ByteDanceOmniHuman 2 is ByteDance's audio-driven human video model. It turns one image of a person and an audio track of 4 to 300 seconds into a video whose expressions, gestures, and movement follow the audio, with an optional prompt to direct the scene.
No usage yet