OmniHuman 2.0 Talking Avatar
Upload one portrait and 4–300 seconds of MP3 or WAV audio. The video follows the audio length at 720p. Add an optional prompt of up to 300 characters to guide gestures and camera.
Open workspaceOmniHuman 2.0 AI Talking Avatar Generator: Lifelike Videos up to 5 Minutes
OmniHuman 2.0 is an audio-driven AI video model that turns one image and one audio file into a lifelike talking avatar video, with natural lip sync, expressive faces, body gestures and videos up to 5 minutes long.
Image + audio to video · 720p · 4–300 seconds · optional prompt
OmniHuman 2.0 Image and Audio to Video
One workflow: upload a portrait, add speech or singing, and optionally describe the performance.
Explore OmniHuman 2.0 Features
OmniHuman 2.0 turns one image and your audio into a talking video with natural lip sync, expressive faces and body gestures.
Turn One Image and Audio into a Talking Video with OmniHuman 2.0
Turn a single image and an audio file into a realistic talking avatar video with OmniHuman 2.0. The model brings characters to life with natural speech, facial expressions and body movements, all driven by the audio. Create AI presenters, virtual characters and digital avatars without filming anyone. Upload a portrait photo, an illustration or a stylized character, and OmniHuman 2.0 keeps the face, outfit and style true to your image.
- One portrait plus one audio file is all you need
- Speech, expressions and gestures follow the audio
- AI presenters, virtual hosts and digital humans
Natural Lip Sync for Speech and Singing
OmniHuman 2.0 creates talking videos with lip movements that closely follow spoken words and singing. Mouth shapes stay in sync with the audio while the rest of the face keeps moving naturally. Use it for virtual presenters, singing avatars and voice-driven character animation. Lip sync holds up in close-ups and in fast lyrics, so talking-head videos and music performances look convincing on any screen.
- Mouth shapes match every word and lyric
- Works for talking and for singing
- Speech in Chinese, English, Japanese, Korean, Spanish and more
Expressive Faces and Body Gestures
OmniHuman 2.0 goes beyond basic lip sync. It generates facial expressions, head turns, hand gestures and body motion that respond to the rhythm and emotion of the audio. Give AI avatars more personality for storytelling, entertainment and social content. A calm narration stays gentle, while an excited voice brings bigger gestures and livelier expressions.
- Emotion in the voice shows on the face
- Head turns, hand gestures and posture shifts
- Performances that feel alive, not stiff
Create Talking Videos Up to 5 Minutes Long
OmniHuman 2.0 supports talking videos up to 300 seconds. The video length follows your audio, from 4 seconds to 5 minutes. Produce full speeches, presentations, tutorials and songs in one take instead of stitching short clips together. One continuous take keeps the same face, lighting and framing from the first word to the last.
- Audio from 4 to 300 seconds
- Video length matches your audio exactly
- 720p output for long-form content
Animate Several People in One Scene
OmniHuman 2.0 can animate several characters in a single scene. Create conversations, interviews and group performances with audio-driven expressions and natural body movement. Turn podcast audio and dialogues into engaging videos. Use a group photo or a scene with several people, and give each speaker room in the frame for the clearest results.
- Conversations, interviews and group scenes
- Each person reacts with natural motion
- Ideal for visual podcasts
Guide the Performance with a Text Prompt
Add an optional text prompt to your image and audio in OmniHuman 2.0. Describe actions, gestures, scene behavior and camera movement to shape the result. Combine audio-driven animation with prompt guidance for more controlled talking videos. For example, write “smiles and waves at the camera, slow push-in” to open a product video with a friendly greeting.
- Optional prompt of up to 300 characters
- Direct gestures, scene and camera
- Works with photos, illustrations and stylized characters
How to Create Talking Videos with OmniHuman 2.0
Make a lifelike talking video in OmniHuman 2.0 in three simple steps:
Upload your portrait
Sign in and open the OmniHuman 2.0 workspace. Upload one clear JPG, PNG or WEBP image of up to 30 MB, 300–6,000 px per side, with an aspect ratio between 0.4 and 2.5. Photos, illustrations and stylized characters all work.
Add audio and an optional prompt
Upload speech or singing as MP3 or WAV, up to 15 MB and 4 to 300 seconds long. The audio sets the video length. If you like, describe gestures, scene or camera moves in up to 300 characters.
