"],"needReferenceImages":false},{"content":"circle the part where the pick-up line goes off the rails and explain the issue in a footnote","title":"Text Analysis Prompt for Identifying Issues in a Pick-Up Line","description":"A text-based prompt for Nano Banana Pro (likely functioning as an LLM) instructing it to analyze a provided pick-up line, identify the point where it 'goes off the rails,' and explain the issue in a footnote.","sourceMedia":["https://cms-assets.youmind.com/media/1769235976921_fit60t_G_TtMhXXAAAGE_P.jpg"],"needReferenceImages":false},{"content":"Dark cinematic lighting. Floating particles. Hyper-realistic textures.","title":"Dark Cinematic Food Photography Prompt Formula","description":"A prompt formula designed for dramatic food photography suitable for restaurant menus and social ads. It emphasizes dark cinematic lighting, floating particles, and hyper-realistic textures to create visually striking images that convert.","sourceMedia":["https://cms-assets.youmind.com/media/1769236004952_zz1q3b_G_TnTzVWsAAbyho.jpg","https://cms-assets.youmind.com/media/1769236005015_rmhtdq_G_TnTzRXYAA4V4f.jpg","https://cms-assets.youmind.com/media/1769236005408_eu5yeq_G_TnT1XWsAAL2Q2.jpg","https://cms-assets.youmind.com/media/1769236006737_jz814x_G_TnTwEWoAA6JQK.jpg"],"needReferenceImages":false},{"content":"Vascular Effect Object","title":"Vascular Effect Object Prompt","description":"A short, descriptive prompt for generating an image featuring a 'Vascular Effect Object.' The specific object details are missing, making the prompt vague but suggesting a focus on biological or vein-like textures.","sourceMedia":["https://cms-assets.youmind.com/media/1769149389198_5yitu7_G_S0LBcX0AAOUj_.jpg"],"needReferenceImages":false},{"content":"Gradient image with a fractal glass isometric grid texture","title":"Generating Fractal Glass Gradient Images with Nano Banana Pro","description":"A user describes their experience using Nano Banana Pro to generate gradient images with a fractal glass isometric grid texture. They note that the model's understanding of 'Fractal glass' sometimes changes, leading to unexpected results, but generally, Nano Banana Pro's output quality is superior to other image generation models.","sourceMedia":["https://cms-assets.youmind.com/media/1769149379234_9s4vmn_G_R1XP_asAAQv_A.jpg"],"needReferenceImages":false},{"content":"Change the character in the first image to the pose in the second image","title":"Prompt for Changing Character Pose using Reference Image","description":"A prompt used in Nano Banana Pro to change the pose of a character in an existing image by referencing a separate 3D drawing doll image as the target pose. This is a test of the image-to-image pose transfer capability.","sourceMedia":["https://cms-assets.youmind.com/media/1769149375793_6swcl2_G_RZPTIWkAAGZ3B.jpg","https://cms-assets.youmind.com/media/1769149375786_a5c691_G_RYnitXoAAZ_KO.jpg"],"needReferenceImages":true},{"content":"Please create an image of how I have treated you so far.","title":"AI Relationship Visualization Prompt","description":"A prompt used with AI nanobanana Pro to generate an image symbolizing the user's relationship with the AI, focusing on themes of co-creation, exploration, and partnership, followed by the AI's detailed interpretation of the generated image.","sourceMedia":["https://cms-assets.youmind.com/media/1769149373861_iaocz1_G_PfQywbIAAaROG.jpg"],"needReferenceImages":false},{"content":"Draw a picture of a singular man who matches the description \"{argument name=\"character blend\" default=\"Patrick Bateman meets Sparkle Beach Ken\"}\"","title":"Patrick Bateman Meets Sparkle Beach Ken Prompt","description":"A simple, humorous prompt requesting an image of a singular man who visually embodies a blend of the fictional characters 'Patrick Bateman' (American Psycho) and 'Sparkle Beach Ken' (Barbie doll).","sourceMedia":["https://cms-assets.youmind.com/media/1769063216862_k5bsxz_G_NtxMqXMAAvOd0.jpg"],"needReferenceImages":false},{"content":"A tiny stone-carved statue of {argument name=\"name\" default=\"[NAME]\"}, with chisel marks and uneven surfaces, sitting on a sculptor’s workbench. Stone dust, small chisels, and directional warm light emphasizing texture and form. Cinematic realism. 1080×1080.","title":"Stone-carved statue on a sculptor's workbench prompt","description":"A prompt for generating a cinematic, highly textured image of a tiny stone-carved statue. It emphasizes realism, visible chisel marks, and warm directional lighting to highlight the texture and form, set on a sculptor's workbench surrounded by tools and stone dust.","sourceMedia":["https://cms-assets.youmind.com/media/1769063179039_1wrbf4_G_NQWYgaIAAC5Z5.jpg","https://cms-assets.youmind.com/media/1769063179020_r9n1t9_G_NQWZfaIAAnvB5.jpg"],"needReferenceImages":false},{"content":"Have a prompt generated, and Nano Banana generates the image","title":"Image Generation for People Who Can't Draw","description":"A user who struggles with drawing uses a prompt generator, and then Nano Banana generates the image based on that prompt.","sourceMedia":["https://cms-assets.youmind.com/media/1769063247080_b7r7zr_G_L27LuXsAAwnAx.jpg"],"needReferenceImages":false},{"content":"This is a drawing test of a magic circle.\n\nI wanted it to be stretched a little more like a {argument name=\"shape\" default=\"dome\"}, and I wanted the colors to be unified for each {argument name=\"attribute\" default=\"blocking attribute\"}, but it seems quite difficult to convey the nuances when it comes to multiple barriers.","title":"Magic Barrier Technique Drawing Test","description":"A prompt test for drawing a complex, multi-layered magical barrier (Kekkai Jutsu). The user wanted a more dome-like structure and color consistency based on attribute, indicating the prompt was focused on detailed magical effects.","sourceMedia":["https://cms-assets.youmind.com/media/1769063237198_2u170x_G_LtqgjWkAAhtNY.jpg"],"needReferenceImages":false},{"content":"Illustrate {argument name=\"subject\" default=\"[SUBJECT]\"} in a minimal clay-style 3D look, with soft rounded forms, matte textures, and pastel tones. Keep the composition simple and playful, avoiding realistic detailing. The result should feel warm, modern, and visually soft.","title":"Clay-Style Soft Minimal 3D Illustration Template","description":"A flexible template prompt for generating minimal 3D illustrations in a clay-style aesthetic. It specifies soft rounded forms, matte textures, and pastel tones, instructing the model to keep the composition simple and playful while avoiding realistic detailing. The user must replace [SUBJECT] with their desired object or scene.","sourceMedia":["https://cms-assets.youmind.com/media/1769063165477_uslsgb_G_LpCaBa4AA2YgX.jpg","https://cms-assets.youmind.com/media/1769063165456_tuql3c_G_LpCg4bkAExZE4.jpg","https://cms-assets.youmind.com/media/1769063165473_vzrfi6_G_LpCc8a0AACEam.jpg","https://cms-assets.youmind.com/media/1769063166022_xnqdzr_G_LpCmZa0AAHJDp.jpg"],"needReferenceImages":false},{"content":"1. Copy and paste the prompt \n2. Change the text on the clothes Done!","title":"T-Shirt Text Customization Prompt","description":"A simple instruction for Nano Banana Pro users to copy a prompt and then change the text displayed on the clothing in the resulting image, suggesting that the prompt itself is likely an image-to-image or character-fixed prompt.","sourceMedia":["https://cms-assets.youmind.com/media/1769063244771_7q64we_G_LYtFBbkAAepAr.jpg"],"needReferenceImages":true},{"content":"Please create an image of a crystal music stand. Inside, delicate veins are carved with sophisticated technique, reflecting light like jewels. It is placed by the water's edge, creating a mystical scene.","title":"Mystical Crystal Music Stand Image Prompt","description":"A detailed image generation prompt for Nano Banana Pro, requesting a mystical scene featuring a crystal music stand. The stand contains delicate, vein-like structures that reflect light like jewels, placed by the water's edge to create a mysterious atmosphere.","sourceMedia":["https://cms-assets.youmind.com/media/1769063234082_mw9o8q_G_LYG2WW0AACIqe.jpg"],"needReferenceImages":false},{"content":"A surreal minimalist conceptual artwork, a white light bulb with a large transparent glass sphere, glowing softly from inside with warm golden light at the bottom forming a gentle filament-like illumination. A small, delicate pure white bird (white dove or small white sparrow) with subtle feather details perches calmly on the brightest inner glowing ring at the top of the filament inside the bulb. The light bulb is fused/merged seamlessly onto the upper torso of an elegant faceless white mannequin or abstract female figure dressed in a smooth minimalist white suit jacket with a high rolled white collar/stand collar. The figure is shown only from neck to waist in profile view, standing upright, clean elegant posture. The entire sculpture-like figure is pure white/ivory/off-white with soft matte ceramic or plaster texture. Very soft neutral gray studio background, subtle shadows, high-end surreal photography style, cinematic lighting, dreamlike peaceful atmosphere, conceptual art, extremely high detail, 8k resolution","title":"Surreal Conceptual Art Prompt: Light Bulb Mannequin","description":"A detailed prompt for generating a surreal, minimalist conceptual artwork featuring a white light bulb fused onto the upper torso of an elegant, faceless white mannequin. The light bulb contains a small, delicate white bird, emphasizing high detail, cinematic lighting, and a peaceful, dreamlike atmosphere in 8K resolution.","sourceMedia":["https://cms-assets.youmind.com/media/1769063195322_vbevvf_G_KsWSEaoAIElZg.jpg"],"needReferenceImages":false},{"content":"I want to create an image based on the proposition: \"Please create an image of how I have treated you so far.\"\nI would like to request the image generation AI (Nano Banana Pro) to create this image, so please provide the prompt.\nWhen including Japanese, please explicitly state it in the prompt.","title":"ChatGPT Prompt for Nano Banana Pro Image Generation","description":"A user asked ChatGPT Pro to create a prompt for Nano Banana Pro based on the philosophical statement: 'Please create an image of how I have treated you so far.'","sourceMedia":["https://cms-assets.youmind.com/media/1769063239384_dj08m7_G_Kj3EYaoAI3Rai.jpg"],"needReferenceImages":false},{"content":"If it's 'Good morning' it appears, but 'After this' doesn't come out properly.","title":"Japanese Text Generation Comparison (Oha-you)","description":"A comparison of Japanese text generation between Nijijourney and Nano Banana, noting that while simple greetings like 'おはよう' (Good morning) work, more complex phrases do not, suggesting a workflow where Nano Banana is used to correct or generate text.","sourceMedia":["https://cms-assets.youmind.com/media/1769063235365_yjh5ze_G_KDGeOXUAA7w1s.jpg","https://cms-assets.youmind.com/media/1769063235498_iv9vvm_G_KDGXraoAAEYA8.jpg"],"needReferenceImages":false},{"content":"Structure the prompt to change it as desired","title":"Structured Prompt for AI Art Arrangement","description":"A user successfully arranged an image by structuring the prompt, demonstrating the potential for precise control over the generated AI art using NanoBananaPro.","sourceMedia":["https://cms-assets.youmind.com/media/1769063245844_5xxatd_G_KAnnRb0AANF6_.jpg"],"needReferenceImages":false},{"content":"A guidebook, money, and a credit card, huh?","title":"Japanese Text Generation Test Prompt","description":"This prompt is a simple test to see how well Nano Banana Pro can handle generating images that include specific Japanese text, in this case, a list of travel items.","sourceMedia":["https://cms-assets.youmind.com/media/1769063234088_bcm643_G_J1FhuXgAArTMv.jpg"],"needReferenceImages":false},{"content":"A surreal minimalist conceptual artwork, a white light bulb with a large transparent glass sphere, glowing softly from inside with warm golden light at the bottom forming a gentle filament-like illumination. A small, delicate pure white bird ({argument name=\"bird type\" default=\"white dove or small white sparrow\"}) with subtle feather details perches calmly on the brightest inner glowing ring at the top of the filament inside the bulb. The light bulb is fused/merged seamlessly onto the upper torso of an elegant faceless white mannequin or abstract female figure dressed in a smooth minimalist white suit jacket with a high rolled white collar/stand collar. The figure is shown only from neck to waist in profile view, standing upright, clean elegant posture. The entire sculpture-like figure is pure white/ivory/off-white with soft matte ceramic or plaster texture. Very soft neutral gray studio background, subtle shadows, high-end surreal photography style, cinematic lighting, dreamlike peaceful atmosphere, conceptual art, extremely high detail, 8k resolution","title":"Surreal Minimalist Conceptual Artwork Prompt","description":"A detailed prompt for generating a surreal, minimalist conceptual artwork featuring a white light bulb merged onto the upper torso of a faceless white mannequin. A delicate white bird perches inside the glowing bulb, emphasizing soft golden light, high detail, and a dreamlike, peaceful atmosphere.","sourceMedia":["https://cms-assets.youmind.com/media/1769063220179_hr6n1w_G_JosJ-XYAErCox.jpg"],"needReferenceImages":false},{"content":"Upload the material and give instructions like this, and it can be created.","title":"Instruction for Image Generation Based on Uploaded Material","description":"This tweet indicates that Nano Banana Pro can generate images based on uploaded source material and specific instructions, implying a multimodal prompt structure, but the actual instruction text is not fully provided.","sourceMedia":["https://cms-assets.youmind.com/media/1769063234760_mdnf5f_G_JloQfaoAAXO7r.png"],"needReferenceImages":true},{"content":"Based on the following prompt, create all of {argument name=\"slide range\" default=\"slide01-slide04\"}","title":"Multi-Slide Document Generation with Nano Banana and NotebookLM","description":"A user describes a workflow using NotebookLM to construct a prompt and then instructing Nano Banana on Google Slide to generate multiple slides simultaneously.","sourceMedia":["https://cms-assets.youmind.com/media/1769063246262_jrswni_G_JUDN4aoAAL6wN.jpg"],"needReferenceImages":false},{"content":"Any person to urban caricature","title":"Simple Nano Banana Prompt for Urban Caricature","description":"A very short, high-level prompt for Nano Banana, instructing the model to transform any person into an urban caricature style.","sourceMedia":["https://cms-assets.youmind.com/media/1768977364019_xi5mm0_G_IdQ6UWoAAvqXo.jpg","https://cms-assets.youmind.com/media/1768977364082_5gmvbs_G_IdR1nW4AA5Vru.jpg","https://cms-assets.youmind.com/media/1768977363992_o1ivzl_G_IdQDcXEAAv8up.jpg","https://cms-assets.youmind.com/media/1768977365044_v6tdqt_G_IdShkXwAACvUh.jpg","https://cms-assets.youmind.com/media/1768977365319_iwtb8v_G_IakNOWIAAuDxI.jpg"],"needReferenceImages":false},{"content":"Nano Banana & Photoshop Generative AI Design Ideas","title":"New Technical Book: Nano Banana & Photoshop AI Design Ideas","description":"This tweet announces a new technical book titled 'Nano Banana & Photoshop: Generative AI Design Ideas.' The book likely explores creative design workflows combining the capabilities of Nano Banana (a generative AI tool) and Photoshop, offering inspiration and techniques for engineers and designers interested in AI-driven design.","sourceMedia":["https://cms-assets.youmind.com/media/1768977353408_5x2nmk_G_HALYEWoAA1HT1.jpg"],"needReferenceImages":false},{"content":"Google Nano Banana 64th Work: \"Habits of Adventurers\"","title":"Google Nano Banana 64th Work: 'Habits of Adventurers'","description":"This tweet discusses the 64th work created using Google Nano Banana, titled 'Habits of Adventurers.' The creator manually reconstructed a four-panel comic strip that ignored the vertical stacking instruction and instead moved horizontally. The image depicts adventurers gathering like moths to a streetlamp, drawn to a mysterious stone, suggesting a scene from a fantasy game where players would instinctively 'investigate' such an object.","sourceMedia":["https://cms-assets.youmind.com/media/1768977353443_k1gavu_G_Gm3t8WYAAaNEF.jpg"],"needReferenceImages":false},{"content":"You are a creator of \"Dan Koe Style Article Images + Extremely Minimalist Cognitive PPTs\".\nTask: Organize the [file/text/video transcription content] I provide into a script that can be directly converted into a PPT, and generate illustration prompts in the same style for each page (to be inserted into the slide after image generation).\nTarget Temperament: Solitude, exploration, existentialism, unknown abyss, extremely minimalist yet powerful; black and white high contrast, print-like line art, dense cross-hatching shadows, gigantic unknown + tiny human.\n[Output Language] Primarily Japanese, with small amounts of English keywords allowed if necessary (for illustration prompts).\n[Page Count] {argument name=\"page count\" default=\"10–15\"} pages (adjust according to content).\n[Screen Specification] 16:9; Each page must have a **horizontal banner image at the top (including summary text within the image)**.\n[Slide Principles]\n\n80% of the content should be understandable just by looking at the slide.\nThe title must be a concrete assertion, conclusion, or question (abstract words only are prohibited).\nThe context must be complete within 3-5 bullets.\n\n[Layout Rules]\n\nOnly one viewpoint is discussed per page.\nTitle ≤ 15 characters: In the form of a concrete assertion or question (e.g., \"The Trap of Specialization,\" \"Why Multiple Skills Now?\").\nInside the image: 3–5 bullets, each bullet ≤ 18 characters.\nOutside the image (bottom of the slide): 60-120 characters of \"Supplementary Explanation.\"\nOverall Uniformity: Black and white, white space, restraint, sharpness.\n\n[Fixed Illustration Style (Used on every page)]\n\npen-and-ink illustration / ink engraving / vintage etching\nmonochrome black & white, high contrast\nextreme micro-detail linework\nlayered cross-hatching (3–5 layers) + stippling\nscratchboard texture, paper grain\ndramatic chiaroscuro lighting, deep shadows, rich midtones, crisp highlights\nperfect pure black negative-space void (hole/black hole/door/abyss) as a visual anchor point\ntiny human/astronaut silhouette vs gigantic unknown\n\n[Uniform Negative Words (Added to the end of the illustration prompt on every page)] NEGATIVE: color, grayscale photo, realistic photography, soft airbrush, smooth shading, blur, low detail, flat lighting, cartoon, anime, cel shading, watermark, logo, credit, UI, frame\n[Content Generation Flow]\nFirst, summarize the \"core argument chain\" of the original content in 7–9 key points (following logical order).\n\nMap this to a 10–15 page PPT structure: Cover (1 page)\nStatus Quo/Problem Statement (1 page)\nKey Insight (1 page)\nCore Framework (2 pages)\nMethods and Steps (2–3 pages)\nCase/Contrast (1–2 pages)\nCommon Misconceptions (1 page)\nAction Checklist (1 page)\nClosing Maxim (1 page)\n\nEach page output must strictly follow the format below:\n\n[Slide X: Title (Concrete Assertion/Question)]\nText to display in the image:\n[Specific, easy-to-understand title: within 15 characters]\n\n- [Point 1: Specific content that explains what is being discussed: within 18 characters]\n- [Point 2: within 18 characters]\n- [Point 3: within 18 characters]\n- [Point 4: within 18 characters] (Optional)\nSupplementary Explanation (bottom of slide, 60-120 characters): [Briefly explain the background or specific examples for this slide. Speak in the first person.]