AI video model · by ByteDance
OmniHuman talking video generator
OmniHuman is ByteDance's audio-driven human animation model: you give it an image of a person and an audio track, and it generates a video of that person speaking or singing in time with the audio.
OmniHuman variants available now
These are the OmniHuman variants currently enabled on MotionLab and the settings each one accepts. The table is generated from the live model catalog, so it changes when a variant is added, updated or paused.
| Variant | Modes | Duration | Resolution | Aspect ratios | Audio / end frame |
|---|---|---|---|---|---|
| OmniHuman v2 | Image to video | Follows your source media | auto | auto | Silent · No end frame |
What to use OmniHuman for
- Talking-head explainers and announcements from a single portrait.
- Lip-synced music snippets and character performances.
- Localised versions of a message by pairing one image with different voice tracks.
- Presenter clips for product demos and course content.
How to use OmniHuman on MotionLab
- Sign in and open the video creation page.
- Choose a OmniHuman variant in the model picker. Only the variants listed above are shown.
- Write your prompt, add a start image or references if the mode needs them, and pick duration, resolution and aspect ratio.
- Check the credit cost shown for your settings, then generate. Failed jobs are refunded automatically.
- Download the result or keep refining it in your library.
OmniHuman prompt tips
- Use a clear, front-facing portrait with the mouth visible and good light.
- Use clean audio — background noise and music under speech reduce lip-sync quality.
- Trim silence at the start and end of the audio; the video length follows the audio.
- Only animate people who have agreed to it, and follow the acceptable use policy.
More technique in the MotionLab prompt guide and the camera motion guide.
Compare OmniHuman with other models
OmniHuman questions
How long can an OmniHuman video be?
The video length follows your audio track within the limits shown in the specs table on this page.
Can I animate any photo of a person?
Only images you have the rights and the person's consent to use. MotionLab's safety checks and acceptable use policy apply to every generation.
How much does OmniHuman cost on MotionLab?
OmniHuman generations spend MotionLab credits from the same wallet used by every model on the site. The credit cost for your exact settings is shown before you generate, and a job that fails is refunded automatically. Credit packs are listed on the pricing page.
Do I need a separate OmniHuman account or API key?
No. MotionLab runs OmniHuman for you: sign in, pick the model in the create page and generate. Your results are saved to your MotionLab library next to work from every other model.