Give your image a voice. Turn a portrait and audio into expressive video, with natural lip sync, gestures and movement.
Preview your cost without signing in. 10 credits = 1 billed second.
First video free, up to 6 seconds after email verification. No card required. Dollar equivalent uses Starter monthly (about $0.82/s). Other plans have different rates.
AI-generated demonstration from a stock photo and synthetic narration. Native 720p. The person pictured is not endorsing this service.
Original portrait used for this generation. Click to view the full image.
First video free, up to 6 seconds after email verification. No card required. Your creations stay private.
An independent creation platform. Not affiliated with or endorsed by ByteDance.
A real 15-second result generated with the same model used in the studio. Compare the animated performance with the original image.
AI-generated demonstration from a stock photo and synthetic narration. Native 720p. The person pictured is not endorsing this service.
Original portrait used for this generation. Click to view the full image.
Start with the character you imagine. Add the audio you want them to perform.
A portrait, an illustrated character or an original avatar establishes the appearance of your video.
Speech and song guide the character’s lip movements, facial expressions and gestures.
An optional prompt guides the scene, character movements and camera.
Create 720p videos with audio from 4 seconds up to 5 minutes.
Choose one clear image of the person or character you want to animate.
Upload speech or music, then add an optional prompt to guide the performance.
Sign in, check the credit cost, and find the finished video in your private library.
See the verified credit cost before you start a generation.
Failed generations return their credits to your balance.
Your source files and generated videos are accessible only through your account.
Subscription credits and credit packs work together in one account.
Compare duration, published API prices and the editing needed for longer audio. Read the full comparison for provider-specific limits, Kling Avatar and lip-sync alternatives.
| What changes | OmniHuman 2.0 | OmniHuman 1.5 |
|---|---|---|
| Audio per generation | 4–300 seconds on KIE | Less than 60 seconds; BytePlus documents this limit |
| A 300-second performance | Fits one supported request; no manual segment stitching needed | Exceeds one request; split into shorter clips and assemble in an editor |
| BytePlus public API price | $0.07 per second | $0.12 per second |
API prices are the named providers’ published list prices, checked 27 October 2026, before applicable discounts or taxes. They are not this website’s retail prices. The stitching row is a workflow conclusion from the documented duration limits, not a visual-quality benchmark.
For a presenter-style clip, write one clear opening and one main point before recording. A short introduction is easier to review than a long performance with several unrelated ideas. Leave a little space between sentences so the delivery sounds natural, then listen for background noise and accidental interruptions. Choose an image whose expression fits the topic: an excited expression may feel out of place in a calm explanation. The uploaded recording supplies the actual spoken words. If a sentence needs changing, edit or record the audio again before generating another version. The tool does not rewrite a script after you upload it, and its movement direction does not function as an audio editor.
An original illustrated character can be a starting point for a narrated story or a musical performance. Select a single clear subject rather than a crowded scene in which the intended speaker is ambiguous. Keep the visible face, posture, and clothing consistent with the performance you are planning. Your optional direction can then focus on a modest change, such as a friendly expression or gentle gestures, instead of trying to repair an unsuitable source image. Results are generative, and a direction is guidance rather than a guaranteed choreography. Compare the finished clip with your creative aim, and change the source image if the same problem persists across different directions.
Decide where the finished clip will be used before spending credits. A short speaker introduction, an explanation inside a longer presentation, and a vocal character performance need different source recordings even when they use the same portrait. Review the downloaded clip in the editing or publishing software you normally use. If you need captions, music mixing, a title card, or a different final layout, prepare those elements in that software; this generator focuses on the portrait performance itself. Keep a copy of the original image and edited recording so you can produce another version deliberately. A fresh confirmed task uses credits again, so make changes to the source before starting the next attempt.
Choose what the person or character should say, then record a short introduction, product explanation, original story, or vocal performance that you have permission to use. Listen to the recording before uploading it. Audio determines the length and rhythm of the result, so removing an unwanted pause before generation is more useful than asking a prompt to remove it afterward. Choose a portrait that fits the recording and shows the intended subject clearly. This workflow creates an audio-driven performance from your files. The optional direction does not create a voice from text, replace the recording, or provide a frame-by-frame animation timeline.
Use one still JPEG, PNG, or WebP image up to 10 MB and four megapixels. Both dimensions must be between 300 and 6000 pixels, and width divided by height must fall between 0.4 and 2.5. A 1000 by 1500 portrait meets these rules. A 3000 by 3000 image exceeds the site pixel limit even if its compressed file is small. These site limits keep validation manageable. Select a clear face with an unobstructed mouth for a talking portrait. Results depend on the source and model, so treat the first performance as a creative draft to inspect before publication.
The generator measures your uploaded audio, rounds its duration upward to a whole second, and charges 10 credits per quoted second. For example, a 12.2-second recording requires 130 credits; an exact 30-second recording requires 300. These examples explain billing and are not generated-video samples. Preview the cost as a guest, then sign in to confirm generation. Annual subscriptions are billed for a year and issue allowances monthly. Unused subscription credits expire at the end of their allowance period; separately purchased packs do not expire. Both contribute to one balance, with expiring credits used before nonexpiring packs. Verify your email to receive 60 trial credits once, enough for one 4–6 second video.
Watch a completed 720p performance through before publishing it. Check the lip movement, expression, hands, and beginning and ending of the clip. The direction guides the model but does not guarantee exact gestures. If the browser loses its connection after submission, look for the existing attempt in your private library before starting another generation. The site returns the credits once the provider confirms a failure. An attempt awaiting confirmation may still be processing and is retained for reconciliation. Your uploads remain private by default, while the generation provider receives temporary access needed to process them. Use only source files you are authorized to submit.
Model input requirements: KIE OmniHuman 2.0 documentation. Open the video workspace
Try your first video free, up to 6 seconds. Then choose monthly credits or a one-time pack. 10 credits = 1 billed second.
600 credits each month
Billed monthly · cancel anytime
Up to 60 seconds of video
≈ $0.82 / second
2,000 credits each month
Billed monthly · cancel anytime
Up to 200 seconds of video
≈ $0.50 / second
4,800 credits each month
Billed monthly · cancel anytime
Up to 480 seconds of video
≈ $0.41 / second
Checking checkout availability…
Audio is rounded up to whole seconds for each video. Dollar rates assume the full allowance is used. Annual plans are prepaid for a year and deliver credits monthly. Subscription credits expire at the end of each allowance month. Credit packs never expire.
One portrait image and one MP3 or WAV audio track. You can also add an optional prompt to guide the scene, movement or camera.
This model requires an image and audio. Your optional text prompt provides creative direction; it does not replace either input.
Select your image and audio to preview the cost without signing in. Audio duration rounds up to whole seconds, at 10 credits per second. A 30-second clip uses 300 credits. Each plan shows its dollar equivalent per second. Sign in only when you are ready to confirm generation.
Yes. Verify your email to receive 60 credits once, enough for one 4–6 second video. No card is required. Longer videos require purchased credits. Repeated sign-ins do not reset the trial, and confirmed failed generations return the credits.
Your videos and uploads are private by default. Sign in to your account to view and download your work.
When the provider confirms a failed generation, the credits used for that task are returned automatically.
You pay for the year once and receive a new credit allowance every month. Each monthly allowance expires at the end of its allowance period.