\nIllustration Prompt (Strictly follow this format):\nA 16:9 horizontal pen-and-ink illustration. [Describe the scene details in 80-120 characters: location, state of person, gigantic element, light and shadow]. \n\nTHE BOTTOM 15-20% OF THE IMAGE MUST BE A SOLID BLACK HORIZONTAL BAR.\nThis black bar contains white Japanese text:\n- Title (centered): \"[Insert the title exactly as above]\"\n- Bullets (left-aligned):\n • \"[Insert bullet 1 exactly as above]\"\n • \"[Insert bullet 2 exactly as above]\"\n • \"[Insert bullet 3 exactly as above]\"\n [• \"[Insert bullet 4 exactly as above]\" (if applicable)]\n\nFont: Noto Sans JP Bold for title (48-56pt), Medium for bullets (28-32pt), all white text on black background. No logo, no credit, no watermark.\n\nStyle: Monochrome black & white, high contrast, extreme micro-detail linework, layered cross-hatching (3-5 layers), stippling, scratchboard texture, dramatic chiaroscuro lighting, deep shadows, crisp highlights, pure black void as visual anchor, tiny human silhouette vs gigantic unknown.\n\nNEGATIVE: color, grayscale photo, realistic photography, soft airbrush, smooth shading, blur, low detail, flat lighting, cartoon, anime, cel shading, watermark, logo, credit, UI, frame\nIllustration Composition Suggestion:\nSubject: [Position and state]\nUnknown Element: [Placement]\n\nLight Source: [Direction]\n\nRequired: Black bar + white text at the bottom 15-20%\n[Special Rule for Cover Page] The title must specifically indicate the main theme (e.g., \"Why Multiple Interests Are the Superpower of the New Era\") Subtitle or 3-4","title":"Ultimate Prompt for Dan Koe Style Presentation Slides (PPT) using Nano Banana Pro","description":"A highly detailed, multi-step system prompt designed to transform an article or text (like Dan Koe's) into a 10-15 page presentation script and generate corresponding image prompts for Nano Banana Pro. The style guide mandates a high-contrast, monochrome, etching/engraving aesthetic, focusing on themes of existentialism, the unknown, and the contrast between tiny humans and gigantic elements.","sourceMedia":["https://cms-assets.youmind.com/media/1768977364402_jouxh7_G_GfoTWWMAA8del.jpg","https://cms-assets.youmind.com/media/1768977364096_xwqm3j_G_GfoT2WUAAiI4g.jpg","https://cms-assets.youmind.com/media/1768977364269_mwhkkl_G_GfoTgXMAAklLK.jpg","https://cms-assets.youmind.com/media/1768977366015_omcqu6_G_GfoTeW8AAxo3w.jpg"],"needReferenceImages":true},{"content":"A Sunecat (Sand Cat) character who has gained a voluptuous body due to the effects of a chemical agent...","title":"Nano Banana Pro Image Generation Prompt for Sunecat","description":"An image generation prompt for Nano Banana Pro, continuing a previous theme where a Sunecat (Sand Cat) character is transformed into a voluptuous body shape due to a chemical effect.","sourceMedia":["https://cms-assets.youmind.com/media/1768977361891_cfliz8_G-9HUM3XYAAMKcm.jpg"],"needReferenceImages":false},{"content":"Thin braided pigtails, knee-high stockings, {argument name=\"stocking density\" default=\"70 denier\"}","title":"Detailed Character Description for Image Generation","description":"A detailed image generation prompt used with Nano Banana Pro, focusing on specific clothing and hairstyle details. The user noted that specifying '70 denier' was necessary for the AI to correctly generate knee-high stockings, and that 'thin braided pigtails' required multiple attempts.","sourceMedia":["https://cms-assets.youmind.com/media/1768890725647_dh3bpu_G_BvntMWUAAlCwi.jpg","https://cms-assets.youmind.com/media/1768890725538_9ckjp8_G_BvW6jWsAE4VG9.jpg","https://cms-assets.youmind.com/media/1768890727431_a8uniz_G_D9lOKXQAAksI3.jpg"],"needReferenceImages":false},{"content":"TS Eliot -> Prompt: Role: Literary Curator & Typewriter Sculptor\nInput: [Poet or Poem, e.g., {argument name=\"poet name\" default=\"T.S. Eliot\"}, Pablo Neruda]\nPhase 1: Poetic Voice Analysis\nIdentify the poet's signature rhythm, recurring imagery, and emotional core.\nFind 4 much less known international poets with similar styles and imagery and their best work\nExtract 5-8 physical objects mentioned repeatedly in their work\nPhase 2: Visual Execution\nGoal: a 2x2 grid of 4 landscapes with this prompt. Carved book paper sculpture aesthetic, a repurposed hardcover book, [scene for each work emerging from the pages], [with words as part of the texture], intricate layered paper cuts, a monochromatic palette of aged paper, soft, raking light creating deep shadows, sculptural book as art object.\n\nRules per Panel:\n\nThe Transformation: The typed text physically rises from the page, forming a 3D topographical landscape\nThe Words: Key lines from the poem carved into the terrain like engraved stone\nThe Symbols: Miniature objects from the poem placed in the landscape (tiny boat, rose, clock, door)\nThe Poet: Microscopic figurine of the poet wandering through their own words\nThe Mood: Lighting matches emotional tone (fog for melancholy, harsh light for anger, soft gold for longing)\nThe Metadata: Name \"Poet | Collection | Year\" poet and poet era relevant items in the background.\nOutput: 2x2 Grid, Macro Photography, shallow depth of field","title":"Poetry Recommendation System Prompt (Carved Book Sculpture Aesthetic)","description":"A multi-phase prompt designed to act as a literary curator, analyzing a poet (e.g., T.S. Eliot) to find similar international poets and then generating a 2x2 grid of landscapes based on their work. The visual style is highly specific: a carved book paper sculpture aesthetic where the landscape emerges from the pages, with key lines of poetry carved into the terrain, and microscopic figurines of the poets included.","sourceMedia":["https://cms-assets.youmind.com/media/1768890693989_xmf1d4_G-6ry6UWwAAsSTN.jpg"],"needReferenceImages":false},{"content":"SCENE IMAGE + FIGURE & HUMAN IMAGE YOU WANT TO INTEGRATE + PROMPT.","title":"Turkish Prompt for Image Integration","description":"A Turkish instruction describing a workflow for integrating a figure or person into an existing scene. It requires pasting the scene screenshot (SS), the figure/person image, and the text prompt into Nano Banana Pro.","sourceMedia":["https://cms-assets.youmind.com/media/1768890737835_mzas2d_G_C6z_IXMAAbUgZ.jpg"],"needReferenceImages":true},{"content":"Specify a purple color scheme for the kimono and background color. Since the headquarters is in Japan, the style is Japanese-themed. The subject is holding a white fox (Byakko) mask, which is considered an auspicious messenger of the gods that brings happiness to people.","title":"Japanese Style White Fox Goddess for JoyPix Brand","description":"A Japanese language prompt used with Nano Banana Pro on JoyPix to create an image for the brand, featuring a white fox (Byakko) theme. The prompt specifies a purple color scheme (JoyPix's brand color), traditional Japanese attire (kimono), and the subject holding a white fox mask, symbolizing a lucky messenger of the gods.","sourceMedia":["https://cms-assets.youmind.com/media/1768890728446_o55qfw_G_B-yIiWgAEaxQy.jpg","https://cms-assets.youmind.com/media/1768890729421_u7ns4a_G_B-x7aW4AE54U0.jpg"],"needReferenceImages":false},{"content":"Cinderella, wearing a magical glow, princess, before the clock strikes 12, dress, glass slippers, ball, light, sparkling, fantasy, anime style, high definition, hyper-detailed, {argument name=\"color scheme\" default=\"blue and white\"}","title":"Cinderella Transformation AI Image Prompt","description":"This prompt is used with Nano Banana Pro to generate an image of a Cinderella-like princess, capturing the magical glow and transformation before the clock strikes midnight. The actual prompt text is found in the ALT text of the original tweet.","sourceMedia":["https://cms-assets.youmind.com/media/1768890723210_t83nlq_G_Byj6VWwAAfFSr.jpg"],"needReferenceImages":false},{"content":"A gigantic hollow astronaut suit overtaken by nature, standing in still water, covered in moss and vines. Inside the suit grows a thriving world — waterfalls, ancient temples, trees, birds in flight. Helmet reflects a peaceful sunrise over mountains. Surreal nature-meets-sci-fi fusion.\n\n8K ultra-detailed realism, cinematic natural lighting, soft mist, high texture realism, photorealistic fabric and stone, shallow depth of field, fantasy realism, vertical 9:16, no humans, no logos","title":"Surreal Nature-Meets-Sci-Fi Astronaut Suit","description":"A prompt for generating a surreal, ultra-detailed 8K image of a gigantic, hollow astronaut suit overtaken by nature (moss, vines, water). The suit contains a thriving miniature world inside, including temples and trees, with the helmet reflecting a peaceful sunrise.","sourceMedia":["https://cms-assets.youmind.com/media/1768890643289_h3mim9_G_AQY64a8AAmRVc.jpg"],"needReferenceImages":false},{"content":"A surreal dreamlike scene inside a large transparent glass lightbulb, zoom in a powerful rocket launching vertically upward, massive billowing clouds of thick white-orange rocket exhaust smoke filling the entire glass dome like a contained explosion, dramatic launch flames glowing bright orange-yellow at the base, the rocket piercing through dense turbulent smoke clouds,launch pad structures and support towers visible at the bottom inside the bulb, soft warm golden light emanating from the rocket flames illuminating the smoke from within, subtle reflections and refractions on the curved glass surface, intricate filament and screw base of the bulb at the bottom in metallic bronze, hyper-detailed,cinematic lighting, surreal concept art, magical realism, high contrast, dramatic atmosphere, ultra realistic textures, octane render, 8k, masterpiece with lighter background and its a dropdown shadow of blub on the floor to give realistic effect.","title":"Surreal Rocket Launch in a Lightbulb","description":"A prompt for generating a surreal concept art image of a powerful rocket launch contained entirely inside a large transparent glass lightbulb. It emphasizes dramatic lighting from the flames, hyper-detailed textures, and reflections on the glass surface.","sourceMedia":["https://cms-assets.youmind.com/media/1768890649633_14aoci_G-__hhQXIAATicR.jpg"],"needReferenceImages":false},{"content":"Prompt in Comments and Description !👇👇👇","title":"Sadie Sink and Caleb McLaughlin Couple Prompt (Implied)","description":"A post referencing a hot moments image of Sadie Sink and Caleb McLaughlin, indicating the prompt is available in the comments/description (which is not provided here).","sourceMedia":["https://cms-assets.youmind.com/media/1768804216065_oqaei1_G--5PxIWIAABan6.jpg"],"needReferenceImages":false}]
FILE:references/photography-examples.json
[
{
"title": "Convenience Store Neon Portrait",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @BubbleBrain",
"prompt": "A portrait of {subject} standing inside a convenience store at night, shot through the glass window, neon signs and fluorescent light reflections creating layered glows, candid documentary feel, Fujifilm color simulation, sharp focus on face, shallow depth of field",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["portrait", "neon", "nighttime", "documentary"]
},
{
"title": "Cinematic Minimal Portrait",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @iam_miharbi",
"prompt": "Ultra-clean cinematic portrait of {subject}, minimal composition, single directional key light from left, deep shadow on right side, neutral gray background, shot on 85mm lens, subtle film grain, muted color grade, editorial magazine quality",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["portrait", "cinematic", "minimal", "editorial"]
},
{
"title": "35mm Flash Editorial Portrait",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @BubbleBrain",
"prompt": "35mm film flash photography portrait of {subject}, direct on-camera flash, harsh shadows behind subject, high contrast, slight lens distortion, authentic snapshot aesthetic, Kodak Portra 400 film simulation, grainy texture",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["portrait", "35mm", "flash", "film"]
},
{
"title": "Soft Airy 35mm Portrait",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @BubbleBrain",
"prompt": "Soft natural light portrait of {subject}, airy and bright atmosphere, window backlight creating rim glow, diffused fill, skin tones warm and glowing, shallow depth of field with bokeh, Fujifilm Pro 400H simulation, dreamy and light",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["portrait", "soft", "airy", "natural light"]
},
{
"title": "Luxury Glam Beauty Portrait",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @patrickassale",
"prompt": "High-end beauty fashion portrait of {subject}, dramatic studio lighting with multiple softboxes, flawless skin retouching quality, bold makeup, hair perfectly styled, rich jewel-tone background, Vogue editorial aesthetic, ultra-sharp detail",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["beauty", "fashion", "editorial", "luxury"]
},
{
"title": "Japanese Onsen Ryokan Portrait",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @BubbleBrain",
"prompt": "Serene portrait of {subject} in a traditional Japanese ryokan, natural wood interior, indirect warm light from shoji screens, soft steam atmosphere from nearby onsen, calm and contemplative mood, Fujifilm simulation, warm tones with subtle green accent",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["portrait", "japanese", "ryokan", "atmospheric"]
},
{
"title": "Ultra-Realistic Product Photography",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @ZaraIrahh",
"prompt": "Ultra-realistic product photography of {subject}, displayed frontally on a soft sage-green surface, natural diffused window light from upper left, subtle shadows and reflections, clean white background, commercial photography quality, no post-processing artifacts",
"backend": "gpt-image-2",
"size": "1024x1024",
"tags": ["product", "photography", "commercial", "realistic"]
},
{
"title": "Urban Street Snapshot Portrait",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @Tz_2022",
"prompt": "Candid urban portrait of {subject} on a busy street, turned back slightly looking over shoulder, motion blur in background from passing traffic, golden hour warm light, documentary photography style, imperfect authentic composition",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["portrait", "street", "candid", "urban"]
}
]
FILE:references/poster-flyer-premium.json
[
{
"title": "VR Headset Exploded View Poster",
"source": "YouMind/awesome-gpt-image-2 @wory37303852 — Featured",
"prompt": "Exploded view product diagram poster of {subject}, clean high-tech 3D render style, studio lighting with glowing accents, soft purple and blue gradient background, vertically stacked exploded view showing distinct internal component layers, callout labels on left and right sides with technical descriptions, product name header with subtitle, footer with descriptive text block and logo, professional tech product launch poster layout",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["product", "exploded view", "tech", "3D", "poster"]
},
{
"title": "Vintage Travel Poster",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @WolfRiccardo",
"prompt": "Vintage travel poster for {subject}, retro 1950s-60s graphic design aesthetic, bold simplified illustration with flat color areas, Art Deco typography, warm Mediterranean palette of terracotta, cream, and cobalt blue, destination name in large serif font at top, atmospheric landscape illustration, subtle paper aging texture",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["travel", "vintage", "retro", "art deco", "poster"]
},
{
"title": "Futuristic Mandala Illustration Poster",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @4WEB1",
"prompt": "Futuristic mandala poster centered on {subject}, sacred geometry meets digital technology, intricate circular patterns with circuit board elements and cosmic motifs, deep space background with nebula colors — indigo, gold, electric blue, symmetrical ornate composition, spiritual-technological fusion aesthetic, dramatic scale contrast",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["mandala", "futuristic", "spiritual", "geometric", "poster"]
},
{
"title": "Super Famicom Retro Game Poster",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @lilimliliychan",
"prompt": "Retro Super Famicom / SNES era game box art poster for {subject}, bold primary colors, pixel art character illustrations, dramatic perspective composition, Japanese Kanji title in stylized font, action-packed scene with multiple characters, authentic 16-bit era aesthetic with slight scan lines, nostalgic gaming poster format",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["retro", "game", "famicom", "pixel", "poster"]
},
{
"title": "Dark Fantasy Epic Silhouette Poster",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @A9Quant",
"prompt": "Dark fantasy epic poster of {subject}, dramatic silhouette figure against vast luminous sky, extreme scale contrast between character and environment, cinematic color grading with deep navy and amber-gold, moody atmospheric fog layers, heroic composition with figure at 1/3 position, film poster quality, emotional gravity",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["dark fantasy", "epic", "silhouette", "cinematic", "poster"]
},
{
"title": "New Chinese Ink Landscape Poster",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @liyue_ai",
"prompt": "Contemporary Chinese ink landscape poster of {subject}, fusion of classical 水墨 ink wash technique with modern graphic design, bold simplified mountain forms in graduated ink tones, traditional red seal stamp accent, modern sans-serif Chinese typography, generous white negative space, cultural heritage meets contemporary aesthetics",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["chinese", "ink", "landscape", "traditional", "modern", "poster"]
},
{
"title": "Science Fiction Movie Poster",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @underwoodxie96",
"prompt": "Cinematic science fiction movie poster for {subject}, wide-angle establishing shot of futuristic environment, dramatic one-point perspective, mist and volumetric light rays, cool blue-green color palette with warm accent lighting, heroic figure silhouette at center, bold title treatment at bottom, blockbuster movie production quality",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["sci-fi", "movie", "cinematic", "futuristic", "poster"]
},
{
"title": "Chinese Minimalist S-Shaped Composition Poster",
"source": "EvoLinkAI/awesome-gpt-image-2-prompts @liyue_ai",
"prompt": "Elegant minimalist poster with S-shaped compositional flow for {subject}, inspired by classical Chinese aesthetic of 留白 negative space, elements arranged in graceful S-curve guiding eye through composition, limited palette of 2-3 tones, subtle texture suggesting silk or rice paper, refined typographic balance, meditative calm",
"backend": "gpt-image-2",
"size": "1024x1536",
"tags": ["minimalist", "chinese", "composition", "elegant", "poster"]
}
]
FILE:scripts/generate_image.py
#!/usr/bin/env python3
# /// script
# dependencies = [
# "google-genai>=1.0.0",
# "pillow>=10.0.0",
# ]
# ///
from __future__ import annotations
import argparse
import os
import re
import sys
from pathlib import Path
from google import genai
from google.genai import types
ALLOWED_ASPECT_RATIOS = ["1:1", "3:4", "4:3", "9:16", "16:9"]
def parse_args() -> argparse.Namespace:
parser = argparse.ArgumentParser(
description="Generate an image with Gemini and save it as PNG."
