Drop files here, paste files, or add a URL. Required. Exactly one character image. JPEG, PNG, WebP, BMP, TIFF, GIF, HEIC or HEIF, up to 30 MB; 300–6,000 px per side and an aspect ratio between 0.4 and 2.5.
Drop files here, paste files, or add a URL. Required. One WAV or MP3 voice, song or dialogue track, 4–300 seconds and up to 15 MB. The video length follows the audio.
Example Output
This is sample data. Generate a new result to see real output.
Transparent pricing with no hidden fees. Pay as you go.
omnihuman-2.0| Model | Spec | PoYo price | Official price | You save |
|---|---|---|---|---|
| omnihuman-2.0 | 720p | $0.105/sec 21 credits /sec | - | - |
* Actual fees are based on the final output.
One image and one audio track into a talking, singing video
Use omnihuman-2.0 with exactly one image in image_urls and one WAV or MP3 file in audio_url. Audio must be 4–300 seconds and no larger than 15MB. Add an optional prompt to direct actions, scene elements and camera movement. Output is 720p and runs as long as the audio.
Official OmniHuman 2.0 samples show emotional delivery, lip-sync, multi-person dialogue, singing and non-human characters. Turn on the sound to hear the source audio.
01
OmniHuman 2.0 follows the rhythm, intonation and feeling of the audio rather than only moving the lips. Facial expressions, head motion and body language rise and fall with the delivery, so dramatic monologues and calm explanations each look natural.
02
Mouth shapes stay aligned with every syllable, even for characters that are not human. Animals, cartoon mascots, 3D figures and stylized illustrations can all speak or sing while keeping their original design.
03
Start from an image with several people and OmniHuman 2.0 can stage a natural conversation, with the active speaker talking while the others listen and react. Use it for interviews, sketches and dialogue scenes without compositing separate clips.
04
OmniHuman 2.0 accepts audio of up to 300 seconds, enough for a full song, lesson or product walkthrough in one generation. Singing performances follow the melody with matching breath, expression and body rhythm.
05
Give a voice to mascots, pets and animated characters. OmniHuman 2.0 adapts facial movement to each character's anatomy, so a raccoon or a cartoon hero can talk with believable expressions.
Turn recorded lessons and narration into presenter videos. Long-audio support covers full explanations in a single take, and a consistent instructor builds familiarity across a course.
Create spokesperson videos, product explainers and localized ads from one photo. Swap the audio to produce versions in different languages while keeping the same presenter.
Produce singing performances, lyric videos and character skits. OmniHuman 2.0 follows melody and rhythm, making it suitable for covers, virtual artists and short-form entertainment.
Animate a virtual anchor, customer-service avatar or brand ambassador with natural speech and gestures for websites, apps and social channels.
Bring pets, mascots and illustrated characters to life for campaigns, children's content and storytelling without traditional animation.
Add audio-driven avatar generation to your product through PoYo's asynchronous API, with task polling, callbacks and per-second billing.
OmniHuman 2.0 keeps the one-image, one-audio workflow of OmniHuman 1.5 and extends a single take from under a minute to 5 minutes. The trade-off is resolution: OmniHuman 2.0 outputs 720p, while OmniHuman 1.5 also offers 1080p.
| Feature | OmniHuman 2.0 on PoYo | OmniHuman 1.5 |
|---|---|---|
| Inputs | 1 image + 1 audio track + optional prompt | 1 image + 1 audio track + optional prompt |
| Maximum audio and video length | 4–300 seconds (up to 5 minutes) | Under 60 seconds at 720p; under 30 seconds at 1080p |
| Output resolution | 720p | 720p or 1080p |
| Prompt control | Optional; directs actions, scene elements and camera movement | Optional; directs actions, objects and camera movement |
| Multi-person scenes | Supported | Supported |
| Singing | Supported | Supported |
| Non-human characters | Pets, cartoon and anime characters | Animals, anthropomorphic characters and cartoons |
The OmniHuman 2.0 column lists the parameters PoYo accepts. The OmniHuman 1.5 column is based on ByteDance's OmniHuman 1.5 project page, the BytePlus product page and fal's API documentation. For both models, the video runs as long as the input audio.
Step 1: Register and create an API Key.
Sign in to PoYo and create an access key for your application.
Get API Key →
Step 2: Top up credits.
Add credits to your account. The Playground estimates credits from your audio length before you generate.
Add Credits →
Step 3: Integrate the API.
Send model: "omnihuman-2.0" with image_urls, audio_url and an optional prompt to /api/generate/submit, then query the task ID or provide a callback URL.
View API Reference →
No subscription or minimum. You are charged for the seconds of video produced, and the Playground shows an estimate before you submit.
Upload an image and an audio file in the Playground, add a prompt, and inspect the exact JSON request before integrating.
Use the same authentication and task workflow for OmniHuman 2.0, Kling Avatar 2.0 and other video, image and audio models on PoYo.
Poll the task ID or provide a callback URL, then retrieve the finished video through the same asynchronous workflow.
