Tongyi Wanxiang 2.7 video model 鈥?one model covers text-to-video, first/last-frame-to-video, and reference-to-video.