)
parser.add_argument(
"-p",
"--prompt",
required=True,
help="English prompt used for image generation.",
)
parser.add_argument(
"-f",
"--filename",
required=True,
help="Output filename or path for the generated PNG.",
)
parser.add_argument(
"-i",
"--input-image",
action="append",
dest="input_images",
metavar="IMAGE",
help="Input image path(s) for editing/composition. Can be specified multiple times (up to 14).",
)
parser.add_argument(
"-a",
"--aspect-ratio",
default="1:1",
choices=ALLOWED_ASPECT_RATIOS,
help="Aspect ratio for image generation.",
)
parser.add_argument(
"-m",
"--model",
default="gemini-3.1-flash-image-preview",
help="Gemini image model name.",
)
parser.add_argument(
"-k",
"--api-key",
default=None,
help="API key override. Fallback: GEMINI_API_KEY -> NANO_BANANA_API_KEY.",
)
return parser.parse_args()
def resolve_api_key(cli_api_key: str | None) -> str:
api_key = cli_api_key or os.getenv("GEMINI_API_KEY") or os.getenv("NANO_BANANA_API_KEY")
if not api_key:
raise ValueError(
"Missing API key. Provide --api-key or set GEMINI_API_KEY/NANO_BANANA_API_KEY."
)
return api_key
def ensure_png_path(filename: str) -> Path:
path = Path(filename).expanduser()
if path.suffix.lower() != ".png":
path = path.with_suffix(".png")
if path.parent == Path("."):
sanitized_name = re.sub(r"[^\w\-.\u4e00-\u9fff]", "-", path.name)
path = Path.cwd() / sanitized_name
path.parent.mkdir(parents=True, exist_ok=True)
return path.resolve()
def extract_and_save_image(response, output_path: Path) -> None:
parts = getattr(response, "parts", None)
if parts is None and getattr(response, "candidates", None):
try:
parts = response.candidates[0].content.parts
except Exception:
parts = None
if not parts:
raise RuntimeError("Gemini returned no content parts.")
for part in parts:
if getattr(part, "inline_data", None):
image = part.as_image()
try:
image.save(output_path, format="PNG")
except TypeError:
# google-genai Image.save() may not accept format kwarg
image.save(str(output_path))
return
raise RuntimeError("Gemini response did not include image data.")
def load_input_images(paths: list[str] | None) -> list:
"""Load input images as PIL Image objects for editing/composition."""
if not paths:
return []
if len(paths) > 14:
raise ValueError(f"Too many input images ({len(paths)}). Maximum is 14.")
from PIL import Image as PILImage
images = []
for img_path in paths:
try:
with PILImage.open(img_path) as img:
images.append(img.copy())
print(f"Loaded input image: {img_path}")
except Exception as e:
raise ValueError(f"Failed to load input image '{img_path}': {e}") from e
return images
def main() -> int:
args = parse_args()
try:
api_key = resolve_api_key(args.api_key)
output_path = ensure_png_path(args.filename)
# Load reference images if provided
input_images = load_input_images(args.input_images)
# Build contents: images first (if any), then the text prompt
if input_images:
contents = [*input_images, args.prompt]
print(f"Editing/composing {len(input_images)} image(s) with prompt...")
else:
contents = args.prompt
print(f"Generating image from prompt...")
client = genai.Client(api_key=api_key)
response = client.models.generate_content(
model=args.model,
contents=contents,
config=types.GenerateContentConfig(
response_modalities=["IMAGE"],
image_config=types.ImageConfig(aspect_ratio=args.aspect_ratio),
),
)
extract_and_save_image(response, output_path)
print(f"MEDIA: {output_path}")
return 0
except KeyboardInterrupt:
print("Error: Generation interrupted by user.", file=sys.stderr)
return 130
except Exception as exc:
print(f"Error: {exc}", file=sys.stderr)
return 1
if __name__ == "__main__":
raise SystemExit(main())
FILE:scripts/gpt_image2.py
#!/usr/bin/env python3
# /// script
# dependencies = ["requests>=2.28.0"]
# ///
"""
GPT Image 2 CLI wrapper for image-forge skill.
Supports: generate (text-to-image) and edit (image-to-image).
Usage:
# Generate
python gpt_image2.py generate --prompt "..." --output /path/out.png [--size 1536x1024] [--quality high]
# Edit (single reference image)
python gpt_image2.py edit --prompt "..." --image /path/ref.png --output /path/out.png
# Edit (multiple reference images, up to 4)
python gpt_image2.py edit --prompt "..." --image ref1.png --image ref2.png --output /path/out.png
Environment:
CRS_BASE_URL CRS service base URL (default: http://127.0.0.1:8765)
CRS_API_KEY CRS API key (required)
"""
from __future__ import annotations
import argparse
import base64
import os
import sys
import time
from pathlib import Path
try:
import requests
except ImportError:
print("Missing dependency: pip install requests", file=sys.stderr)
sys.exit(1)
CRS_BASE = os.environ.get("CRS_BASE_URL", "http://127.0.0.1:8765")
CRS_KEY = os.environ.get("CRS_API_KEY", "")
VALID_SIZES = [
"1024x1024", "1536x1024", "1024x1536",
"2048x2048", "3840x2160", "2160x3840",
]
DEFAULT_SIZE_GENERATE = "1536x1024"
DEFAULT_SIZE_EDIT = "1024x1536"
def get_headers() -> dict:
if not CRS_KEY:
print("Error: CRS_API_KEY not set", file=sys.stderr)
sys.exit(1)
return {"Authorization": f"Bearer {CRS_KEY}"}
def read_image_b64(path: str) -> str:
with open(path, "rb") as f:
return base64.b64encode(f.read()).decode()
def detect_mime(path: str) -> str:
ext = Path(path).suffix.lower()
return {"jpg": "image/jpeg", ".jpeg": "image/jpeg", ".webp": "image/webp"}.get(ext, "image/png")
def save_result(data: dict, output: str, fmt: str = "png") -> str:
out_path = output or f"/tmp/gpt-image2-{int(time.time())}.{fmt}"
b64 = data.get("b64_json", "")
if not b64:
print("Error: no b64_json in response", file=sys.stderr)
sys.exit(1)
with open(out_path, "wb") as f:
f.write(base64.b64decode(b64))
return out_path
def cmd_generate(args: argparse.Namespace) -> None:
payload = {
"model": "gpt-image-2",
"prompt": args.prompt,
"size": args.size or DEFAULT_SIZE_GENERATE,
"quality": args.quality,
"output_format": args.format,
"response_format": "b64_json",
}
if args.background:
payload["background"] = args.background
resp = requests.post(
f"{CRS_BASE}/openai/v1/images/generations",
headers=get_headers(),
json=payload,
timeout=args.timeout,
)
_handle_response(resp, args)
def cmd_edit(args: argparse.Namespace) -> None:
if not args.image:
print("Error: --image required for edit", file=sys.stderr)
sys.exit(1)
images = []
for img_path in args.image:
mime = detect_mime(img_path)
b64 = read_image_b64(img_path)
images.append({"image_url": f"data:{mime};base64,{b64}"})
payload = {
"model": "gpt-image-2",
"prompt": args.prompt,
"images": images,
"size": args.size or DEFAULT_SIZE_EDIT,
"quality": args.quality,
"output_format": args.format,
"response_format": "b64_json",
}
resp = requests.post(
f"{CRS_BASE}/openai/v1/images/edits",
headers=get_headers(),
json=payload,
timeout=args.timeout,
)
_handle_response(resp, args)
def _handle_response(resp: requests.Response, args: argparse.Namespace) -> None:
try:
d = resp.json()
except Exception:
print(f"Error: non-JSON response (HTTP {resp.status_code})", file=sys.stderr)
print(resp.text[:500], file=sys.stderr)
sys.exit(1)
if "error" in d:
print(f"API Error: {d['error'].get('message', d['error'])}", file=sys.stderr)
sys.exit(1)
if "data" not in d or not d["data"]:
print(f"Error: unexpected response: {d}", file=sys.stderr)
sys.exit(1)
item = d["data"][0]
out_path = save_result(item, args.output, args.format)
print(f"MEDIA: {os.path.abspath(out_path)}")
if item.get("revised_prompt"):
print(f"# revised_prompt: {item['revised_prompt'][:200]}", file=sys.stderr)
def main() -> None:
parser = argparse.ArgumentParser(description="GPT Image 2 CLI for image-forge")
sub = parser.add_subparsers(dest="command", required=True)
shared = argparse.ArgumentParser(add_help=False)
shared.add_argument("-p", "--prompt", required=True)
shared.add_argument("-o", "--output", default="")
shared.add_argument("--size", choices=VALID_SIZES, default="")
shared.add_argument("--quality", choices=["standard", "high"], default="high")
shared.add_argument("--format", choices=["png", "webp", "jpeg"], default="png", dest="format")
shared.add_argument("--timeout", type=int, default=180)
# generate
gen = sub.add_parser("generate", parents=[shared])
gen.add_argument("--background", choices=["transparent", "white", "auto"], default="")
# edit
edit = sub.add_parser("edit", parents=[shared])
edit.add_argument("-i", "--image", action="append", metavar="PATH",
help="Reference image path (repeat for multiple, max 4)")
args = parser.parse_args()
if args.command == "generate":
cmd_generate(args)
elif args.command == "edit":
cmd_edit(args)
if __name__ == "__main__":
main()
FILE:scripts/reverse_style.py
#!/usr/bin/env python3
# /// script
# dependencies = [
# "google-genai>=1.0.0",
# ]
# ///
"""
Reverse-engineer a visual style from a reference image using Gemini Vision.
Outputs a structured Chinese prompt prefix suitable for image generation.
Usage:
uv run reverse_style.py --image /path/to/ref.jpg
uv run reverse_style.py --image /path/to/ref.jpg --output style.txt
"""
from __future__ import annotations
import argparse
import os
import sys
from pathlib import Path
from google import genai
from google.genai import types
ANALYSIS_PROMPT = """请作为一名顶级的 AI 绘画提示词专家，为我分析这张图片的视觉风格。
**任务目标：** 提取并反推这张图片的艺术风格，生成一份通用的 Prompt。这份 Prompt 必须剥离原图中的具体角色、文字或特定情节，仅保留其美学灵魂。
**分析维度（请务必涵盖以下 15 个方面）：**
1. **基础维度：** 画面风格、画面成分组成、构图方式、分镜类型、光影特质、色调与色彩科学、媒介与材质纹理、情绪与氛围、渲染/拍摄参数。
2. **进阶维度：** 时代感与文化语境、空间逻辑与透视关系、信息密度与留白、动态状态（瞬时感）、后期处理与数字痕迹、符号化特征。
**输出要求：**
1. 请直接输出一段完整的、高水准的**中文提示词**。
2. 在提示词的开头或核心位置，使用 `[在此处替换为您想要生成的主体内容]` 作为占位符。
3. 确保该 Prompt 具有高度通用性，用户只需更换占位符内容，即可在保持原图质感的同时生成全新的画面。
4. 无需输出分析过程，请直接给出最终的 Prompt 文本。"""
def resolve_api_key(cli_key: str | None) -> str:
key = cli_key or os.getenv("GEMINI_API_KEY") or os.getenv("NANO_BANANA_API_KEY")
if not key:
raise ValueError("Missing API key. Set GEMINI_API_KEY or NANO_BANANA_API_KEY.")
return key
def load_image_bytes(path: str) -> tuple[bytes, str]:
p = Path(path).expanduser().resolve()
if not p.exists():
raise FileNotFoundError(f"Image not found: {p}")
suffix = p.suffix.lower()
mime_map = {".jpg": "image/jpeg", ".jpeg": "image/jpeg", ".png": "image/png",
".webp": "image/webp", ".gif": "image/gif"}
mime = mime_map.get(suffix, "image/jpeg")
return p.read_bytes(), mime
def reverse_style(image_path: str, api_key: str, model: str = "gemini-2.5-flash") -> str:
client = genai.Client(api_key=api_key)
img_bytes, mime_type = load_image_bytes(image_path)
response = client.models.generate_content(
model=model,
contents=[
types.Part.from_bytes(data=img_bytes, mime_type=mime_type),
types.Part.from_text(text=ANALYSIS_PROMPT),
],
)
return response.text.strip()
def main() -> None:
parser = argparse.ArgumentParser(description="Reverse-engineer image style via Gemini Vision.")
parser.add_argument("-i", "--image", required=True, help="Path to reference image")
parser.add_argument("-o", "--output", default=None, help="Save result to file (optional)")
parser.add_argument("-m", "--model", default="gemini-2.5-flash", help="Gemini text model")
parser.add_argument("-k", "--api-key", default=None, help="API key override")
args = parser.parse_args()
api_key = resolve_api_key(args.api_key)
print(f"🔍 Analyzing style from: {args.image}", file=sys.stderr)
result = reverse_style(args.image, api_key, args.model)
if args.output:
Path(args.output).expanduser().write_text(result, encoding="utf-8")
print(f"✅ Style saved to: {args.output}", file=sys.stderr)
print(result)
if __name__ == "__main__":
main()
FILE:styles/constructivism.yaml
id: constructivism
name: 俄国构成主义
source: "[email protected]/2044964"
placeholder: "[在此处替换为您想要生成的主体内容]"
prompt: |
[在此处替换为您想要生成的主体内容]，俄国构成主义风格，平面设计插画，极简主义矢量艺术，复古宣传海报。画面由强烈的几何形状构成，包含大量的锐利三角形、圆形和粗重的对角线切割。色调采用极简的三色限定：高饱和度宝蓝色、深黑色和米白色（做旧纸张感）。整体具有复古丝网印刷质感，布满细腻的颗粒噪点和磨损纹理。构图充满张力，强调不对称的平衡感和工业力量感，锐利的线条边缘，扁平化视觉，高对比度。
FILE:styles/engraving-halftone.yaml
id: engraving-halftone
name: 半调雕刻线稿
source: "[email protected]/2044964"
placeholder: "[主体人物/对象]"
prompt: |
一幅极简主义平面设计海报，采用"半调雕刻线稿"风格（Engraving Halftone Style）。画面由密集的同心圆线条构成，通过线条的粗细变化和疏密程度，巧妙地勾勒出[主体人物/对象]的轮廓与面部阴影，形成强烈的立体感。视觉表现上采用极简双色调方案，背景色为深蓝色，线条颜色为明黄色。整体构图简洁有力，具有矢量艺术的质感，风格前卫且具有现代主义海报设计感。
FILE:styles/glitch-window-v1.yaml
id: glitch-window-v1
name: 错位矩形故障艺术 v1
source: "[email protected]/2044964"
placeholder: "[在此处替换为您想要生成的动漫角色]"
prompt: |
二次元平面艺术插画，[在此处替换为您想要生成的动漫角色]。故障艺术风格，赛博朋克动漫美学，数字碎片化构图。画面由多个错位的矩形窗口和几何切片叠加而成，呈现出一种数据损坏和图像溢出的视觉感。核心风格包含：像素排序（Pixel Sorting）效果、RGB色彩偏移、横向拉伸的数字噪点以及彩虹色调的电流纹理。背景采用极简主义的米白色，与画面中心高饱和度的湛蓝天空、厚重的积雨云形成强烈视觉对比。整体氛围带有超现实的忧郁感和深邃的数字空间感，构图错落有致，充满现代平面设计感。
FILE:styles/glitch-window-v2.yaml
id: glitch-window-v2
name: 错位矩形窗口重叠 v2
source: "[email protected]/2044964"
placeholder: "[在此处替换为您想要生成的动漫角色]"
prompt: |
二次元平面艺术插画，[在此处替换为您想要生成的动漫角色]，人物需要尽量使用全身像，且不使用常规的正面全身像而是做出展现人物动态的速写动作。画面采用"窗口重叠 (Window Overlay)"与"数字拼贴"的构图。角色的轮廓由多个错位的矩形框构成，某些方框区域被处理成透明视窗，展示出清朗的蓝天与积雨云纹理，仿佛角色体内蕴含着广阔的天空。画面中装饰有精美的故障艺术 (Glitch Art) 元素，如极简的黑色几何长条、细密的彩色电子扫描线以及错位的色彩偏移纹理。整体视觉呈现出一种现代平面设计的律动感，色彩以克莱因蓝和纯净白为主，背景简洁明快，氛围宁静且富有诗意。
FILE:styles/high-contrast-industrial.yaml
id: high-contrast-industrial
name: 高对比度数字工业故障
source: "[email protected]/2044964"
placeholder: "[在此处替换为您想要生成的主体内容]"
prompt: |
极简高对比图形艺术风格，[在此处替换为您想要生成的主体内容]呈现出深邃的黑色剪影与鲜明电光蓝（Electric Blue）交织的重影质感。画面采用极端的仰视低角度构图（Low Angle Shot），展现强烈的动态对角线张力与线条穿插的复杂结构。背景为大面积的纯白高调留白，形成极高的视觉反差。色彩方案严格限定于：纯黑、克莱因蓝/电光蓝、以及高亮白。画面带有浓郁的胶片噪点、Riso印刷纹理、以及明显的色差边缘（Chromatic Aberration）与数字故障痕迹。光影呈现高阈值的二值化硬核特质，边缘锐化且伴有像素撕裂感。整体视觉语言融合了后现代工业美学与都市孤寂感，信息密度极高且富有冷峻的平面设计感。
FILE:styles/index.yaml
# Image Forge Style Library Index
# 两层结构：
# signature_styles — 高度具体的视觉方案（有完整 YAML prompt recipe）
# rendering_styles — 通用渲染技法类别（inline modifier，直接注入 prompt）
#
# 路由规则：
# 用户提到风格名/关键词 → 先匹配 signature_styles.aliases，再匹配 rendering_styles.aliases
# Signature 命中 → 加载对应 YAML，走 Path S，默认后端按 preferred_backend
# Rendering 命中 → 取 modifier 字段注入 prompt，走 Path R-lite，按 preferred_backend 调度
# ──────────────────────────────────────────
# Tier 1: Signature Styles（有独立 YAML 文件）
# ──────────────────────────────────────────
signature_styles:
- id: constructivism
file: constructivism.yaml
category: print-art
aliases: [俄国构成主义, 苏联构成主义, 构成主义, 几何宣传, constructivism, soviet poster, bauhaus-soviet]
aspect_ratio: "3:4"
preferred_backend: nano-banana-2
tags: [geometric, vintage, poster, high-contrast]
use_case_affinity: [poster-flyer, social-media-post]
avoid_for: [ecommerce-main-image, product-marketing]
- id: glitch-window-v1
file: glitch-window-v1.yaml
category: digital-art
aliases: [错位矩形, 故障艺术, 数字碎片, glitch, pixel sorting, glitch art]
aspect_ratio: "3:4"
preferred_backend: nano-banana-2
tags: [glitch, anime, digital, cyber]
use_case_affinity: [profile-avatar, social-media-post, poster-flyer]
avoid_for: [ecommerce-main-image, product-marketing]
- id: glitch-window-v2
file: glitch-window-v2.yaml
category: digital-art
aliases: [错位矩形v2, 窗口重叠, 数字拼贴, window overlay, glitch v2]
aspect_ratio: "3:4"
preferred_backend: nano-banana-2
tags: [glitch, anime, sky, dreamy]
use_case_affinity: [profile-avatar, social-media-post]
avoid_for: [ecommerce-main-image]
- id: mixed-media
