Skip to content

OmniHuman-1

Generate human videos from one portrait and motion signals

Visit official site

What OmniHuman-1 is

The framework is described as an end-to-end multimodal conditional human video generation framework that creates human videos from a single portrait and motion signals such as audio, video or both, using a mixed training strategy to address the scarcity of high-quality data.

What you can do with it

  • Generate a video from a single portrait plus an audio track
  • Feed portrait images of different aspect ratios, including headshots, half-body shots and full-body shots
  • Drive generation with audio, video or both signals together
  • Produce videos with realistic motion, lighting and texture detail

Application scenarios

  • Generate a smooth speech video for a virtual host
  • Create a singer performance video for a music clip across music styles
  • Produce realistic motion and expression videos for animated characters

Getting started

  • Visit the project page for the framework
  • Prepare a high-quality portrait image as the input
  • Choose a motion signal such as an audio file or a video file
  • Upload the portrait image and the motion signal to the model

Pros & cons

✓ What we like

  • Motion signals can be audio, video or a combination of both
  • Music styles, singing forms and hand gestures are listed as supported

! What to watch out for

  • Each run requires a portrait image and a motion signal as inputs
  • Training relies on a mixed strategy to address scarce high-quality data

FAQ

What inputs does it need?

The material names a single high-quality portrait image and a motion signal such as an audio or video file.

Which motion signals are supported?

The feature list names audio, video, or both combined.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review