Prepare the portrait
Use a clear frontal portrait so the face and shoulders are readable for lip-sync generation.
Omnihuman 2.0 Avatar Studio
Turn a portrait and an audio clip into an AI avatar video
Upload one portrait photo and one audio track, and Omnihuman 2.0 generates a lip-synced avatar video whose motion follows the audio.
Key points
1+1Required inputs
Exactly one portrait image and one audio track per task.
720pOutput resolution
Avatar videos are rendered at 720p.
4-300sAudio length
Output length and Star cost follow the uploaded audio.
Highlights
- One portrait image plus one audio track drive every generation.
- Output length matches the uploaded audio, from 4 to 300 seconds.
- Video output is rendered at 720p.
- An optional prompt of up to 300 characters can guide the scene and motion.
- Star cost is calculated per second of generated video.
- Each generation appears in a conversation view with status and the final video.
Public Sunday routes
HomeA focused Seedance 3.0 entry point for AI video creation.MembershipCompare Stars, subscriptions, one-time top-ups, and creator access.AI Video GeneratorCreate AI videos with prompts, reference uploads, model controls, and task progress.AI Video EnhancementImprove uploaded video resolution, frame rate, and bitrate with MediaKit.AI Image GeneratorPrepare AI image prompts, reference images, model choices, and image size controls.World Cup 2026Generate football-style AI clips from a portrait with player and stadium presets.HappyHorse 1.1Review the HappyHorse 1.1 landing page, benchmarks, and integrated workflow notes.Seedance 2.5Create 4-30 second multimodal videos with Seedance 2.5 at 480p, 720p, or 1080p.AI Avatar VideoTurn one portrait photo and one audio track into a lip-synced avatar video.CommunityBrowse creator posts, published AI videos, and remixable generation notes.Help CenterRead workflow guidance, model notes, Star cost examples, and support information.AI BotRead API key, upload, task creation, and polling docs for bots, scripts, and agents.BlogRead the latest creator guides, product updates, and AI video insights.Commercial GuideRead commercial AI video guides, prompt examples, and ad workflow ideas.Privacy PolicyUnderstand how Seedance handles accounts, data, payments, and cookies.Terms of ServiceReview platform rules, acceptable use, membership terms, and open-source notices.
About this route
Omnihuman 2.0 is an avatar video workflow: the appearance comes from the uploaded portrait, while lip movement, expression, and timing follow the driving audio. Because the output length equals the audio length, the audio file also determines the Star cost.
Accepted images are JPG, PNG, WebP, BMP, TIFF, GIF, HEIC, or HEIF up to 30 MB with an aspect ratio between 0.4 and 2.5 and each side between 300 and 6000 pixels. Accepted audio is WAV or MP3 between 4 and 300 seconds and up to 15 MB.
The interactive client handles uploads, task creation, progress tracking, playback, and downloads when JavaScript is available.
How the avatar workflow works
Add the driving audio
Upload WAV or MP3 audio from 4 to 300 seconds. The generated video uses the same length.
Optionally describe the shot
A prompt is optional and can add up to 300 characters of scene, motion, and camera guidance.
Review and download
Task progress and the finished video appear in the conversation panel, ready to open or download.
Omnihuman 2.0 questions
What inputs does Omnihuman 2.0 need?
Omnihuman 2.0 requires exactly one portrait image and one audio track. The image supplies the appearance and the audio drives the mouth movement and expression.
How long can the generated video be?
The output length equals the uploaded audio, which must be between 4 and 300 seconds. Star cost is charged per second of the generated video.
Is a prompt required?
No. The prompt is optional and limited to 300 characters. It can describe the scene, the character motion, and camera movement.
What resolution does the output use?
Avatar videos are currently generated at 720p.