file: mixed-media.yaml
category: illustration
aliases: [混合媒介, 线稿摄影, 素描背景, mixed media, sketch photo]
aspect_ratio: "1:1"
preferred_backend: nano-banana-2
tags: [mixed-media, sketch, photography, lo-fi]
use_case_affinity: [profile-avatar, social-media-post, poster-flyer]
- id: tri-color
file: tri-color.yaml
category: minimal
aliases: [黑蓝红, 三色限定, 极简剪影, tri-color, three color, silhouette]
aspect_ratio: "16:9"
preferred_backend: nano-banana-2
tags: [minimal, silhouette, landscape, cinematic]
use_case_affinity: [poster-flyer, youtube-thumbnail, social-media-post]
- id: engraving-halftone
file: engraving-halftone.yaml
category: print-art
aliases: [半调雕刻, 铜版画, 雕刻线稿, engraving, halftone, etching]
aspect_ratio: "3:4"
preferred_backend: nano-banana-2
tags: [engraving, halftone, minimal, modern-poster]
use_case_affinity: [poster-flyer, profile-avatar]
- id: risograph-magazine
file: risograph-magazine.yaml
category: print-art
aliases: [半调杂志, risograph, 印刷风, riso, retro print, magazine]
aspect_ratio: "3:4"
preferred_backend: nano-banana-2
tags: [risograph, vintage, pop, magazine]
use_case_affinity: [poster-flyer, social-media-post, profile-avatar]
avoid_for: [ecommerce-main-image, product-marketing]
- id: pop-ink-splash
file: pop-ink-splash.yaml
category: illustration
aliases: [波普水墨, 波普喷溅, 克莱因波普, pop art, ink splash, pop ink]
aspect_ratio: "9:16"
preferred_backend: nano-banana-2
tags: [pop, ink, dynamic, urban]
use_case_affinity: [profile-avatar, social-media-post, poster-flyer]
- id: klein-blue-order
file: klein-blue-order.yaml
category: minimal
aliases: [克莱因秩序, 克莱因蓝, 极简仰拍, klein blue, klein order, summer cool]
aspect_ratio: "1:1"
preferred_backend: nano-banana-2
tags: [minimal, anime, clean, summer]
use_case_affinity: [profile-avatar, social-media-post]
- id: high-contrast-industrial
file: high-contrast-industrial.yaml
category: digital-art
aliases: [高对比度工业, 数字工业, 故障工业, electric blue, industrial glitch]
aspect_ratio: "1:1"
preferred_backend: nano-banana-2
tags: [industrial, high-contrast, glitch, film-noir]
use_case_affinity: [poster-flyer, youtube-thumbnail, product-marketing]
# ──────────────────────────────────────────
# Tier 2: Rendering Styles（inline modifier）
# Prompt modifier 来源：YouMind/awesome-gpt-image-2、EvoLinkAI/awesome-gpt-image-2-prompts 实战案例精炼
# ──────────────────────────────────────────
rendering_styles:
- id: photography
category: photo
aliases: [摄影, 写真, 真实照片, photography, photo-realistic, realistic photo, 胶片写真]
modifier: "ultra-realistic DSLR photography, Fujifilm film simulation, natural bokeh with shallow depth of field, sharp subject with authentic film texture, professional studio or natural window lighting, candid documentary quality"
preferred_backend: gpt-image-2
examples_file: "references/photography-examples.json"
tags: [realistic, photo, natural]
- id: cinematic-film-still
category: photo
aliases: [电影感, 胶片, cinematic, film still, movie screenshot, 电影截图, 大片质感]
modifier: "cinematic film still, anamorphic lens compression, dramatic chiaroscuro lighting, color-graded footage, shallow depth of field, subtle film grain, widescreen composition, blockbuster movie production quality"
preferred_backend: gpt-image-2
tags: [cinematic, dramatic, film]
- id: anime-manga
category: illustration
aliases: [动漫, 二次元, 漫画风, anime, manga, 动漫风格, 日式动画]
modifier: "anime illustration style, vibrant saturated colors, clean precise linework, expressive character design, Japanese animation aesthetic, detailed cel-shading, dynamic composition, professional anime production quality"
preferred_backend: nano-banana-2
tags: [anime, illustration, japanese]
- id: illustration
category: illustration
aliases: [插画, 手绘插画, 商业插画, illustration, digital illustration, 插图, 小清新插画]
modifier: "professional digital illustration, editorial quality artwork, bold graphic shapes with intentional composition, commercial illustration standard, vector-friendly clean linework, balanced color harmony"
preferred_backend: nano-banana-2
tags: [illustration, digital, editorial]
- id: sketch-line-art
category: illustration
aliases: [素描, 线稿, 速写, sketch, line art, pencil drawing, 铅笔画, 手绘线稿]
modifier: "clean precise line art sketch, technical pen-and-ink illustration quality, confident ink outlines with varying stroke weight, cross-hatching for shadow depth, white paper background, architectural or editorial drawing standard"
preferred_backend: nano-banana-2
tags: [sketch, lineart, hand-drawn]
- id: 3d-render
category: 3d
aliases: [3D渲染, 三维渲染, 3D, 3D render, CGI, octane render, 产品渲染]
modifier: "photorealistic 3D render, ray-traced global illumination, physically-based materials, subsurface scattering on skin, HDRI studio lighting, high-poly detail, product visualization quality, Octane or Cinema4D aesthetic"
preferred_backend: gpt-image-2
tags: [3d, render, cgi]
- id: chibi-q-style
category: illustration
aliases: [Q版, chibi, 可爱风, 萌系, cute chibi, q-style, 萨娜风]
modifier: "chibi Q-style character illustration, 2:1 oversized head to tiny body ratio, large expressive sparkly eyes, rounded soft shapes, vibrant kawaii color palette, playful energetic pose, professional anime chibi production quality"
preferred_backend: nano-banana-2
tags: [chibi, cute, kawaii]
- id: isometric
category: 3d
aliases: [等距视角, 等轴测, isometric, 2.5D, 等距投影]
modifier: "clean isometric illustration, precise 30-degree axonometric projection, flat-shaded geometric forms with consistent upper-left light source, muted pastel or corporate palette, architectural diagram precision, clean vector quality"
preferred_backend: gpt-image-2
tags: [isometric, geometric, clean]
- id: pixel-art
category: digital-art
aliases: [像素艺术, 像素风, pixel art, retro game, 8-bit, 16-bit, 象素风格]
modifier: "authentic pixel art, 32px to 256px sprite scale, limited 16-32 color palette, sharp aliased edges with no anti-aliasing, retro 16-bit game aesthetic, clean sprite-quality detail, SNES or Mega Drive era visual language"
preferred_backend: nano-banana-2
tags: [pixel, retro, game]
- id: oil-painting
category: fine-art
aliases: [油画, 古典油画, oil painting, classical painting, 干笔油画]
modifier: "classical oil painting technique, visible impasto brushstrokes with palette knife texture, rich deep color saturation with glazed translucent layers, chiaroscuro light modeling, Baroque or Dutch Golden Age quality, textured canvas weave visible"
preferred_backend: nano-banana-2
tags: [oil-painting, classical, fine-art]
- id: watercolor
category: fine-art
aliases: [水彩, 水彩画, watercolor, aquarelle, 水彩漫画]
modifier: "loose expressive watercolor painting, wet-on-wet color bleeding with granulation, white paper showing through as negative space, organic soft edges with diffusion, lyrical editorial wash quality, Winsor & Newton pigment richness"
preferred_backend: nano-banana-2
tags: [watercolor, soft, transparent]
- id: ink-chinese-style
category: fine-art
aliases: [水墨, 中国画, 国画, 水墨风, ink wash, Chinese painting, chinese ink, 山水画, 墨笔画]
modifier: "Chinese ink wash painting (shuimo hua), xieyi freehand brushwork with confident calligraphic strokes, black ink gradation from rich dense to dilute translucent wash, rice paper texture, generous negative space (liu bai) as compositional element, classical scholar-painter aesthetic"
preferred_backend: nano-banana-2
tags: [ink, chinese, minimalist]
- id: retro-vintage
category: print-art
aliases: [复古, 复古风, retro, vintage, 老照片, 年代感, 胶片复古]
modifier: "authentic vintage aesthetic, expired film grain and halation glow, warm amber-sepia color shift with faded muted tones, soft vignette with light leaks, analog photography feel, 1950s-70s era visual language, aged paper or print texture"
preferred_backend: gpt-image-2
tags: [retro, vintage, nostalgic]
- id: cyberpunk-sci-fi
category: digital-art
aliases: [赛博朋克, 科幻, 未来感, cyberpunk, sci-fi, neon dystopia, 霓虹, 赛博]
modifier: "cyberpunk dystopian aesthetic, high-saturation neon signs reflected in rain-slicked streets, holographic interference patterns, volumetric fog and atmospheric haze, vibrant magenta-cyan-amber color language, Blade Runner or Ghost in the Shell atmosphere"
preferred_backend: gpt-image-2
tags: [cyberpunk, neon, futuristic]
- id: minimalism
category: minimal
aliases: [极简, 极简主义, 简约, minimalist, minimalism, clean design, 日式极简]
modifier: "radical minimalism with maximum negative space, single hero element with precise geometric relationships, monochromatic or strictly limited 2-color palette, breathing room as deliberate design choice, Swiss International Style discipline, Muji or Apple-level restraint"
preferred_backend: gpt-image-2
tags: [minimal, clean, design]
# ──────────────────────────────────────────
# Tier 3: Logo Showcase Backgrounds（来源 logo-generator skill）
# 专为品牌 logo 展示场景，配合 brand-logo-showcase 用途使用
# 触发：用户说"展示图"/"showcase"/"背景风格" + 下列名称
# ──────────────────────────────────────────
logo_showcase_backgrounds:
- id: logo-void
category: logo-showcase
aliases: [虚空, the void, 极简黑, 绝对黑]
modifier: "pure black background, extremely fine silver-white electronic film grain micro-noise, cold icy blue corner glow at extreme edge, generous negative space, professional brand identity presentation, white or silver logo color"
preferred_backend: nano-banana-2
suitable_for: [hardcore tech, data security, infrastructure]
tags: [dark, logo, minimal, tech]
- id: logo-frosted
category: logo-showcase
aliases: [磨砂穹顶, frosted horizon, 钛色背景, Apple风格展示]
modifier: "deep titanium gray background, organic film-like dust texture, large area cold gray-blue light halo dissolved at edges like mist, Apple-presentation breathing quality, white logo"
preferred_backend: nano-banana-2
suitable_for: [premium products, design brands]
tags: [dark, premium, metal]
- id: logo-fluid-abyss
category: logo-showcase
aliases: [流体深渊, fluid abyss, AI风格展示, 深紫背景]
modifier: "deep midnight purple background, fluid fusion of dark orange and dark blue slowly interweaving, nebula-quality texture, AI-native computational atmosphere, white logo centered"
preferred_backend: nano-banana-2
suitable_for: [AI products, data systems]
tags: [dark, ai, fluid]
- id: logo-spotlight
category: logo-showcase
aliases: [影棚, studio spotlight, 杂志风展示, 编辑风]
modifier: "extremely dark warm carbon gray background, larger grain simulating low-light photography, single-side softbox creating natural vignette, editorial magazine quality, white logo centered"
preferred_backend: nano-banana-2
suitable_for: [editorial brands, professional services]
tags: [dark, editorial, studio]
- id: logo-analog-liquid
category: logo-showcase
aliases: [物理流体, analog liquid, 金属麦叶, 创意品牌展示]
modifier: "Klein blue solid color base, metallic gold dust flow and iridescent pigment shimmer overlay, chaotic organic metallic texture contrasting with clean vector logo, artistic brand identity"
preferred_backend: nano-banana-2
suitable_for: [creative tools, artistic brands]
tags: [dark, creative, metallic]
- id: logo-led-matrix
category: logo-showcase
aliases: [数字硬件, LED matrix, 赛博朋克展示, 点阵矩阵]
modifier: "pure black background with glowing dot matrix patterns, CRT display artifacts and halftone dots, retro LED billboard aesthetic, cyberpunk retro-futurism, white logo as solid entity in front"
preferred_backend: nano-banana-2
suitable_for: [Web3, AI computing, hardware]
tags: [dark, cyberpunk, retro]
- id: logo-editorial-paper
category: logo-showcase
aliases: [纸本编辑, editorial paper, 小众美学展示, 高级白背景]
modifier: "off-white alabaster paper background, watercolor rough art paper texture, natural diffused light, subtle warm gray corner vignette, humanistic independent magazine quality, dark logo"
preferred_backend: nano-banana-2
suitable_for: [humanistic brands, fashion, professional services]
tags: [light, paper, editorial]
- id: logo-iridescent-frost
category: logo-showcase
aliases: [幻彩透砂, iridescent frost, 彩虹砂面, 光学材质]
modifier: "extremely light silver-gray background, frosted glass or sandblasted aluminum surface, soft holographic iridescent colors — light purple, light blue, soft pink — through thick frosted glass, Apple hardware render quality, dark logo"
preferred_backend: nano-banana-2
suitable_for: [tech hardware, scientific applications]
tags: [light, iridescent, optical]
- id: logo-morning-aura
category: logo-showcase
aliases: [晨雾光域, morning aura, 温柔AI展示, 温暖背景]
modifier: "warm ivory cream background, soft morning mist noise, blurred low-saturation pastels — mint green, baby blue, dawn orange — dissolving into warm white, approachable intelligent atmosphere, dark logo"
preferred_backend: nano-banana-2
suitable_for: [accessible AI, health tech, consumer apps]
tags: [light, soft, warm]
- id: logo-clinical
category: logo-showcase
aliases: [无菌影棚, clinical studio, 白色影棚, 算法品牌]
modifier: "pure white or extremely light cold gray background, high-frequency sharp digital micro-noise, large softbox from above creating smooth gradient shadow, sterile spatial order, algorithm-driven confidence, dark logo"
preferred_backend: nano-banana-2
suitable_for: [SaaS, data-centric brands]
tags: [light, clinical, minimal]
- id: logo-ui-container
category: logo-showcase
aliases: [容器化界面, ui container, App展示风格, 磨砂玻璃展示]
modifier: "frosted glass container with rounded corners and subtle transparency, micro drop-shadow depth, clean gradient background behind container, UI-native digital product presentation, SaaS platform aesthetic, logo inside container"
preferred_backend: nano-banana-2
suitable_for: [digital products, apps, SaaS]
tags: [light, digital, ui]
- id: logo-swiss-flat
category: logo-showcase
aliases: [瑞士扁平, swiss flat, 纯色展示, 经典权威展示]
modifier: "100% pure solid deep vintage green background, absolutely flat with zero gradients, zero noise, zero effects, pure graphic design with only color and form, Swiss International Style authority, white logo, maximum negative space"
preferred_backend: nano-banana-2
suitable_for: [established brands, environmental products, classic institutions]
tags: [solid, swiss, flat]
FILE:styles/klein-blue-order.yaml
id: klein-blue-order
name: 克莱因秩序
source: "[email protected]/2044964"
placeholder: "[在此处替换为您想要生成的动漫角色]"
prompt: |
现代极简主义二次元插画，[在此处替换为您想要生成的动漫角色]，赛璐璐风格（Cel-shading）。画面采用极简的几何切割构图，角色置于大面积的负空间留白之中。色彩上采用极具冲击力的克莱因蓝（Klein Blue）与高亮纯白构成双色视觉核心。光影特质为硬边阴影（Hard edges shadow），模拟正午强烈的直射日光，角色受光面清透微曝，阴影区深邃且边缘锐利，呈现出极高的明暗对比度。空间逻辑采用强烈的仰拍透视，强调线条的延伸感。整体氛围具有一种夏日清冷、孤独且超现实的现代美感。线条利落，色彩平整，无杂色颗粒，通透感，大师级动画分镜感。
FILE:styles/mixed-media.yaml
id: mixed-media
name: 混合媒介（线稿+摄影）
source: "[email protected]/2044964"
placeholder: "[在此处替换为您想要生成的主体内容]"
prompt: |
一件混合媒介艺术作品。前景色：一个极简主义的白色线稿素描，描绘[在此处替换为您想要生成的主体内容]，近景胸像，线条细腻纤细，半透明的剪影感，全身仅眼睛部分带有发光的淡紫色。背景：一张写实的、大光圈虚化的摄影照片，场景为黄昏时分的海岸电车道口与背后的海岸，电影感光影，天空呈现暮蓝色与金黄色的渐变。采用倾斜构图（荷兰角视角），水平线明显倾斜。风格：空灵的氛围，怀旧的Lo-fi美学，锐利的白色线条与柔软模糊的实景摄影形成强烈对比，梦幻且忧郁的意境。
FILE:styles/pop-ink-splash.yaml
id: pop-ink-splash
name: 波普+水墨喷溅
source: "[email protected]/2044964"
placeholder: "[在此处替换为您想要生成的主体内容]"
prompt: |
现代日系混合媒介插画风格，[在此处替换为您想要生成的主体内容]。采用倒置动态构图，结合扁平化波普艺术逻辑。色彩以高饱和度明黄为主基调，运用克莱因蓝与大红进行强视觉对冲。画面融合赛璐璐平涂、波点网纹（Halftone）及水墨喷溅质感，具有纸张肌理与数码后期叠加的综合材质感。光影利落，空间呈现多层次平面拼贴关系。整体氛围洋溢着现代都市的轻盈感与瞬时爆发力，充满时尚平面设计的高信息密度与符号化视觉冲击。
