Upload one portrait and a voice clip. OmniHuman 2.0 animates the face and body to the audio and returns a 720P video as long as the clip, from 4 to 300 seconds.
Motion inspiration
These clips come from the shared AI video gallery and are visual inspiration only. They are not OmniHuman 2.0 outputs; use them to plan framing and pacing.
Framing the portrait
Input: 1 portrait, JPEG/PNG/WebP
A head-and-shoulders portrait with the face clearly visible gives the model the most to work with.
Input: 1 portrait, JPEG/PNG/WebP
A clear recording without music or background noise keeps the mouth movement readable.
Input: WAV or MP3, 4-300 sec
Test with a short clip first; the price grows with every second of audio.
Pricing: 14 credits/sec, 4-sec minimum
Why OmniHuman 2.0
AISnapEdit exposes a focused lip sync surface: one image, one audio file, and an optional prompt. Output is 720P and as long as the audio.
Provider
ByteDance
ByteDance's OmniHuman model line for audio-driven human video.
Inputs
Image + audio
One portrait and one WAV or MP3 voice clip.
Length
4-300 sec
The video follows the audio length.
Pricing
14 credits/sec
Per started second of audio, 4-second minimum.
Pick the tool by the input you already have.
Portrait + audio required
Portrait + audio required
Visual inspiration, not OmniHuman 2.0 output
Upload one portrait and one WAV or MP3 clip of 4-300 seconds.
Prompt-driven motion
Prompt-driven motion
Visual inspiration, not OmniHuman 2.0 output
Use the Image to Video tool and describe the motion in the prompt.
Prepare the audio first; it decides both the result and the price.
Own voice
Best when you speak the script yourself.
Generated voice
Best when you start from a written script.
These are planning examples, not verified model outputs.
Product walkthrough
Visual inspiration, not OmniHuman 2.0 output
A founder wants a short spoken intro for a landing page.
Prompt cue
Friendly, confident delivery, small hand gestures, steady head-and-shoulders framing.
Settings
Front-facing portrait, 20-30 sec MP3
Expected motion
Natural speech rhythm with restrained gestures
Course intro
Visual inspiration, not OmniHuman 2.0 output
An instructor needs a welcome clip for a new lesson.
Prompt cue
Warm, calm expression, occasional nod, eye contact with the camera.
Settings
Well-lit portrait, 45-60 sec MP3
Expected motion
Even pacing over a longer take
Character line
Visual inspiration, not OmniHuman 2.0 output
A game or story character delivers one recorded line.
Prompt cue
Expressive face, slight head tilt, keep the illustrated style.
Settings
Illustrated portrait, 5-10 sec WAV
Expected motion
Lip movement matched to a short line
The AISnapEdit generator for this model focuses on one lip sync workflow.
Inputs
One portrait image and one WAV or MP3 audio file are required; the prompt is optional.
Prepare a clean voice track before you start.
Length
The video is as long as the audio, from 4 to 300 seconds.
Trim silence at the start and end to avoid paying for it.
Pricing
14 credits per started second of audio, 4-second minimum, measured from the uploaded file.
Test with a short clip before a long take.
Media proof
Gallery media is inspiration and is not OmniHuman 2.0 output.
Judge the model on your own short test clip.
Find answers to common questions about this model
OmniHuman 2.0 is a ByteDance model that turns one portrait and one voice recording into a talking video. On AISnapEdit it powers the Lip Sync tool.
One JPEG, PNG, or WebP image up to 30 MB and one WAV or MP3 audio file of 4-300 seconds up to 15 MB.
The video is as long as the audio, from 4 to 300 seconds.
720P MP4. There is no resolution, duration, or aspect ratio setting.
14 credits per started second of audio, with a 4-second minimum: 4 seconds costs 56 credits, 60 seconds costs 840 credits. The server measures the uploaded audio and charges by that length.
No. The audio drives the video. An optional prompt of up to 300 characters can guide expression, gestures, or framing.
Not yet. Create the voice first, for example with the AI Voice Generator, download it, and upload the file.
Only ones you have the right to use, such as your own or with the person's permission. Do not use someone's likeness or voice to mislead others; see the Terms of Service.
Upload a portrait and a voice clip, check the estimate, and create a 720P talking video.
