多模态参考生视频 AI
通过参考图片、视频与音频进行多模态视频生成，智能提炼角色动作、外貌特征与镜头运动
近期创作
灵感画廊
探索 AI 创作示例，在 AI Kenerate 开启下一份作品。
MiniMax-H3
Multi-reference Cinematic Motion
“Character consistent camera tracking shot through misty cyberpunk alleyway”
Seedance 2.5
Dynamic Subject Reference Transition
“Seamless dynamic lighting transformation respecting subject identity and atmosphere”
常见问题
Reference to Video allows you to supply multiple reference images (up to 9) and videos (up to 3) as conditioning inputs to guide character consistency, camera motion, and action style.
Currently, MiniMax-H3 and Seedance 2.5 support multimodal reference-based video generation. Unsupported models are automatically hidden on this page.