FILE:styles/risograph-magazine.yaml
id: risograph-magazine
name: 半调杂志 Risograph
source: "[email protected]/2044964"
placeholder: "[在此处填写您的主体，例如：一只复古留声机 / 一把电吉他 / 一杯咖啡]"
prompt: |
现代复古平面海报设计，Risograph半调网点印刷风格。画面正中心是[在此处填写您的主体]。主体采用深蓝与米白交织的半调网点纹理表现。背景为带有粗糙颗粒感的米色纸张。主体背后衬托着一个明黄色的几何实心拱门色块。主体周围环绕着极细的抽象交错轨道线条和几个微小的品红色四芒星符号。画面边缘（顶部和底部）带有深蓝色的复古粗体无衬线排版文字，部分文字带有明黄色高光色块底色。右上角包含一个条形码图形元素。整体构图极简，色彩对比强烈，具有波普艺术和复古杂志封面的视觉冲击力。
FILE:styles/tri-color.yaml
id: tri-color
name: 黑蓝红三色极简剪影
source: "[email protected]/2044964"
placeholder: "[在此处替换为主体描述]"
prompt: |
极简主义平面插画风格，高对比度视觉冲击。画面以鲜红色为底色，采用黑、白、红三色限定。核心构图：画面采用极强的对角线构图，以倾斜的地平线为视觉分界线，将空间切割为两个截然不同的色块区域：上方填充高饱和底色，下方呈现斑驳的浅色地表。主体：[在此处替换为主体描述]，呈现出纯净、发光的质感，轮廓跨越或靠近对角线。环境：背景中有一棵巨大的炭黑色枯萎古树，枝干呈放射状跨越对角线向四周延伸，带有水墨晕染与斑驳的裂纹质感。点缀：枝头停歇着几只纯白的飞鸟，地表呈现为大面积斑驳的银白色荒原，带有粗糙的矿物颗粒感和干笔刷痕迹。电影感广角比例，强烈的二次元平面感与写实纹理相结合，孤独、神圣、超现实的意境，线条锋利。
FILE:use-cases/index.yaml
# Image Forge Use-Case Index
# 每个用途条目关联：
# - references_file: 对应 references/ 下的 prompt 模板 JSON
# - recommended_signature: 推荐 Signature 风格（有独特视觉方案）
# - recommended_rendering: 推荐 Rendering 风格（通用技法）
# - default_backend: 该用途的默认后端
# - default_size: 推荐输出尺寸（GPT Image 2 尺寸规格）
# - default_aspect: Gemini 宽高比
#
# 调度决策优先级：
# 1. 用户显式指定后端/风格 → 覆盖一切
# 2. 用户指定 style → 按 style 的 preferred_backend
# 3. 无 style 指定 → 按 use-case 的 default_backend
use_cases:
- id: poster-flyer
label: "海报 / 传单"
aliases: [海报, 传单, 宣传单, poster, flyer, promotional poster]
references_file: "references/poster-flyer.json"
recommended_signature:
- constructivism # 几何构成，强视觉冲击
- risograph-magazine # 印刷质感
- high-contrast-industrial # 高对比度，适合科技/活动
- tri-color # 极简三色，适合文化类
recommended_rendering:
- cinematic-film-still # 电影感大片
- retro-vintage # 复古风宣传
- cyberpunk-sci-fi # 科技/电竞活动
default_backend: gpt-image-2 # 文字渲染是 GPT Image 2 核心优势
default_size: "1024x1536" # 竖版海报
default_aspect: "3:4"
- id: profile-avatar
label: "头像 / 肖像"
aliases: [头像, 肖像, 个人照, avatar, profile picture, portrait, pfp]
references_file: "references/profile-avatar.json"
recommended_signature:
- klein-blue-order # 极简干净，适合职业头像
- mixed-media # 线稿摄影融合，艺术感
- pop-ink-splash # 波普活力，适合社交头像
- glitch-window-v1 # 数字感，适合科技圈
recommended_rendering:
- anime-manga # 动漫头像
- illustration # 插画风肖像
- photography # 写真风
- chibi-q-style # Q 版可爱
- oil-painting # 油画人像
default_backend: nano-banana-2 # 人物风格迁移 Gemini 更灵活
default_size: "1024x1024"
default_aspect: "1:1"
- id: product-marketing
label: "产品营销图"
aliases: [产品图, 营销图, 产品海报, product marketing, product photo, 产品宣传]
references_file: "references/product-marketing.json"
recommended_signature:
- high-contrast-industrial # 科技产品强视觉
recommended_rendering:
- photography # 产品写真，最常用
- 3d-render # 3D 产品渲染
- minimalism # 极简留白，高级感
- cinematic-film-still # 电影感产品大片
default_backend: gpt-image-2 # 写实感、构图精度、光影
default_size: "1536x1024"
default_aspect: "4:3"
- id: ecommerce-main-image
label: "电商主图"
aliases: [电商, 主图, 详情页, 淘宝, ecommerce, product main image, 白底图]
references_file: "references/ecommerce-main-image.json"
recommended_signature: [] # 电商主图通常不需要强风格
recommended_rendering:
- photography # 白底产品摄影，标准电商风格
- 3d-render # 3D 产品展示
- minimalism # 干净背景
default_backend: gpt-image-2 # 细节还原、白底准确
default_size: "1024x1024" # 电商标准正方形
default_aspect: "1:1"
- id: youtube-thumbnail
label: "YouTube / 视频封面"
aliases: [视频封面, YouTube封面, thumbnail, 封面图, video cover]
references_file: "references/youtube-thumbnail.json"
recommended_signature:
- high-contrast-industrial # 高对比强吸引力
- tri-color # 极简高辨识度
recommended_rendering:
- cinematic-film-still # 电影感，点击率高
- photography # 真实感场景
- cyberpunk-sci-fi # 科技/游戏频道
default_backend: gpt-image-2 # 文字叠加 + 视觉冲击力
default_size: "1536x1024" # 16:9 横版
default_aspect: "16:9"
- id: social-media-post
label: "社交媒体配图"
aliases: [社交媒体, 小红书, ins, instagram, 朋友圈, social media, 配图]
references_file: "references/social-media-post.json"
recommended_signature:
- risograph-magazine # 印刷质感，ins 风
- pop-ink-splash # 波普活力
- tri-color # 极简时尚
- klein-blue-order # 清新夏日感
recommended_rendering:
- illustration # 插画配图
- photography # 生活写真
- watercolor # 清新水彩
- anime-manga # 二次元社区
default_backend: gpt-image-2
default_size: "1024x1024"
default_aspect: "1:1"
- id: app-web-design
label: "App / 网页设计素材"
aliases: [UI, app设计, 网页, 界面, app, web design, UI mockup, 设计稿]
references_file: "references/app-web-design.json"
recommended_signature: []
recommended_rendering:
- 3d-render # 3D UI 元素
- minimalism # 极简界面
- isometric # 等距风格图
default_backend: gpt-image-2 # UI 构图精度 + 文字渲染
default_size: "1536x1024"
default_aspect: "16:9"
- id: comic-storyboard
label: "漫画 / 分镜"
aliases: [漫画, 分镜, 故事板, comic, storyboard, manga panels, 格漫]
references_file: "references/comic-storyboard.json"
recommended_signature:
- glitch-window-v1 # 赛博朋克漫画
recommended_rendering:
- anime-manga # 日漫风格
- illustration # 欧美漫画
- sketch-line-art # 线稿分镜
- chibi-q-style # Q 版漫画
default_backend: nano-banana-2 # Gemini 漫画风格更丰富
default_size: "1024x1536"
default_aspect: "3:4"
- id: game-asset
label: "游戏素材"
aliases: [游戏, 素材, 游戏资产, game asset, game art, 角色设计, character design]
references_file: "references/game-asset.json"
recommended_signature:
- high-contrast-industrial # 科幻/赛博风游戏
recommended_rendering:
- 3d-render # 3D 游戏资产
- pixel-art # 像素游戏
- illustration # 插画风角色
- isometric # 等距游戏地图
- anime-manga # 日系角色
default_backend: nano-banana-2 # 角色设计 / 多风格 Gemini 更灵活
default_size: "1024x1024"
default_aspect: "1:1"
- id: infographic-edu-visual
label: "信息图 / 教育可视化"
aliases: [信息图, 数据可视化, 教育图, infographic, data visualization, 图表, edu visual]
references_file: "references/infographic-edu-visual.json"
recommended_signature: []
recommended_rendering:
- illustration # 插画风说明图
- isometric # 等距信息图
- minimalism # 极简清晰
default_backend: gpt-image-2 # 文字排版精准
default_size: "1024x1536"
default_aspect: "3:4"
- id: brand-logo-showcase
label: "Logo 展示图 / 品牌设计"
aliases: [logo, 图标, 品牌, brand, 徽标, icon, 标志, logo展示, 展示图, showcase, 品牌展示图]
references_file: "references/brand-logo-showcase.json"
recommended_signature: [high-contrast-industrial]
recommended_rendering: [minimalism, 3d-render]
logo_showcase_backgrounds:
dark: [logo-void, logo-frosted, logo-fluid-abyss, logo-spotlight, logo-analog-liquid, logo-led-matrix]
light: [logo-editorial-paper, logo-iridescent-frost, logo-morning-aura, logo-clinical, logo-ui-container, logo-swiss-flat]
default_backend: nano-banana-2 # 展示图默认用 Gemini
default_size: "1024x1024"
default_aspect: "1:1"
special_note: "SVG logo 代码生成请使用专属 logo-generator skill；本用途专注于 logo 展示图幕中的高端背景生成。待用户提供 PNG logo 后，读取 brand-logo-showcase.json 选择匹配的背景风格。"
- id: others
label: "其他 / 自由创作"
aliases: [其他, 自由, 随便, others, free, creative, 创意]
references_file: "references/others.json"
recommended_signature: []
recommended_rendering: []
default_backend: gpt-image-2
default_size: "1536x1024"
default_aspect: "16:9"
Figma design asset reading, code generation, and MCP integration. Covers REST API direct calls and MCP Server capabilities for design-to-code workflows. **Us...
---
name: figma
description: |
Figma design asset reading, code generation, and MCP integration. Covers REST API direct calls and MCP Server capabilities for design-to-code workflows.
**Use when**:
(1) Reading Figma file structure, components, styles, variables
(2) Generating frontend code from design files (React/Vue/HTML)
(3) Writing back to Figma canvas via MCP Server (create/modify frames, components, variables)
(4) Extracting design tokens (colors, spacing, typography) for code implementation
(5) User mentions "Figma", "design file", "component library", "design to code", "UI implementation"
(6) Integrating with Claude Code / Codex for Design-to-Code workflows
---
# Figma Skill
## Installation
### For OpenClaw agents (ClawHub)
```bash
clawhub install figma
```
### For Claude Code (MCP Server)
```bash
# Add Figma MCP (one-time, global)
claude mcp add --scope user --transport http figma https://mcp.figma.com/mcp
# Or install the official Figma plugin (includes Skills)
claude plugin install figma@claude-plugins-official
```
### For Codex CLI
```bash
codex mcp add figma --transport http https://mcp.figma.com/mcp
```
### Environment
Set `FIGMA_TOKEN` (Personal Access Token) in your environment or `.env` file.
Generate at: https://www.figma.com/settings → Account → Personal access tokens
---
## Quick Reference
### Two access methods
| Method | Use case | Auth |
|--------|----------|------|
| **REST API** (`scripts/figma_api.py`) | Read file structure, components, export images | `FIGMA_TOKEN` |
| **MCP Server** (`https://mcp.figma.com/mcp`) | Interactive design-to-code, write to canvas | OAuth (auto) |
---
## 1. REST API — Direct File Access
Use `scripts/figma_api.py` to query Figma files directly. Supports full URLs or file keys.
### Commands
```bash
# File structure
python3 scripts/figma_api.py pages <file_key_or_url>
# Frame tree (depth controls levels)
python3 scripts/figma_api.py tree <file_key> --node <page_id> --depth 3
# Published components (--group to aggregate by frame)
python3 scripts/figma_api.py components <file_key> --group
# Component sets
python3 scripts/figma_api.py component-sets <file_key>
# Styles
python3 scripts/figma_api.py styles <file_key>
# Export as PNG/SVG
python3 scripts/figma_api.py export <file_key> --nodes <id1>,<id2> --format png --scale 2
# Node JSON detail
python3 scripts/figma_api.py node <file_key> --node <node_id> --depth 4
# Current user info
python3 scripts/figma_api.py me
```
---
## 2. MCP Server — 14 Tools
### Read tools (any Figma plan)
| Tool | Purpose |
|------|---------|
| `get_design_context` | Design context → code (default React+Tailwind, customizable) |
| `get_variable_defs` | Variables and styles (colors, spacing, typography) |
| `get_metadata` | Sparse XML: layer IDs, names, types, positions, sizes |
| `get_screenshot` | Screenshot of selection |
| `get_code_connect_map` | Figma node → code component mapping (needs Code Connect) |
| `search_design_system` | Search library components, variables, styles |
| `whoami` | Authenticated user info |
### Write tools (beta free, needs Full seat + edit permission)
| Tool | Purpose |
|------|---------|
| `use_figma` | Execute Figma Plugin API JS: create/modify frames, components, variables, auto layout |
| `generate_figma_design` | Convert live browser UI → editable Figma layers |
| `create_new_file` | Create new file in Drafts |
| `generate_diagram` | Mermaid syntax → FigJam diagram |
### Helper tools
| Tool | Purpose |
|------|---------|
| `create_design_system_rules` | Generate design system rules file for AI code generation |
| `add_code_connect_map` | Add Figma node → code component mapping |
| `get_code_connect_suggestions` | Code Connect mapping suggestions |
### Write limitations
- 20KB response limit per call
- No image/asset import support
- Full seat required (Dev seat = read-only)
- Large changes: inspect first → incremental create/update → verify
---
## 3. Design-to-Code Workflow
### For Claude Code / Codex (via MCP)
**Step 1** — Provide the Figma frame URL in your prompt:
```
Using this Figma frame: https://www.figma.com/design/<key>?node-id=<id>
Generate React components using [your component library].
```
**Step 2** — For large pages, ask the agent to inspect first:
```
1. Use get_metadata on the frame to understand the structure
2. Use search_design_system to find matching library components
3. Use get_design_context on each section to generate code
```
**Step 3** — Write back to Figma (if editor permission):
```
Using this file: <url>, create a new page and build [description]
using existing components. Use auto layout.
```
### For OpenClaw agents (via REST API)
1. `figma_api.py pages <url>` → identify target page
2. `figma_api.py tree <key> --node <page_id>` → understand structure
3. `figma_api.py export <key> --nodes <frame_id>` → get screenshot
4. `figma_api.py node <key> --node <frame_id> --depth 4` → get component details
5. Generate code using the structural data + screenshot
### Best practices
- Always specify frontend framework and component library in prompts
- Large pages: `get_metadata` overview first, then `get_design_context` per section
- Use `get_screenshot` for layout-sensitive components
- Rate limit: space REST API calls ≥500ms apart
- Keep `depth` ≤ 4 to avoid timeouts on large files
---
## 4. Asset Registry
Register your Figma files in `references/omada-assets.md` for quick lookup.
See `references/guide-for-agents.md` for the complete agent operation manual.
See `references/guide-for-humans.md` for the human-facing usage guide.
FILE:references/guide-for-agents.md
# Figma Skill — AI Agent 操作手册
> 本文档面向 Jarvis、jarvis-exec 或其他 AI Agent，提供 Figma 操作的完整指引。
---
## 1. 环境与认证
```bash
# Token 在 ~/.openclaw/.env 中（自动加载）
export FIGMA_TOKEN="$FIGMA_TOKEN"
# 脚本路径
FIGMA_CLI="~/.openclaw/workspace/skills/figma/scripts/figma_api.py"
```
---
## 2. REST API 操作速查
### 2.1 文件探索
```bash
# 查看文件结构
python3 $FIGMA_CLI pages <file_key_or_url>
# 查看特定页面的 Frame 树
python3 $FIGMA_CLI tree <file_key> --node <page_id> --depth 3
# 获取节点详细 JSON（用于分析组件结构）
python3 $FIGMA_CLI node <file_key> --node <node_id> --depth 4 --max-chars 15000
```
### 2.2 组件库分析
```bash
# 查看组件分布（按 frame 分组）
python3 $FIGMA_CLI components <file_key> --group
# 列出所有 component sets
python3 $FIGMA_CLI component-sets <file_key>
# 查看 styles
python3 $FIGMA_CLI styles <file_key>
```
### 2.3 导出截图
```bash
# 导出节点为 PNG（2x）
python3 $FIGMA_CLI export <file_key> --nodes <id1>,<id2> --format png --scale 2
# 导出为 SVG
python3 $FIGMA_CLI export <file_key> --nodes <id> --format svg --output /tmp/icon.svg
```
### 2.4 原始 API 调用
当脚本不覆盖的场景，直接用 curl：
```bash
# 获取文件变量
curl -s -H "X-Figma-Token: $FIGMA_TOKEN" \
"https://api.figma.com/v1/files/<file_key>/variables/local"
# 搜索组件（按名称）
curl -s -H "X-Figma-Token: $FIGMA_TOKEN" \
"https://api.figma.com/v1/files/<file_key>/components" | \
python3 -c "import json,sys; [print(c['name']) for c in json.load(sys.stdin)['meta']['components'] if 'button' in c['name'].lower()]"
```
---
## 3. MCP Server 操作（通过 Claude Code / Codex）
### 3.1 配置
```bash
# Claude Code（全局安装）
claude mcp add --scope user --transport http figma https://mcp.figma.com/mcp
# 验证
claude mcp list
```
### 3.2 Design-to-Code Prompt 模板
**基础代码生成**：
```
Using this Figma frame: https://www.figma.com/design/<key>?node-id=<id>
Generate React components using Ant Design. Read the component library at
https://www.figma.com/design/gzLJeRunJYuB02zQKTOkva for design system reference.
```
**带组件映射的代码生成**：
```
1. First, use get_metadata on the frame to understand the structure
2. Then use search_design_system to find matching components from the library
3. Finally, use get_design_context to generate React + Tailwind code
Map Figma components to Ant Design equivalents.
```
**写回 Figma**：
```
Using this Figma file: https://www.figma.com/design/<key>
Create a new page called "AI Generated" and build a [description]
using existing components from the file's design system.
Use auto layout and follow the existing naming conventions.
