Please note the compute consumption.
Parameters: Complex actions perform better when character orientation matches the video; it responds to camera movement when matching the image.
Default is character orientation matching the video, where character actions, expressions, camera movement, and orientation are generated according to the motion video. Other information can be controlled via prompt. Supports up to 30s generation length.
Selecting character orientation matching the image, character actions and expressions will follow the motion video, while the orientation matches the person in the image. Camera movement and other info can be customized via prompt.
Supports up to 10s generation length.