- Portrait photo plus speech audio
Inputs
- 4 to 300 seconds
Audio length
- Matches the audio
Video length
- Optional
Prompt
About OmniHuman v2
OmniHuman v2 is ByteDance's audio-driven video model, which Masonry runs through fal. You give it a photo of a person and a speech audio file, and it animates the person to match the audio. fal describes the result as vivid, high-quality video in which the character's emotions and movements stay strongly correlated with the audio. The video runs as long as the audio, and an optional text prompt adds control over the scene, movements, and camera.
For business teams, that makes OmniHuman v2 a way to turn a recorded or generated voiceover into a presenter video without a shoot: product explainers, onboarding messages, social updates, and localized versions of the same script. Because the audio drives the clip, the voice track is the script, so approve it before generating. Masonry bills by the length of the input audio, and the editor shows the cost before you generate.
Why teams choose OmniHuman v2
Choose OmniHuman v2 when you already have the voice track and want a talking video from a single photo. The audio sets the timing and the clip length, so there is no separate duration to manage. For video generated from a prompt alone, use a text-to-video model such as WAN 3.0 or FLUX 3 Max. Always get consent before animating a real person's likeness.
What OmniHuman v2 can do
The capabilities that set OmniHuman v2 apart and earn its place in a brief
Photo Plus Audio In, Video Out
Supply one portrait and a speech audio file, and the model animates the person to match the audio.
Movement Tied to the Audio
fal describes the character's emotions and movements as maintaining a strong correlation with the audio.
Length Follows the Audio
The video runs as long as your audio. fal accepts WAV or MP3 audio between 4 and 300 seconds, up to 15 MB.
Optional Prompt Control
Add a text prompt to steer the scene, movements, and camera. fal lists prompt support in Chinese, English, Japanese, Korean, Spanish (Mexican), and Indonesian.
Flexible Portrait Input
fal accepts common image formats up to 30 MB, 300 to 6000 pixels per side, with an aspect ratio between 0.4 and 2.5.
Where teams reach for OmniHuman v2
- Presenter videos for product explainers and tutorials
- Onboarding and internal update messages
- Social posts with a talking spokesperson
- Localized versions of one script with different voice tracks
Frequently asked questions
What teams need to know about creating with OmniHuman v2 in Masonry
What is OmniHuman v2?
OmniHuman v2 is ByteDance's audio-driven video model. It animates the person in a photo to match a speech audio file you supply, with an optional text prompt. Masonry runs it through fal.
How long can an OmniHuman v2 video be?
The video follows your audio. fal accepts audio between 4 and 300 seconds.
What audio formats does OmniHuman v2 accept?
WAV or MP3, up to 15 MB.
What image does OmniHuman v2 need?
A single photo of the person, in a common format such as JPEG, PNG, or WebP, up to 30 MB, 300 to 6000 pixels per side, and an aspect ratio between 0.4 and 2.5.
Does OmniHuman v2 need a prompt?
No. The prompt is optional and adds control over the scene, movements, and camera.
Does OmniHuman v2 generate the voice?
No. You supply the speech audio. Record it or generate it with a voice model first, then pair it with the photo.
How is OmniHuman v2 billed in Masonry?
By the length of the input audio, in credits. The editor shows the exact cost before you generate.
Can OmniHuman v2 output be used commercially?
fal marks the endpoint for commercial use. Get consent from anyone whose likeness or voice you use, and review the current fal and ByteDance terms.
How does my team use OmniHuman v2 in Masonry?
Open a Masonry canvas, pick OmniHuman v2 from the model selector, and describe the video you need: a product shot, an ad creative, a social post. Masonry generates it, then you refine, edit, and combine OmniHuman v2 with other models in one workspace.
Is OmniHuman v2 free to try?
New Masonry accounts get 50 free credits to try GPT Image 2.5 Flare. OmniHuman v2 is included in every paid plan, and the editor shows the exact credit cost before you generate.
How do I write good prompts for OmniHuman v2?
Start with a clear, well-lit, front-facing photo and a clean voice recording without background music. Use the optional prompt to describe gestures, framing, and camera, for example "speaks warmly with small hand gestures, slow push-in".
Who makes OmniHuman v2?
OmniHuman v2 is built by fal-ai. Inside Masonry it runs alongside 50+ image and video models, so your team can pick the right one for each brief without switching tools.
Start creating with OmniHuman v2
Generate, edit, and compare across 50+ models in one workspace.
Explore more AI models
Compare OmniHuman v2 with other models teams run in Masonry