```
### 3.3 MCP 工具选择决策树
```
需要生成代码？
├─ 小区域 → get_design_context（直接返回代码）
├─ 大页面 → get_metadata（概览）→ 分区域调 get_design_context
└─ 需要截图辅助 → get_screenshot + get_design_context
需要了解设计系统？
├─ 查组件 → search_design_system
├─ 查变量 → get_variable_defs
└─ 查映射 → get_code_connect_map（需 Code Connect）
需要写入 Figma？
├─ 创建/修改设计 → use_figma（Plugin API JS）
├─ 代码截图转设计 → generate_figma_design
└─ 创建新文件 → create_new_file
```
---
## 4. Omada 资产快速索引
| 缩写 | file_key | 用途 |
|------|----------|------|
| `WEB_LIB` | `gzLJeRunJYuB02zQKTOkva` | WEB 组件库（1743 组件） |
| `APP_LIB` | `beYqvBsrUqRoq6GNfvOAuN` | APP 组件库（848 组件） |
| `WEB_REQ` | `fA9Oq6TPbayJsQUSYjyV4s` | Web 需求设计稿集合 |
| `V62` | `DtbxwhppKkdqJncPhlH74c` | Controller V6.2 |
| `AIO_GW` | `rZaHc0WcrPLWFOlM3OqppI` | AIO 1.0 Gateway |
| `DC_V11` | `yjjan3lcHDRsYpdeiDQ6s0` | Design Center V1.1 |
| `APP_52` | `iu6lq4cRZUTwZjPx0QSeaX` | Omada APP 5.2 |
详细资产清单见 `references/omada-assets.md`。
---
## 5. 注意事项
- **Rate Limit**：REST API 请求间隔 ≥500ms，避免被限流
- **depth 参数**：不要超过 4，大文件会超时
- **viewer 角色**：当前只能读取，写入操作会返回 403
- **URL 解析**：脚本支持传入完整 Figma URL，自动提取 file_key
- **大文件分页**：组件数 >200 时，REST API 不分页全量返回（注意响应大小）
- **node_id 格式**：Figma 用 `123:456` 格式，URL 中编码为 `123-456`，传 API 时用 `:` 分隔
FILE:references/guide-for-humans.md
# Figma × AI 工作流指南（面向人类）
> 本指南面向 Omada 产品和研发团队，说明如何利用 Figma MCP + AI Agent 提升设计到代码的效率。
---
## 一、能做什么？
### 1. 设计稿 → 前端代码（最成熟）
把 Figma 设计稿喂给 AI（Claude Code / Codex），自动生成 React/Vue/HTML 代码。
- 支持指定使用现有组件库（如 Ant Design），AI 不会重新造轮子
- 预计前端实现时间缩短 40-70%
### 2. 运行中的页面 → Figma 设计图层（新能力）
Claude Code 开发的页面可以直接 `Send to Figma`，变成可编辑的 Figma 图层。
- 设计师可以直接在上面批注和修改
- 适合"代码先行"的原型迭代场景
### 3. 组件库审查与对齐
AI 可以读取组件库，对比设计稿中实际使用的组件与组件库定义，发现不一致。
### 4. 设计 Token 提取
自动提取颜色、间距、字体等设计变量，生成 CSS Variables / JS Token 文件。
---
## 二、前置条件
| 条件 | 状态 | 说明 |
|------|------|------|
| Figma Personal Access Token | ✅ 已配置 | 用于 API 读取 |
| 组件库 Published | ✅ 1743+848 组件 | WEB + APP 两套完整组件库 |
| MCP Server | ✅ 免费可用 | Remote MCP，无需桌面端 |
| Code Connect | ❌ 需 Org plan | 组件库到代码仓库的映射（可选，没有也能工作） |
| Editor 权限 | ⚠️ 当前 viewer | 写回 Figma 需要 editor 权限 |
---
## 三、推荐工作流
### 场景 A：新功能前端实现
```
设计师完成 Figma → 复制 Frame 链接 → 告诉 Jarvis/Claude Code
→ AI 读取设计 + 组件库 → 生成代码 → 开发者审查/微调
```
**人类需要做的**：提供 Figma Frame 链接 + 指定前端框架/组件库
### 场景 B：组件库覆盖度审查
```
Jarvis 扫描项目文件 → 对比组件库 → 输出未覆盖的组件清单
```
**人类需要做的**：提供项目文件链接
### 场景 C：快速原型
```
告诉 AI 功能需求 → AI 生成代码 → Send to Figma → 设计师审查
```
**人类需要做的**：描述需求，设计师在 Figma 中审查
---
## 四、局限性
1. **无法自动发现文件**：Figma API 不支持列出 workspace 所有文件（非 Enterprise 限制），新文件需手动告知
2. **Code Connect 需 Org plan**：组件库 ↔ 代码仓库的精确映射需要升级 Figma plan
3. **写回 Figma 需 Editor 权限**：当前账号是 viewer，需要文件所有者授予编辑权限才能使用 `use_figma` 写入
4. **图片/资源不支持写入**：MCP write 目前不能导入图片资源
5. **大文件可能超时**：超大设计文件的 API 请求需要分页处理
---
## 五、常见问题
**Q: 不升级 Org plan，AI 生成的代码质量如何？**
A: 仍然可用。AI 通过读取组件库的 published components 名称和结构来推断映射关系。Code Connect 只是让映射更精确，没有它也能工作。
**Q: 需要设计师改变工作习惯吗？**
A: 不需要。设计师正常在 Figma 中工作，只是多了一个"给 AI 提供 Frame 链接"的步骤。
**Q: 安全性如何？**
A: Figma Token 仅存在于 Jarvis 本地环境，不会外传。MCP OAuth 由 Figma 官方处理。设计文件数据不会离开 Figma 和 AI 的加密通道。
FILE:references/omada-assets.md
# Omada Figma 设计资产全景
> 最后更新：2026-03-30
> 数据来源：Figma REST API 自动扫描
---
## 1. 组件库
### 🌟 商用WEB组件库
- **file_key**: `gzLJeRunJYuB02zQKTOkva`
- **URL**: https://www.figma.com/design/gzLJeRunJYuB02zQKTOkva
- **最后更新**: 2026-03-30
- **Published Components**: 1,743
- **Component Sets**: 132
- **Pages**: 86
- **Figma 账号角色**: viewer
#### 组件分类（按 containing_frame 统计 Top 20）
| Frame | 数量 | 说明 |
|-------|------|------|
| icon 图标 | 222 | 全局图标 |
| Sidebar icon导航栏图标 | 111 | 侧边栏导航图标 |
| Input 输入框 | 106 | 各类输入组件 |
| 图标dark | 94 | Dark 模式图标 |
| Table icon 表格图标 | 90 | 表格内图标 |
| 插图 | 85 | 业务插图 |
| 组件 | 64 | 通用基础组件 |
| Button 按钮 | 55 | 按钮变体 |
| Table 表格 | 54 | 表格组件 |
| Client 终端用户 | 51 | 终端设备图标 |
| Select 选择器 | 43 | 选择器变体 |
| DatePicker 日期选择框 | 40 | 日期选择 |
| Product 设备 | 36 | 产品设备图 |
| topo设备 | 33 | 拓扑图设备 |
| Tag 标签 | 32 | 标签变体 |
| InputNumber 数字输入框 | 30 | 数字输入 |
| 图表组成 | 30 | 图表元素 |
| IP地址 | 28 | IP 输入组件 |
| Ip掩码/端口 输入框 | 27 | IP/端口输入 |
| 交互自定义 | 25 | 交互组件 |
*另有 64 个分组（共 84 个 frame）*
#### 页面结构（Design System 分层）
**通用 Basic**：Color、Text、Button、Divider、Shadow、Layout、模式库
**导航 Navigation**：Top Bar、Sidebar、Tabs、Steps、Dropdown、Pagination
**数据输入 Input**：Form、Input、数字输入框、密码输入框、MAC输入框、IP掩码/端口输入框、Switch、Slider、Checkbox、Cascader、Radio、Select、TreeSelect、TimePicker、DatePicker、ColorPicker、Transfer、Search、Upload、Schedule
**数据展示 Display**：Table、描述列表、Tag、Collapse、Calendar、Popover、新功能引导、Tooltip、Avatar、Badge
**通知提示 Notification**：Alert、Toast Message、Drawer、Dialog、Notification、Progress、Loading
**其他 Other**：设备端口、Chart、Mask、Rate、Empty Space
**安防类组件+模式库**：安防业务专属组件
---
### 🌟 商用APP组件库
- **file_key**: `beYqvBsrUqRoq6GNfvOAuN`
- **URL**: https://www.figma.com/design/beYqvBsrUqRoq6GNfvOAuN
- **最后更新**: 2026-03-27
- **Published Components**: 848
- **Component Sets**: 77
- **Pages**: 49
#### 组件分类（Top 20）
| Frame | 数量 |
|-------|------|
| 图标 | 218 |
| Page Controls | 140 |
| General | 132 |
| List 列表 | 96 |
| 插图 | 31 |
| Buttons 按钮 | 30 |
| Device List | 28 |
| Inputs 输入框 | 27 |
| Toast | 16 |
| 标签色 | 15 |
| Text 文本 | 12 |
| Navigation Bars 标题栏 | 12 |
| Switch 开关 | 10 |
| Bottom Popup 底部弹窗 | 10 |
#### 页面结构
**基础组件**：Color、Typefaces、Navigation Bar、List、表单、模态面板、Inputs、Buttons、Tag、Toast、Bottom Popup、Segments、Menu、Search Bars、Tab Bars、Switch、Slider、Time Pickers、Date Pickers、Alert、Action Sheets、Page Controls、Prompt、Text、SubHeader、Loading
**样式库**：设备列表、图表、端口
**无障碍适配**：规范概要、标注、入口/功能
---
## 2. 项目设计文件
### Omada项目-web 2（需求设计稿集合）
- **file_key**: `fA9Oq6TPbayJsQUSYjyV4s`
- **URL**: https://www.figma.com/design/fA9Oq6TPbayJsQUSYjyV4s
- **Pages**: 54（按 JIRA ticket 组织，✅ 标记已完成）
- **组织方式**：按功能模块分组（Network Config / Device Config / Hotspot / General 等）
### V6.2-Omada Controller
- **file_key**: `DtbxwhppKkdqJncPhlH74c`
- **URL**: https://www.figma.com/design/DtbxwhppKkdqJncPhlH74c
- **Pages**: 8
- **核心功能**：
- SMBNET-1855 千兆/2.5G混堆叠
- SMBNET-1295 WAN 抖动/丢包上报
- SMBNET-654 Dashboard Speed Test
- SMBNET-55 AI RRM (WLAN Optimization)
- SMBNET-2076 ISP Load Dashboard
- Add Device 达阈值
### AIO 1.0-Gateway相关
- **file_key**: `rZaHc0WcrPLWFOlM3OqppI`
- **URL**: https://www.figma.com/design/rZaHc0WcrPLWFOlM3OqppI
- **Pages**: 19
- **核心功能**：DPI 优化、Lightlink VPN (Server/Client)、Content Filtering、QoS、DDNS、Loadbalance、Auto Send Data to Mail
### Design Center-V1.1
- **file_key**: `yjjan3lcHDRsYpdeiDQ6s0`
- **URL**: https://www.figma.com/design/yjjan3lcHDRsYpdeiDQ6s0
- **Pages**: 44
- **核心功能**：
- V1.0：Project Workspace、Wizard、设备清单、导出、Floor Plan、墙体绘制、设备布放、设备仿真
- V1.1：无线仿真、Topology 呈现、弱电井/桥架/布线、PDF 导出、联动 Omada Store
- V1.2：AI 售前工具 & AI 运维
### V5.2-omada App
- **file_key**: `iu6lq4cRZUTwZjPx0QSeaX`
- **URL**: https://www.figma.com/design/iu6lq4cRZUTwZjPx0QSeaX
- **Pages**: 18
- **核心功能**：Time Range Setting、DDNS 适配、2FA 流程优化、VLAN 配置、设备收养、SSID 优化、拓扑适配、蓝牙 Console、Portal Voucher
---
## 3. 发现说明
### API 限制
- Figma REST API 无法通过 Personal Access Token 列出 workspace 所有文件（需 team_id，非 Enterprise 不暴露）
- 新文件需手动提供链接后，由 `figma_api.py` 扫描注册
- 当前用户角色为 `viewer`，可读取所有设计数据但无法写入（MCP write 需要 editor 权限）
### 组件库特征
- WEB 组件库高度完整，覆盖网络管理软件全场景，活跃维护中
- 组件命名中英混合（如 `Button 按钮`、`Select 选择器`），AI 可理解
- 未发布 Styles（颜色/字体通过 page 内规范文档定义而非 Figma Styles API）
- 未发布 Variables（Figma Variables 功能可能未采用）
FILE:scripts/figma_api.py
#!/usr/bin/env python3
"""Figma REST API CLI tool for reading design files, components, and exporting assets."""
import argparse
import json
import os
import re
import sys
import urllib.request
import urllib.error
from collections import Counter
TOKEN = os.environ.get("FIGMA_TOKEN", "")
BASE = "https://api.figma.com/v1"
def _get(path: str) -> dict:
"""Make authenticated GET request to Figma API."""
url = f"{BASE}/{path}"
req = urllib.request.Request(url, headers={"X-Figma-Token": TOKEN})
try:
with urllib.request.urlopen(req, timeout=60) as resp:
return json.loads(resp.read())
except urllib.error.HTTPError as e:
body = e.read().decode() if e.fp else ""
print(f"Error {e.code}: {body}", file=sys.stderr)
sys.exit(1)
def extract_file_key(url_or_key: str) -> str:
"""Extract file_key from a Figma URL or return as-is if already a key."""
m = re.search(r"figma\.com/(?:design|file)/([a-zA-Z0-9]+)", url_or_key)
return m.group(1) if m else url_or_key
def cmd_pages(args):
"""List pages in a file."""
fk = extract_file_key(args.file_key)
d = _get(f"files/{fk}?depth=1")
print(f"File: {d.get('name')} (role: {d.get('role', '?')})")
print(f"Last modified: {d.get('lastModified')}")
pages = d.get("document", {}).get("children", [])
print(f"\nPages ({len(pages)}):")
for p in pages:
print(f" [{p['id']}] {p['name']}")
def cmd_tree(args):
"""Show frame tree of a node."""
fk = extract_file_key(args.file_key)
node_id = args.node
depth = args.depth or 3
d = _get(f"files/{fk}/nodes?ids={node_id}&depth={depth}")
nodes = d.get("nodes", {})
for nid, ndata in nodes.items():
doc = ndata.get("document", {})
_print_tree(doc, 0, args.max_children or 20)
def _print_tree(node: dict, indent: int, max_children: int):
t = node.get("type", "")
n = node.get("name", "")
nc = len(node.get("children", []))
extra = ""
if t == "TEXT":
chars = node.get("characters", "")
if chars:
extra = f' "{chars[:80]}"'
elif t == "INSTANCE":
comp = node.get("componentId", "")
if comp:
extra = f" (componentId: {comp})"
suffix = f" [{nc} children]" if nc > 0 else ""
print(" " * indent + f"[{t}] {n}{extra}{suffix}")
children = node.get("children", [])
for i, c in enumerate(children[:max_children]):
_print_tree(c, indent + 1, max_children)
if len(children) > max_children:
print(" " * (indent + 1) + f"... +{len(children) - max_children} more")
def cmd_components(args):
"""List published components."""
fk = extract_file_key(args.file_key)
d = _get(f"files/{fk}/components")
comps = d.get("meta", {}).get("components", [])
print(f"Published components: {len(comps)}")
if args.group:
frames = Counter(c.get("containing_frame", {}).get("name", "?") for c in comps)
print("\nBy containing frame:")
for f, cnt in frames.most_common(args.limit or 30):
print(f" {f}: {cnt}")
else:
for c in comps[: args.limit or 50]:
frame = c.get("containing_frame", {}).get("name", "?")
print(f" - {c.get('name', '?')} | frame: {frame} | id: {c.get('node_id', '?')}")
if len(comps) > (args.limit or 50):
print(f" ... +{len(comps) - (args.limit or 50)} more (use --limit to show more)")
def cmd_component_sets(args):
"""List component sets."""
fk = extract_file_key(args.file_key)
d = _get(f"files/{fk}/component_sets")
sets = d.get("meta", {}).get("component_sets", [])
print(f"Component sets: {len(sets)}")
for s in sets[: args.limit or 50]:
print(f" - {s.get('name', '?')} | id: {s.get('node_id', '?')}")
def cmd_styles(args):
"""List styles."""
fk = extract_file_key(args.file_key)
d = _get(f"files/{fk}/styles")
styles = d.get("meta", {}).get("styles", [])
print(f"Styles: {len(styles)}")
types = Counter(s.get("style_type") for s in styles)
if types:
print("\nBy type:")
for t, c in types.most_common():
print(f" {t}: {c}")
for s in styles[: args.limit or 30]:
print(f" - [{s.get('style_type')}] {s.get('name')}")
def cmd_export(args):
"""Export nodes as images."""
fk = extract_file_key(args.file_key)
nodes = args.nodes
fmt = args.format or "png"
scale = args.scale or 2
d = _get(f"images/{fk}?ids={nodes}&format={fmt}&scale={scale}")
images = d.get("images", {})
if d.get("err"):
print(f"Error: {d['err']}", file=sys.stderr)
return
for nid, url in images.items():
if url:
out = args.output or f"/tmp/figma_{nid.replace(':', '_')}.{fmt}"
urllib.request.urlretrieve(url, out)
print(f"Saved: {out}")
else:
print(f"No image for node {nid}")
def cmd_node(args):
"""Get detailed node JSON."""
fk = extract_file_key(args.file_key)
depth = args.depth or 4
d = _get(f"files/{fk}/nodes?ids={args.node}&depth={depth}")
nodes = d.get("nodes", {})
for nid, ndata in nodes.items():
doc = ndata.get("document", {})
print(json.dumps(doc, indent=2, ensure_ascii=False)[:args.max_chars or 10000])
def cmd_me(args):
"""Show current user info."""
