Published: October 19, 2026
Video Model
OmniHuman 2.0 is ByteDance’s audio-driven avatar model: one character image plus a voice or music track becomes a talking or singing video, with lip sync, expression and gestures that follow the audio.
Compared with OmniHuman 1.5 it runs much longer (up to 5 minutes in Artlist, vs 30 seconds) and takes an optional prompt for gestures, camera moves and scene details. HeyGen Avatar 4 stops at 3 minutes and has no prompt in Artlist, so OmniHuman 2.0 is the pick when the performance or camera needs directing.
Key Features
Avatar Generation
- Image + audio to video: Animates one character image to speak or sing the uploaded audio. The clip is exactly as long as the audio.
- Audio-driven performance: Mouth, face and gestures follow the rhythm and emphasis of the audio, for dialogue, voiceover or songs.
Direction
- Optional text prompt: Steer expressions, gestures, camera movement or scene details. Leave it empty and the audio drives everything.
Technical Capabilities
Details |
|
|---|---|
| Quality / Resolution | 720p, shaped to the uploaded image (a square image gives 960×960); no resolution setting |
| Duration | 4 s – 5 min, set by the length of the uploaded audio |
| Aspect Ratios | No setting; framing comes from the uploaded image |
| Inputs | 1 image (JPG, PNG, WebP) + 1 audio file (WAV, MP3), optional text prompt |
| Prompt Languages | Chinese, English, Japanese, Korean, Spanish (Mexico), Indonesian |
Limitations
- 5 minutes is the ceiling. Longer scripts or full songs have to be split into separate generations.
- One image and one audio track per generation: no reference images, start or end frames, or video input.
- No resolution, aspect ratio or duration controls, and audio shorter than 4 seconds won’t go through.
Prompting Tips
- For a straight talking head, leave the prompt empty and let the audio carry the performance.
- Use the prompt for what the audio can’t say: “slow push-in”, “turns to camera on the chorus”, “leans in and gestures with both hands”.
- Start from a clear image with the face fully visible. Clean audio, without long silences or background noise, gives cleaner lip sync.
- For songs, upload the vocal mix you want the character to perform, not the instrumental.