01 —
Let the feeling lead.
OmniHuman 2.0 draws on audio and visual context to shape facial expression, gestures, and body movement. A spoken line or sung phrase becomes a more expressive performance.
See the performanceAudio-driven. Character-led.
OmniHuman 2.0 Make every word move.
One image sets the scene. Your audio gives it a voice. OmniHuman 2.0 brings the two together with expressive faces, gestures, and movement.
Discover the model. Web generation is being prepared.
Official model demo
Audio into expression.
Demonstration by BytePlusImage → performance
A new way to bring characters to life
ConversationsVocal performancesCharacter stories
The model, at a glance
What is OmniHuman 2.0?
OmniHuman 2.0 is an audio-driven human animation model from ByteDance. It turns a reference image and audio into a video, with optional text direction for the performance.
- The image
A visual starting point.
A reference image establishes the character’s appearance, pose, and framing.
- The audio
A voice with intention.
Speech or singing supplies the vocal performance and rhythm for the animation.
- The video
A character in motion.
The output combines facial expression, gesture, and body movement into a video performance.
01 / See the difference
A still image. A living moment.
Watch how a voice becomes expression, presence, and movement. Real examples from the model’s official showcase.
01 / 03
Everyday expression
A voice carries more than words. See the pauses, glances, and gestures that bring a spoken performance into the frame.
Better with sound
Reference image + spoken audio. A closer look at expression and lip synchronization.
Official BytePlus demonstrations. These examples were not generated on this website.
View the official showcase02 /The performance
Every voice deserves a performance.
Explore the expressions, gestures, and movement that give an audio-driven character a sense of presence.
02 —
Give the story room to unfold.
Work with audio from 4 seconds to 5 minutes. An optional text prompt can guide the action, scene, or camera while your reference image establishes the character.
Explore the examples- 4 sec–5 min audio
- 720p video output
- Optional text direction
Model specifications: BytePlus overviewTechnical documentation
03 /How it works
From a still to a scene.
One image and an audio track are the starting point. Add text direction to shape the action, scene, or camera.
One image
Your audio
A performance
- 01
Choose your character.
Start with a clear image you have permission to use. Consider the framing, pose, and personality you want on screen.
- 02
Set the tone.
Choose spoken or sung audio. Add an optional prompt for the action, scene, or camera direction.
- 03
Bring the scene to life.
The model combines the inputs into a video performance. Explore the examples to see expression and movement in context.
04 /Creative possibilities
One character. So many stories.
A new way to think about the people, voices, and characters in your next creative project.
Speak & explain01
Make an idea feel personal.
Explore character-led explanations, educational stories, and dialogue that give a voice a visible presence.
Sing & perform02
Put a face to the music.
Imagine vocal performances where the character’s expression and movement follow the feeling of a song.
Imagine & tell03
Let your characters speak.
Develop narrative concepts, fictional scenes, and expressive characters from a visual starting point.
05 /A little more context
Good questions. Clear answers.
Get to know the model, the terminology, and the possibilities before you go further.
What is OmniHuman 2.0?
OmniHuman 2.0 is an audio-driven human animation model developed by ByteDance. It creates video performances from a reference image and audio, combining facial expression, gestures, and body movement. An optional text prompt can provide further direction.
Is Omni Human 2.0 the same as OmniHuman 2?
People search for this model using several spellings, including Omni Human 2.0, OmniHuman 2, Omni Human 2, and Omni-Human-2.0. The model discussed on this page is OmniHuman 2.0.
What inputs does OmniHuman 2.0 need?
The model takes one reference image and an audio track. You can also supply an optional text prompt to guide the action, scene, or camera. BytePlus documents audio lengths from 4 seconds to 300 seconds and 720p video output.
Can OmniHuman 2.0 animate singing as well as speech?
Yes. Official examples include both spoken and sung audio. The model uses audio and visual context to produce expressions and movement appropriate to the performance.
Can I generate a video on this page?
Video generation is not available on this website yet. You can watch the official model examples, learn the workflow, and visit Pricing for current access information.
Is this the official ByteDance or BytePlus website?
No. This is an independent website about OmniHuman 2.0 and is not affiliated with or endorsed by ByteDance or BytePlus. Model names belong to their respective owners; official demonstrations are credited to their source.
What should I consider before animating a person?
Use images and audio you have the rights and permission to use. Get consent before animating a real person, disclose AI-generated content where appropriate, and avoid misleading impersonation.
See what moves you
Your next idea,
with a human presence.
A conversation, a lesson, a song. Discover the creative possibilities of OmniHuman 2.0.