← Back to feed

Kandinsky S2V 1.0 – an open family of speech-to-video models

AI · Technology · 2026-11-03 09:31

该系列含2B的Lite与21B的Pro两个版本，输入单张参考图、语音和文本提示，即可生成25FPS长视频，唇形、表情、姿态与手势随语音同步。

Source: kandinskyavatar.github.io

Auto-generated summary for reference only. RSS