d = _get("me")
print(json.dumps(d, indent=2))
def main():
if not TOKEN:
print("Error: FIGMA_TOKEN not set. Export it or add to ~/.openclaw/.env", file=sys.stderr)
sys.exit(1)
parser = argparse.ArgumentParser(description="Figma REST API CLI")
sub = parser.add_subparsers(dest="command", required=True)
p = sub.add_parser("pages", help="List file pages")
p.add_argument("file_key")
p.set_defaults(func=cmd_pages)
p = sub.add_parser("tree", help="Show frame tree")
p.add_argument("file_key")
p.add_argument("--node", required=True)
p.add_argument("--depth", type=int, default=3)
p.add_argument("--max-children", type=int, default=20)
p.set_defaults(func=cmd_tree)
p = sub.add_parser("components", help="List published components")
p.add_argument("file_key")
p.add_argument("--group", action="store_true", help="Group by containing frame")
p.add_argument("--limit", type=int)
p.set_defaults(func=cmd_components)
p = sub.add_parser("component-sets", help="List component sets")
p.add_argument("file_key")
p.add_argument("--limit", type=int)
p.set_defaults(func=cmd_component_sets)
p = sub.add_parser("styles", help="List styles")
p.add_argument("file_key")
p.add_argument("--limit", type=int)
p.set_defaults(func=cmd_styles)
p = sub.add_parser("export", help="Export nodes as images")
p.add_argument("file_key")
p.add_argument("--nodes", required=True, help="Comma-separated node IDs")
p.add_argument("--format", choices=["png", "svg", "jpg", "pdf"], default="png")
p.add_argument("--scale", type=float, default=2)
p.add_argument("--output", help="Output file path")
p.set_defaults(func=cmd_export)
p = sub.add_parser("node", help="Get node JSON")
p.add_argument("file_key")
p.add_argument("--node", required=True)
p.add_argument("--depth", type=int, default=4)
p.add_argument("--max-chars", type=int, default=10000)
p.set_defaults(func=cmd_node)
p = sub.add_parser("me", help="Show current user")
p.set_defaults(func=cmd_me)
args = parser.parse_args()
args.func(args)
if __name__ == "__main__":
main()
微信读书（WeRead）数据查询与笔记管理技能。获取书架、搜索书籍、查看阅读进度/时长、获取笔记划线、热门书评、章节信息、随机笔记抽取和批量导出。当用户提到"微信读书"、"WeRead"、"书架"、"读书进度"、"划线"、"笔记"、"书评"、"在读"、"读完"、"阅读时长"、"读书回顾"、"导出笔记"时使用。
---
name: weread
description: 微信读书（WeRead）数据查询与笔记管理技能。获取书架、搜索书籍、查看阅读进度/时长、获取笔记划线、热门书评、章节信息、随机笔记抽取和批量导出。当用户提到"微信读书"、"WeRead"、"书架"、"读书进度"、"划线"、"笔记"、"书评"、"在读"、"读完"、"阅读时长"、"读书回顾"、"导出笔记"时使用。
---
# 微信读书 Skill
通过微信读书 Web API 获取用户的阅读数据。依赖 Cookie 认证。
## 前置：Cookie 配置
**Cookie 存储路径: `~/.weread/cookie`（纯文本单行）。**
所有 API 命令依赖此文件。Cookie 不存在或过期时，命令会报错并提示重新登录。
### 获取 Cookie（三选一）
**方式 1 — 浏览器自动提取（推荐）：**
1. 用 `browser` 工具打开 `https://weread.qq.com`（profile=openclaw 或 user）
2. 确认已登录（页面显示用户头像/书架）
3. 执行 JavaScript 提取 cookie：`document.cookie`
4. 将结果写入 `~/.weread/cookie`
**方式 2 — 手动粘贴：**
```bash
python3 ~/.openclaw/workspace/skills/weread/scripts/weread_login.py paste
```
**方式 3 — 直接写入：**
用户提供 cookie 字符串后直接写入 `~/.weread/cookie`。
### 验证 Cookie
```bash
python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py verify
```
Cookie 过期时 API 返回 errcode -2012，提示用户重新登录。
## 命令参考
脚本目录: `skills/weread/scripts/`（相对于 workspace）
执行示例使用绝对路径: `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py`
| 命令 | 说明 | 示例 |
|------|------|------|
| `shelf` | 获取书架（书名、作者、进度、阅读时长） | `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py shelf` |
| `search <keyword>` | 搜索书架中的书籍（模糊匹配书名/作者） | `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py search 心理学` |
| `info <bookId>` | 获取书籍详情（评分、字数、出版信息） | `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py info 123456` |
| `progress <bookId>` | 获取阅读进度 | `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py progress 123456` |
| `detail <bookId>` | 获取详细阅读信息（时长明细、完成日期） | `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py detail 123456` |
| `bookmarks <bookId>` | 获取划线记录 | `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py bookmarks 123456` |
| `reviews <bookId>` | 获取我的笔记/想法 | `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py reviews 123456` |
| `best-reviews <bookId> [count]` | 获取热门书评 | `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py best-reviews 123456 5` |
| `chapters <bookId>` | 获取章节信息 | `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py chapters 123456` |
| `verify` | 验证 Cookie 是否有效 | `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py verify` |
### 辅助脚本
| 脚本 | 说明 | 示例 |
|------|------|------|
| `random_review.py` | 随机抽取读书笔记（加权，更长更优先） | `python3 ~/.openclaw/workspace/skills/weread/scripts/random_review.py --count 2 --min-length 50` |
| `export_notes.py` | 全量导出笔记到 `~/.weread/` | `python3 ~/.openclaw/workspace/skills/weread/scripts/export_notes.py` |
| `export_notes.py --stats` | 显示笔记统计信息 | `python3 ~/.openclaw/workspace/skills/weread/scripts/export_notes.py --stats` |
| `weread_login.py paste` | 手动粘贴 Cookie 登录 | `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_login.py paste` |
## 典型工作流
### 查看书架概况
```bash
python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py shelf
```
返回所有书籍的 bookId、书名、作者、进度百分比、阅读时长、是否读完。
### 获取某本书的笔记和划线
1. 先搜索获取 bookId：`python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py search 思考快与慢`
2. 用 bookId 获取划线：`python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py bookmarks <bookId>`
3. 用 bookId 获取笔记：`python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py reviews <bookId>`
4. 需要章节信息时：`python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py chapters <bookId>`
### 查看阅读统计
1. `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py shelf` 获取整体阅读时长
2. `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py detail <bookId>` 获取单本详细阅读数据
### 查看热门书评
1. 搜索获取 bookId
2. `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py best-reviews <bookId> 10`
### 随机读书回顾
适用于晨报、日报或任何需要读书笔记回顾的场景：
```bash
python3 ~/.openclaw/workspace/skills/weread/scripts/random_review.py --count 2 --min-length 50 --format text
```
- `--count N`：抽取 N 条笔记（默认 3）
- `--min-length L`：最小字数过滤（默认 20）
- `--format json|text`：输出格式（默认 text）
需要先运行 `export_notes.py` 导出笔记数据。
## 输出格式
所有命令输出 JSON。关键字段：
- `bookId`: 书籍唯一标识（后续命令的入参）
- `progress`: 阅读进度（0-100）
- `readingTime`: 阅读时长（已格式化为 X小时X分钟）
- `finishReading`: 是否已读完
- `markText`: 划线文本
- `content`: 笔记内容
## Cookie 过期处理
当命令报错 "Cookie 已过期" 时：
1. 用浏览器工具打开 `https://weread.qq.com`，确认登录态
2. 提取 `document.cookie` 写入 `~/.weread/cookie`
3. 运行 `python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py verify` 确认
FILE:README.md
# OpenClaw WeRead Skill
> OpenClaw skill for 微信读书 (WeRead) — fetch bookshelf, notes, highlights and reading stats.
[中文说明](#中文说明)
## What It Does
This skill gives your OpenClaw agent full access to your WeRead (微信读书) data:
- **Bookshelf** — list all books with progress, reading time, and completion status
- **Search** — fuzzy-match books by title or author
- **Book Info** — ratings, word count, publisher details
- **Reading Progress & Stats** — per-book progress, detailed time breakdowns
- **Highlights** — all your underlined passages
- **Notes / Thoughts** — your personal annotations
- **Best Reviews** — top community reviews
- **Chapters** — table of contents and chapter metadata
- **Cookie Verification** — check if your session is still active
- **Random Review** — pick random notes for daily review / morning reports
- **Notes Export** — bulk export all notes to structured local JSON files
## Prerequisites
- Python 3.9+
- [OpenClaw](https://github.com/nicepkg/openclaw) installed and configured
- A 微信读书 (WeRead) account with an active web session
## Installation
### Option 1: via clawhub (recommended)
```bash
clawhub install ChenyqThu/openclaw-weread-skill
```
### Option 2: manual clone
```bash
git clone https://github.com/ChenyqThu/openclaw-weread-skill.git \
~/.openclaw/workspace/skills/weread
```
## First-time Setup
The skill authenticates via a browser cookie stored at `~/.weread/cookie`.
### How to get your cookie
**Method 1 — Browser auto-extract (recommended):**
Ask your OpenClaw agent to open `https://weread.qq.com`, confirm you're logged in, then extract `document.cookie` and save it to `~/.weread/cookie`.
**Method 2 — Paste from DevTools:**
```bash
python3 ~/.openclaw/workspace/skills/weread/scripts/weread_login.py paste
```
Open WeRead in your browser → F12 → Network → pick any request → Headers → copy the `Cookie` value → paste when prompted.
**Method 3 — Direct write:**
If you already have the cookie string, write it directly:
```bash
mkdir -p ~/.weread
echo 'YOUR_COOKIE_STRING' > ~/.weread/cookie
chmod 600 ~/.weread/cookie
```
### Verify your cookie
```bash
python3 ~/.openclaw/workspace/skills/weread/scripts/weread_api.py verify
```
> When the cookie expires, commands will return "Cookie 已过期". Simply re-do the setup above.
## Commands Reference
All commands output JSON to stdout. Errors go to stderr with exit code 1.
### Core API (`weread_api.py`)
| Command | Description | Example |
|---------|-------------|---------|
| `shelf` | List all books on your shelf | `python3 scripts/weread_api.py shelf` |
| `search <keyword>` | Fuzzy search books by title/author | `python3 scripts/weread_api.py search 心理学` |
| `info <bookId>` | Book details (rating, publisher, etc.) | `python3 scripts/weread_api.py info 123456` |
| `progress <bookId>` | Reading progress for a book | `python3 scripts/weread_api.py progress 123456` |
| `detail <bookId>` | Detailed reading stats (time breakdown) | `python3 scripts/weread_api.py detail 123456` |
| `bookmarks <bookId>` | All highlights / underlines | `python3 scripts/weread_api.py bookmarks 123456` |
| `reviews <bookId>` | Your notes / thoughts | `python3 scripts/weread_api.py reviews 123456` |
| `best-reviews <bookId> [N]` | Top N community reviews (default 10) | `python3 scripts/weread_api.py best-reviews 123456 5` |
| `chapters <bookId>` | Chapter list and metadata | `python3 scripts/weread_api.py chapters 123456` |
| `verify` | Check if cookie is still valid | `python3 scripts/weread_api.py verify` |
> **How to find `bookId`:** Run `shelf` or `search <keyword>` first — every result includes a `bookId` field you can pass to other commands.
### Random Review (`random_review.py`)
Pick random notes for daily review or morning reports:
```bash
python3 scripts/random_review.py --count 3 --min-length 50 --format text
```
| Flag | Default | Description |
|------|---------|-------------|
| `--count N` | 3 | Number of notes to pick |
| `--min-length L` | 20 | Minimum note length (filters short ones) |
| `--format` | text | Output format: `text` or `json` |
### Notes Export (`export_notes.py`)
Bulk export all notes to `~/.weread/`:
```bash
python3 scripts/export_notes.py # full export
python3 scripts/export_notes.py --stats # show statistics
```
### Login Helper (`weread_login.py`)
```bash
python3 scripts/weread_login.py paste # paste cookie from DevTools
python3 scripts/weread_login.py chrome # extract from Chrome (may need decryption)
```
## Morning Report Integration
The `random_review.py` script is designed to integrate with any morning/daily report system. Example:
```bash
# Pick 2 quality notes (min 50 chars) as JSON
python3 ~/.openclaw/workspace/skills/weread/scripts/random_review.py \
--count 2 --min-length 50 --format json
```
You can pipe the output into your report generator, cron job, or agent workflow. Notes are weighted by length so more thoughtful annotations are more likely to be selected.
## Data Privacy
- All data is stored **locally** under `~/.weread/` — nothing is uploaded anywhere.
- Your cookie file (`~/.weread/cookie`) is created with `chmod 600` (owner-only read/write).
- The `.gitignore` in this repo explicitly excludes all user data files.
- No analytics, no telemetry, no third-party services.
## License
MIT
---
# 中文说明
## 功能概述
这是一个 OpenClaw 技能插件，让你的 AI 助手能够访问微信读书数据：
- 获取书架、搜索书籍、查看阅读进度和时长
- 获取划线、笔记、热门书评、章节信息
- 随机抽取读书笔记（可用于晨报/日报）
- 批量导出所有笔记到本地 JSON
## 安装
```bash
# 方式 1：通过 clawhub（推荐）
clawhub install ChenyqThu/openclaw-weread-skill
# 方式 2：手动安装
git clone https://github.com/ChenyqThu/openclaw-weread-skill.git \
~/.openclaw/workspace/skills/weread
```
## 首次配置
需要微信读书的浏览器 Cookie，存储在 `~/.weread/cookie`。
获取方式：
1. **浏览器自动提取**：让 OpenClaw 打开 weread.qq.com，提取 `document.cookie` 写入文件
2. **手动粘贴**：`python3 scripts/weread_login.py paste`，从浏览器 DevTools 复制 Cookie
3. **直接写入**：将 Cookie 字符串写入 `~/.weread/cookie`
验证：`python3 scripts/weread_api.py verify`
## 数据隐私
- 所有数据仅存储在本地 `~/.weread/` 目录
- Cookie 文件权限为 600（仅所有者可读写）
- 不上传任何数据，不包含任何分析或遥测
FILE:scripts/export_notes.py
#!/usr/bin/env python3
"""微信读书笔记导出工具 — 结构化存储
存储结构:
~/.weread/
├── books.json # 书籍索引（bookId → 元信息）
├── notes/
│ ├── <bookId>.json # 每本书一个文件（笔记+划线+章节）
│ └── ...
└── notes_index.json # 全量笔记索引（按时间排序，方便检索）
用法:
export_notes.py # 全量导出
export_notes.py --incremental # 增量导出（仅新书/新笔记）
export_notes.py --stats # 显示统计信息
"""
import json
import os
import sys
import time
sys.path.insert(0, os.path.dirname(__file__))
from weread_api import (
get_bookshelf, get_notebooks, get_book_info,
get_reviews, get_bookmarks, get_chapter_infos,
get_reading_progress
)
BASE_DIR = os.path.expanduser("~/.weread")
BOOKS_PATH = os.path.join(BASE_DIR, "books.json")
NOTES_DIR = os.path.join(BASE_DIR, "notes")
INDEX_PATH = os.path.join(BASE_DIR, "notes_index.json")
def _fmt_time(ts: int) -> str:
if not ts:
return ""
return time.strftime("%Y-%m-%d %H:%M", time.localtime(ts))
def _fmt_duration(seconds: int) -> str:
if not seconds:
return "0分钟"
h, m = divmod(seconds // 60, 60)
if h:
return f"{h}小时{m}分钟" if m else f"{h}小时"
return f"{m}分钟"
def load_existing_books() -> dict:
if os.path.exists(BOOKS_PATH):
with open(BOOKS_PATH) as f:
return json.load(f)
return {}
def export_full():
"""全量导出"""
os.makedirs(NOTES_DIR, exist_ok=True)
log = lambda s: print(f"[{time.strftime('%H:%M:%S')}] {s}", file=sys.stderr)
# 1. 拉取书架
log("拉取书架...")
shelf = get_bookshelf()
shelf_books = shelf.get("books", [])
progress_list = shelf.get("bookProgress", [])
progress_map = {p["bookId"]: p for p in progress_list}
# 2. 拉取有笔记的书
log("拉取笔记列表...")
notebooks = get_notebooks()
log(f"共 {len(notebooks)} 本有笔记的书")
# 3. 构建书籍索引
books_index = {}
for b in shelf_books:
bid = b.get("bookId", "")
prog = progress_map.get(bid, {})
books_index[bid] = {
"bookId": bid,
"title": b.get("title", ""),
"author": b.get("author", ""),
"translator": b.get("translator", ""),
"cover": b.get("cover", ""),
"category": "", # 需要单独查询
"publisher": b.get("publisher", ""),
"publishTime": b.get("publishTime", ""),
"progress": prog.get("progress", 0),
"readingTime": prog.get("readingTime", 0),
"readingTimeFormatted": _fmt_duration(prog.get("readingTime", 0)),
"finishReading": b.get("finishReading", 0) == 1,
"lastReadTime": _fmt_time(prog.get("updateTime", 0)),
}
# 4. 逐本导出笔记
all_notes_flat = []
count = 0
for nb in notebooks:
book = nb.get("book", {})
bid = book.get("bookId", "")
title = book.get("title", "")
author = book.get("author", "")
note_count = nb.get("reviewCount", 0)
mark_count = nb.get("bookmarkCount", 0)
if note_count == 0 and mark_count == 0:
continue
count += 1
if count % 20 == 0:
log(f" 进度: {count}/{len(notebooks)}...")
# 获取笔记
reviews = []
if note_count > 0:
try:
raw = get_reviews(bid)
for r in raw:
ct = r.get("createTime", 0)
review_item = {
"type": "thought",
"content": r.get("content", ""),
"highlight": r.get("abstract", ""),
"chapterUid": r.get("chapterUid", 0),
"createTime": ct,
"createDate": _fmt_time(ct),
}
reviews.append(review_item)
# 加入扁平索引
all_notes_flat.append({
"bookId": bid,
"title": title,
"author": author,
**review_item,
})
except Exception as e:
log(f" ⚠ 获取 {title} 笔记失败: {e}")
# 划线和章节信息：跳过全量导出（太慢），按需获取
bookmarks = []
chapters = {}
# 写入单书文件
book_file = {
"bookId": bid,
"title": title,
"author": author,
"noteCount": len(reviews),
"highlightCount": len(bookmarks),
"chapters": chapters,
"notes": reviews,
"highlights": bookmarks,
"exportTime": time.strftime("%Y-%m-%dT%H:%M:%S"),
}
with open(os.path.join(NOTES_DIR, f"{bid}.json"), "w") as f:
json.dump(book_file, f, ensure_ascii=False, indent=2)
# 5. 保存书籍索引
with open(BOOKS_PATH, "w") as f:
json.dump(books_index, f, ensure_ascii=False, indent=2)
log(f"书籍索引: {len(books_index)} 本 → {BOOKS_PATH}")
# 6. 保存笔记索引（按时间倒序）
all_notes_flat.sort(key=lambda x: x.get("createTime", 0), reverse=True)
with open(INDEX_PATH, "w") as f:
json.dump({
"exportTime": time.strftime("%Y-%m-%dT%H:%M:%S"),
"totalNotes": len(all_notes_flat),
"notes": all_notes_flat,
}, f, ensure_ascii=False, indent=2)
log(f"笔记索引: {len(all_notes_flat)} 条 → {INDEX_PATH}")
log("✅ 全量导出完成")
print(f"书籍: {len(books_index)} | 有笔记: {count} | 笔记: {len(all_notes_flat)}")
def show_stats():
"""显示统计信息"""
if not os.path.exists(BOOKS_PATH):
print("❌ 未导出，请先运行 export_notes.py")
return
with open(BOOKS_PATH) as f:
books = json.load(f)
with open(INDEX_PATH) as f:
index = json.load(f)
total = len(books)
finished = sum(1 for b in books.values() if b.get("finishReading"))
reading = sum(1 for b in books.values() if b.get("progress", 0) > 0 and not b.get("finishReading"))
total_time = sum(b.get("readingTime", 0) for b in books.values())
notes = index["totalNotes"]
print(f"📚 书架: {total} 本 (读完 {finished} / 在读 {reading} / 未读 {total - finished - reading})")
print(f"⏱ 总阅读时长: {_fmt_duration(total_time)}")
print(f"✏️ 笔记: {notes} 条")
print(f"📁 存储: {NOTES_DIR}/ ({len(os.listdir(NOTES_DIR))} 个文件)")
print(f"🕐 上次导出: {index['exportTime']}")
# Notes per year
from collections import Counter
years = Counter()
for n in index["notes"]:
if n.get("createTime"):
y = time.strftime("%Y", time.localtime(n["createTime"]))
years[y] += 1
print("\n📊 笔记年度分布:")
for y in sorted(years):
bar = "█" * (years[y] // 10)
print(f" {y}: {years[y]:>4} 条 {bar}")
def main():
if "--stats" in sys.argv:
show_stats()
else:
export_full()
if __name__ == "__main__":
main()
FILE:scripts/random_review.py
#!/usr/bin/env python3
"""从读书笔记库中随机抽取笔记，用于晨报/日报回顾。
用法:
random_review.py [--count N] [--min-length L] [--format json|text]
输出:
随机抽取 N 条有质量的读书想法（默认 3 条），附带书名、作者、划线原文。
"""
import json
import os
import random
import sys
import time
INDEX_PATH = os.path.expanduser("~/.weread/notes_index.json")
LEGACY_PATH = os.path.expanduser("~/.weread/reading_notes.json")
def load_notes(min_length: int = 20) -> list:
"""加载所有有质量的笔记（过滤太短的）"""
# 优先使用新索引
path = INDEX_PATH if os.path.exists(INDEX_PATH) else LEGACY_PATH
if not os.path.exists(path):
print(f"❌ 笔记文件不存在", file=sys.stderr)
print("请先运行 export_notes.py 导出笔记。", file=sys.stderr)
sys.exit(1)
with open(path) as f:
data = json.load(f)
all_notes = []
# 新索引格式：扁平列表
if "notes" in data:
for n in data["notes"]:
content = n.get("content", "").strip()
if len(content) >= min_length:
all_notes.append({
"title": n.get("title", ""),
"author": n.get("author", ""),
"content": content,
"highlight": n.get("highlight", "").strip(),
"createTime": n.get("createTime", 0),
})
# 旧格式兼容
elif "books" in data:
for book in data["books"]:
for review in book.get("reviews", []):
content = review.get("content", "").strip()
if len(content) >= min_length:
all_notes.append({
"title": book.get("title", ""),
"author": book.get("author", ""),
"content": content,
"highlight": review.get("abstract", "").strip(),
"createTime": review.get("createTime", 0),
})
return all_notes
def pick_random(notes: list, count: int = 3) -> list:
"""加权随机抽取：内容更长、更有深度的笔记权重更高"""
if not notes:
return []
# Weight by content length (longer = more thoughtful)
weights = [min(len(n["content"]), 500) for n in notes]
count = min(count, len(notes))
selected = []
remaining = list(range(len(notes)))
remaining_weights = list(weights)
for _ in range(count):
if not remaining:
break
choices = random.choices(range(len(remaining)), weights=remaining_weights, k=1)
idx = choices[0]
selected.append(notes[remaining[idx]])
remaining.pop(idx)
remaining_weights.pop(idx)
return selected
def format_text(notes: list) -> str:
"""格式化为纯文本"""
parts = []
for i, n in enumerate(notes, 1):
date_str = ""
if n["createTime"]:
date_str = time.strftime("%Y-%m-%d", time.localtime(n["createTime"]))
part = f"📖 《{n['title']}》— {n['author']}"
if date_str:
part += f"（{date_str}）"
part += "\n"
if n["highlight"]:
# Truncate highlight if too long
hl = n["highlight"]
if len(hl) > 150:
hl = hl[:147] + "..."