Generate and download
Check the credit quote and generate. Your OmniHuman 2.0 video appears in the conversation and My creations, ready to preview and download.
Model comparison
OmniHuman 2.0 vs OmniHuman 1.5: Longer Talking Videos
OmniHuman 2.0 extends audio-driven avatars to videos up to 5 minutes long, ideal for presentations, singing and digital humans. OmniHuman 1.5 makes shorter clips with up to 1080p output. Both support lip sync, expressive gestures and prompt-guided animation.
| Capability | OmniHuman 2.0Up to 5 min | OmniHuman 1.5 |
|---|---|---|
| Positioning | Long-form talking avatars | Short high-resolution clips |
| Input | Image + audio + optional prompt | Image + audio + optional prompt |
| Audio-driven animation | Supported | Supported |
| Lip sync | Supported | Supported |
| Facial expressions | Supported | Supported |
| Body gestures | Supported | Supported |
| Multi-person animation | Supported | Supported |
| Text prompt control | Supported | Supported |
| Maximum video length | Up to 300 seconds | Under 60 seconds |
| Video length | Follows the audio, 4–300 s | Follows the audio |
| Output resolution | 720p | 720p / 1080p |
| Best for | Talks, courses, songs and podcasts | Short high-resolution avatar clips |
OmniHuman 2.0 Frequently Asked Questions
What is OmniHuman 2.0?
OmniHuman 2.0 is ByteDance's audio-driven AI video model. It turns a character image and an audio file into a talking avatar video with synchronized lip movements, facial expressions and body gestures. It works with real photos, illustrations and stylized characters, and makes videos up to 5 minutes long in 720p.
How does OmniHuman 2.0 work?
OmniHuman 2.0 analyzes the speech, rhythm and emotion in your audio, then animates your image with matching lip movements, facial expressions and body gestures. An optional text prompt can guide actions and camera moves. Clear speech with little background noise gives the most accurate lip sync.
Which image and audio files does OmniHuman 2.0 accept?
Upload one JPG, PNG or WEBP image of up to 30 MB, 300–6,000 px per side, with an aspect ratio from 0.4 to 2.5. Audio must be MP3 or WAV, up to 15 MB and 4 to 300 seconds long. The upload box tells you the exact reason if a file doesn't fit.
How long can OmniHuman 2.0 videos be?
OmniHuman 2.0 makes videos from 4 to 300 seconds (5 minutes). The video always matches the length of your audio and is delivered in 720p.
How many credits does OmniHuman 2.0 use?
OmniHuman 2.0 uses 28 credits per second of audio. A 10-second video costs 280 credits, a 1-minute video 1,680 credits and a 5-minute video 8,400 credits. The workspace shows the exact quote after you add the audio. OmniHuman 2.0 requires VIP.
Does OmniHuman 2.0 support several people in one video?
Yes. OmniHuman 2.0 can animate several people in one scene, which works well for interviews, podcasts and group performances. Results depend on the image, the audio and how complex the scene is. For the clearest result, use an image where every face is visible and well lit.
What is the difference between OmniHuman 2.0 and OmniHuman 1.5?
OmniHuman 2.0 makes longer videos of up to 300 seconds at 720p, ideal for talks, lessons and songs. OmniHuman 1.5 focuses on clips under 60 seconds with up to 1080p. Both offer audio-driven lip sync, expressions and gestures.
Can I control how the character moves in OmniHuman 2.0?
Yes. Add an optional prompt of up to 300 characters to describe actions, gestures, the scene or camera moves. Prompts work in Chinese, English, Japanese, Korean, Spanish and Indonesian. The final motion also follows the audio.
Can OmniHuman 2.0 make singing videos?
Yes. Upload a song as MP3 or WAV and OmniHuman 2.0 syncs the lips, breathing and body rhythm to the music. It suits virtual singers, digital idols and music videos up to 5 minutes long.
What can I make with OmniHuman 2.0?
Use OmniHuman 2.0 for AI presenters, UGC-style product ads, course and training videos, visual podcasts, virtual singers and animated illustrations. Any portrait and voice recording can become a talking video without filming.
How do I start using OmniHuman 2.0?
Sign in, activate VIP and open the OmniHuman 2.0 workspace from this page. Upload a portrait and your audio, add a prompt if you like, and generate. No code and no software to install. Your first talking video can be ready in minutes.