OmniHuman 2.0 Video Generator
Upload one portrait photo and one voice track — OmniHuman 2.0 turns them into a talking, singing, or performing video. No camera, no actors, no editing. First video free.
Portrait
Aspect ratio
Voice
Direction
First video is free — just your email, no credit card.
Example output — try it with your own photo & voice
What can this OmniHuman 2.0 generator do?
OmniHuman 2.0 is a audio-driven human animation model released by ByteDance. Give it a single portrait photo and a single audio track, and it returns a realistic video of that person speaking or performing the audio — accurate lip sync, natural facial expressions, head movement, and hand gestures included. People use it to make talking avatars for TikTok and YouTube Shorts, narrate course and product videos without filming, bring historical or fictional characters to life, and create singing portraits from any song. Everything runs in the browser: upload two files, click generate, and download an MP4.
How it works — three steps
Step 1, Portrait: drop in a photo (JPG, PNG, or WEBP up to 5MB, compressed locally in your browser before upload) or describe a fictional person in words and let the model create the portrait. Pick an aspect ratio — 3:4, 1:1, 16:9, or 9:16, which is recommended for TikTok, Reels, and Shorts. Step 2, Voice: upload an MP3 or WAV up to 30 seconds, or type up to 300 characters and choose one of eight built-in voices. Step 3, Direction (optional): add a short prompt describing emotion, camera movement, or scene mood to steer the performance. Accept the terms, click Generate video, and watch the progress panel.
Output, duration, and pricing
Every generation renders a standard MP4 file you can download and post anywhere. Each video runs 30 seconds per generation today, with longer durations up to 300 seconds (5 minutes) rolling out for music and dialogue scripts as the roadmap extends audio limits. A 30-second video costs 30 credits, shown on both this page and the homepage so the number is always the same. If a render fails, your 30 credits are refunded automatically — you are never charged for a failed job.
Frequently asked questions
Is the OmniHuman 2.0 generator free to try?
Yes. Your first video is free — just sign up with an email address, no credit card required. New accounts also receive 50 bonus credits at signup, and every 30-second video after that costs 30 credits.
What kind of photo works best?
Use a clear, front-facing portrait of one person with visible eyes and mouth. JPG, PNG, and WEBP files up to 5MB are supported, and your browser compresses the image before anything is uploaded. Avoid group shots, heavy filters, and sunglasses.
What audio can I use?
Upload an MP3 or WAV file up to 30 seconds long — speech, singing, or music all work. Alternatively, type up to 300 characters and pick one of eight text-to-speech voices. The audio track drives every movement in the final video.
How long does rendering take?
A 30-second video typically finishes rendering in one to two minutes. The preview panel shows live progress while you wait — keep the tab open — and the finished MP4 downloads directly from the page.
What happens if a generation fails?
Nothing is charged for failed renders. If a job fails for any reason, the 30 credits are refunded to your balance automatically, and you can retry immediately with the same inputs or adjusted ones.