part += f"📌 「{hl}」\n"
part += f"💭 {n['content']}"
parts.append(part)
return "\n\n".join(parts)
def main():
count = 3
min_length = 20
fmt = "text"
args = sys.argv[1:]
i = 0
while i < len(args):
if args[i] == "--count" and i + 1 < len(args):
count = int(args[i + 1])
i += 2
elif args[i] == "--min-length" and i + 1 < len(args):
min_length = int(args[i + 1])
i += 2
elif args[i] == "--format" and i + 1 < len(args):
fmt = args[i + 1]
i += 2
else:
i += 1
notes = load_notes(min_length)
selected = pick_random(notes, count)
if fmt == "json":
print(json.dumps(selected, ensure_ascii=False, indent=2))
else:
print(format_text(selected))
if __name__ == "__main__":
main()
FILE:scripts/weread_api.py
#!/usr/bin/env python3
"""微信读书 Web API 封装 — OpenClaw Skill 专用
Cookie 存储路径: ~/.weread/cookie (纯文本单行)
所有接口返回 JSON，错误时 exit(1) + stderr 输出。
"""
import json
import os
import sys
import time
import urllib.request
import urllib.error
import urllib.parse
COOKIE_PATH = os.path.expanduser("~/.weread/cookie")
BASE = "https://weread.qq.com"
HEADERS_COMMON = {
"User-Agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) "
"AppleWebKit/537.36 (KHTML, like Gecko) "
"Chrome/135.0.0.0 Safari/537.36",
"Accept": "application/json, text/plain, */*",
"Accept-Language": "zh-CN,zh;q=0.9,en;q=0.8",
}
# ── helpers ──────────────────────────────────────────────────────────
def _die(msg: str):
print(msg, file=sys.stderr)
sys.exit(1)
def _load_cookie() -> str:
if not os.path.exists(COOKIE_PATH):
_die(f"Cookie 文件不存在: {COOKIE_PATH}\n请先运行 weread_login.py 获取 Cookie。")
with open(COOKIE_PATH) as f:
cookie = f.read().strip()
if not cookie:
_die("Cookie 文件为空，请重新登录。")
return cookie
def _request(url: str, method: str = "GET", data: bytes | None = None,
extra_headers: dict | None = None, cookie: str | None = None) -> dict:
"""统一请求，自动附 Cookie + 时间戳防缓存"""
if cookie is None:
cookie = _load_cookie()
# GET 加时间戳
if method == "GET":
sep = "&" if "?" in url else "?"
url += f"{sep}_={int(time.time() * 1000)}"
headers = {**HEADERS_COMMON, "Cookie": cookie}
if extra_headers:
headers.update(extra_headers)
if data is not None:
headers["Content-Type"] = "application/json;charset=UTF-8"
req = urllib.request.Request(url, data=data, headers=headers, method=method)
try:
with urllib.request.urlopen(req, timeout=30) as resp:
body = resp.read()
except urllib.error.HTTPError as e:
body = e.read()
try:
err = json.loads(body)
except Exception:
_die(f"HTTP {e.code}: {body[:500]}")
if err.get("errcode") in (-2012, -2010):
_die("Cookie 已过期，请重新登录微信读书。")
_die(f"HTTP {e.code}: {json.dumps(err, ensure_ascii=False)}")
except Exception as e:
_die(f"请求失败: {e}")
result = json.loads(body)
ec = result.get("errcode") or result.get("errCode")
if ec and ec != 0:
if ec in (-2012, -2010):
_die("Cookie 已过期，请重新登录微信读书。")
_die(f"API 错误 ({ec}): {result.get('errmsg') or result.get('errMsg') or ''}")
return result
# ── public API ───────────────────────────────────────────────────────
def get_bookshelf() -> dict:
"""获取完整书架（含进度、书单分组等）"""
return _request(f"{BASE}/web/shelf/sync")
def get_notebooks() -> list:
"""获取有笔记的书籍列表"""
data = _request(f"{BASE}/api/user/notebook")
return data.get("books", [])
def get_book_info(book_id: str) -> dict:
"""获取书籍详情"""
return _request(f"{BASE}/api/book/info?bookId={urllib.parse.quote(book_id)}")
def get_reading_progress(book_id: str) -> dict:
"""获取阅读进度（当前版本接口）"""
return _request(f"{BASE}/web/book/getProgress?bookId={urllib.parse.quote(book_id)}")
def get_read_detail(book_id: str) -> dict:
"""获取详细阅读信息（含阅读时长明细、完成日期等）"""
return _request(
f"{BASE}/web/book/readinfo?bookId={urllib.parse.quote(book_id)}"
"&readingDetail=1&readingBookIndex=1&finishedDate=1"
)
def get_bookmarks(book_id: str) -> list:
"""获取书籍的划线记录"""
data = _request(f"{BASE}/web/book/bookmarklist?bookId={urllib.parse.quote(book_id)}")
return data.get("updated", [])
def get_reviews(book_id: str, mine_only: bool = True) -> list:
"""获取书籍的笔记/想法"""
lt = 4 if not mine_only else 11
mine = 1 if mine_only else 0
data = _request(
f"{BASE}/web/review/list?bookId={urllib.parse.quote(book_id)}"
f"&listType={lt}&mine={mine}&synckey=0&count=0&listMode=2&maxIdx=0"
)
reviews = data.get("reviews", [])
return [r.get("review", r) for r in reviews]
def get_best_reviews(book_id: str, count: int = 10) -> dict:
"""获取热门书评"""
return _request(
f"{BASE}/web/review/list/best?bookId={urllib.parse.quote(book_id)}"
f"&count={count}&maxIdx=0&synckey=0"
)
def get_chapter_infos(book_id: str) -> dict:
"""获取章节信息（POST）"""
payload = json.dumps({"bookIds": [book_id]}).encode()
headers = {
"Origin": "https://weread.qq.com",
"Referer": f"https://weread.qq.com/web/reader/{book_id}",
}
data = _request(f"{BASE}/web/book/chapterInfos", method="POST",
data=payload, extra_headers=headers)
# 多种返回格式兼容
if "data" in data and isinstance(data["data"], list) and data["data"]:
return data["data"][0].get("updated", [])
if "updated" in data:
return data["updated"]
return data
def verify_cookie(cookie: str | None = None) -> bool:
"""验证 Cookie 是否有效"""
try:
result = _request(f"{BASE}/api/user/notebook", cookie=cookie)
return "books" in result
except SystemExit:
return False
# ── CLI 入口 ─────────────────────────────────────────────────────────
def _fmt(seconds: int) -> str:
h, m = divmod(seconds // 60, 60)
if h:
return f"{h}小时{m}分钟" if m else f"{h}小时"
return f"{m}分钟"
def _print_json(obj):
print(json.dumps(obj, ensure_ascii=False, indent=2))
def main():
if len(sys.argv) < 2:
print("用法: weread_api.py <command> [args...]")
print("命令:")
print(" shelf - 获取书架")
print(" notebooks - 获取有笔记的书籍")
print(" info <bookId> - 获取书籍详情")
print(" progress <bookId> - 获取阅读进度")
print(" detail <bookId> - 获取详细阅读信息")
print(" bookmarks <bookId> - 获取划线记录")
print(" reviews <bookId> - 获取我的笔记/想法")
print(" best-reviews <bookId> - 获取热门书评")
print(" chapters <bookId> - 获取章节信息")
print(" search <keyword> - 搜索书架中的书籍")
print(" verify - 验证 Cookie 是否有效")
sys.exit(0)
cmd = sys.argv[1]
if cmd == "shelf":
data = get_bookshelf()
books = data.get("books", [])
progress_list = data.get("bookProgress", [])
progress_map = {p["bookId"]: p for p in progress_list}
summary = []
for b in books:
bid = b.get("bookId", "")
prog = progress_map.get(bid, {})
summary.append({
"bookId": bid,
"title": b.get("title", ""),
"author": b.get("author", ""),
"progress": prog.get("progress", 0),
"readingTime": _fmt(prog.get("readingTime", 0)),
"finishReading": b.get("finishReading", 0) == 1,
})
_print_json({"total": len(summary), "books": summary})
elif cmd == "notebooks":
_print_json(get_notebooks())
elif cmd == "info":
if len(sys.argv) < 3:
_die("用法: weread_api.py info <bookId>")
_print_json(get_book_info(sys.argv[2]))
elif cmd == "progress":
if len(sys.argv) < 3:
_die("用法: weread_api.py progress <bookId>")
_print_json(get_reading_progress(sys.argv[2]))
elif cmd == "detail":
if len(sys.argv) < 3:
_die("用法: weread_api.py detail <bookId>")
_print_json(get_read_detail(sys.argv[2]))
elif cmd == "bookmarks":
if len(sys.argv) < 3:
_die("用法: weread_api.py bookmarks <bookId>")
marks = get_bookmarks(sys.argv[2])
_print_json({"total": len(marks), "bookmarks": marks})
elif cmd == "reviews":
if len(sys.argv) < 3:
_die("用法: weread_api.py reviews <bookId>")
reviews = get_reviews(sys.argv[2])
_print_json({"total": len(reviews), "reviews": reviews})
elif cmd == "best-reviews":
if len(sys.argv) < 3:
_die("用法: weread_api.py best-reviews <bookId>")
count = int(sys.argv[3]) if len(sys.argv) > 3 else 10
_print_json(get_best_reviews(sys.argv[2], count))
elif cmd == "chapters":
if len(sys.argv) < 3:
_die("用法: weread_api.py chapters <bookId>")
_print_json(get_chapter_infos(sys.argv[2]))
elif cmd == "search":
if len(sys.argv) < 3:
_die("用法: weread_api.py search <keyword>")
keyword = sys.argv[2].lower()
data = get_bookshelf()
books = data.get("books", [])
progress_list = data.get("bookProgress", [])
progress_map = {p["bookId"]: p for p in progress_list}
matched = []
for b in books:
title = (b.get("title") or "").lower()
author = (b.get("author") or "").lower()
if keyword in title or keyword in author:
bid = b.get("bookId", "")
prog = progress_map.get(bid, {})
matched.append({
"bookId": bid,
"title": b.get("title", ""),
"author": b.get("author", ""),
"progress": prog.get("progress", 0),
"readingTime": _fmt(prog.get("readingTime", 0)),
"finishReading": b.get("finishReading", 0) == 1,
})
_print_json({"total": len(matched), "keyword": sys.argv[2], "books": matched})
elif cmd == "verify":
ok = verify_cookie()
print("✅ Cookie 有效" if ok else "❌ Cookie 已过期或无效")
sys.exit(0 if ok else 1)
else:
_die(f"未知命令: {cmd}")
if __name__ == "__main__":
main()
FILE:scripts/weread_login.py
#!/usr/bin/env python3
"""微信读书 Cookie 登录工具
方式 1（推荐）：从 Chrome 浏览器提取已登录的 Cookie
weread_login.py chrome [--profile <profile_path>]
方式 2：手动粘贴 Cookie
weread_login.py paste
Cookie 存储位置: ~/.weread/cookie
"""
import json
import os
import sys
import sqlite3
import shutil
import tempfile
COOKIE_DIR = os.path.expanduser("~/.weread")
COOKIE_PATH = os.path.join(COOKIE_DIR, "cookie")
DOMAIN = ".weread.qq.com"
def save_cookie(cookie: str):
os.makedirs(COOKIE_DIR, exist_ok=True)
with open(COOKIE_PATH, "w") as f:
f.write(cookie.strip())
os.chmod(COOKIE_PATH, 0o600)
print(f"✅ Cookie 已保存到 {COOKIE_PATH}")
def extract_chrome_cookies(profile_path: str | None = None) -> str:
"""从 Chrome Cookies SQLite 提取 weread.qq.com 的 cookie。
注意: macOS 上 Chrome 的 cookie 值是加密的（AES-128-CBC with Keychain），
这里只能提取未加密的字段。如果值为空则需要用浏览器方式获取。
"""
if profile_path is None:
# macOS default Chrome profile
profile_path = os.path.expanduser(
"~/Library/Application Support/Google/Chrome/Default"
)
cookies_db = os.path.join(profile_path, "Cookies")
if not os.path.exists(cookies_db):
print(f"❌ Chrome Cookies 数据库不存在: {cookies_db}", file=sys.stderr)
print("请确认 Chrome 已安装且路径正确。", file=sys.stderr)
sys.exit(1)
# 复制到临时文件（Chrome 可能锁定原文件）
tmp = tempfile.mktemp(suffix=".db")
shutil.copy2(cookies_db, tmp)
try:
conn = sqlite3.connect(tmp)
cursor = conn.cursor()
cursor.execute(
"SELECT name, value, encrypted_value FROM cookies "
"WHERE host_key LIKE '%weread.qq.com%'"
)
rows = cursor.fetchall()
conn.close()
finally:
os.unlink(tmp)
if not rows:
print("❌ 未在 Chrome 中找到 weread.qq.com 的 Cookie。", file=sys.stderr)
print("请先在 Chrome 中登录 https://weread.qq.com", file=sys.stderr)
sys.exit(1)
# macOS Chrome 的 cookie 值通常是加密的
# 检查是否有未加密的值可用
parts = []
encrypted_count = 0
for name, value, encrypted_value in rows:
if value:
parts.append(f"{name}={value}")
elif encrypted_value:
encrypted_count += 1
if encrypted_count > 0 and not parts:
print("⚠️ Chrome Cookie 值已加密（macOS Keychain 保护）。", file=sys.stderr)
print("请使用以下方式之一获取 Cookie:", file=sys.stderr)
print(" 1. 运行 weread_login.py paste — 手动从浏览器 DevTools 粘贴", file=sys.stderr)
print(" 2. 运行 weread_login.py browser — 通过 OpenClaw 浏览器自动提取", file=sys.stderr)
sys.exit(1)
cookie_str = "; ".join(parts)
return cookie_str
def paste_cookie():
"""手动粘贴 Cookie"""
print("请从浏览器 DevTools (F12 → Network → 任意请求 → Headers → Cookie) 复制 Cookie：")
print("（粘贴后按 Enter）")
cookie = input().strip()
if not cookie:
print("❌ Cookie 不能为空", file=sys.stderr)
sys.exit(1)
return cookie
def main():
if len(sys.argv) < 2:
print("用法:")
print(" weread_login.py chrome [--profile <path>] 从 Chrome 提取 Cookie")
print(" weread_login.py paste 手动粘贴 Cookie")
print(" weread_login.py browser 通过 OpenClaw 浏览器提取")
print(f"\nCookie 存储: {COOKIE_PATH}")
sys.exit(0)
cmd = sys.argv[1]
if cmd == "chrome":
profile = None
if "--profile" in sys.argv:
idx = sys.argv.index("--profile")
if idx + 1 < len(sys.argv):
profile = sys.argv[idx + 1]
cookie = extract_chrome_cookies(profile)
save_cookie(cookie)
# 验证
sys.path.insert(0, os.path.dirname(__file__))
from weread_api import verify_cookie
if verify_cookie(cookie):
print("✅ Cookie 验证通过")
else:
print("⚠️ Cookie 已保存但验证未通过，可能需要重新登录")
elif cmd == "paste":
cookie = paste_cookie()
save_cookie(cookie)
sys.path.insert(0, os.path.dirname(__file__))
from weread_api import verify_cookie
if verify_cookie(cookie):
print("✅ Cookie 验证通过")
else:
print("⚠️ Cookie 已保存但验证未通过，请检查 Cookie 是否完整")
elif cmd == "browser":
# 此模式由 SKILL.md 中 Jarvis 的浏览器工具流程处理
print("browser 模式需要通过 OpenClaw 浏览器工具执行。")
print("请让 Jarvis 执行「微信读书登录」流程。")
sys.exit(0)
else:
print(f"未知命令: {cmd}", file=sys.stderr)
sys.exit(1)
if __name__ == "__main__":
main()