OmniHuman

Overview

What is OmniHuman?

ByteDance’s OmniHuman generates a synced video from a single reference image and an audio track. Trained on 18,700 hours of human motion, it makes audio the main control signal: lips, expressions and movements follow the voice.

In Havincy, upload a portrait and a voice to get a video of a person speaking, with no motion-capture equipment. The length follows the audio.

Key strengths

• Talking portrait
• Audio drive
• Présentation

Specs

OmniHuman: specifications

Publisher

ByteDance

Generation modes

talking avatar (image + voice)

Available since

July 27, 2025

License

Commercial use allowed

Cost in Havincy

From 155 credits

Specs taken from the publisher’s public model page. See the documentation on fal.ai

Use cases

What you can create with OmniHuman

Prompt examples

Prompts to try with OmniHuman

Paste an example into Havincy, then adapt the subject, style and format to your project.

• 01

  Make this portrait speak with this 20-second voiceover.

• 02

  Animate our advisor’s photo so he explains our offer.

• 03

  Create a virtual presenter from this image and this audio script.

Use cases

Compatible skills

Use this model with the following Havincy skills.

How it works

How to create with this model

1. Describe the desired result in a Havincy conversation.
2. Add a reference when the style, product, or character needs to be preserved.
3. Generate, then ask for a variation or improvement without starting over.

FAQ

Frequently asked questions about OmniHuman