OmniHuman 2

Why creators choose OmniHuman 2

OmniHuman 2 animates a whole performance from one photo and one audio track. It goes beyond lip movement: expression, posture, and hand gesture follow the delivery, so a character reads as performing the line rather than mouthing it.

Long takes in one generation

OmniHuman 2 accepts up to five minutes of audio. A full explainer, a podcast segment, or a long monologue renders as one continuous clip, with no splitting the audio and joining shots afterwards.

Expressive hand and body gesture

Hands move with what's being said and stay coordinated with the body. Gesture develops across a long speech instead of repeating a loop.

Prompt control over scene and camera

An optional prompt steers the scene, the character's movement, and the camera. Leave it out and the model performs from the audio alone.

Works across content types

Photographs, AI-generated renders, animated characters, and stylized art all work as the source picture.

How to use OmniHuman 2 in Craze

OmniHuman 2 is a transform model: provide one photo and one audio track, and Craze generates the animated performance.

1. Upload a clear photo of the subject, at least 300 pixels on each side.
2. Upload or record the audio, from 4 seconds to 5 minutes. This can be a voiceover, dialogue, text-to-speech, or any audio track.
3. Ask Craze for OmniHuman 2, or describe the digital human you want to create. If you don't name a model, Craze shows available avatar options with pricing.
4. Craze generates the animated video with gesture, expression, and body language matching the audio.
5. Review in the editor. If you want to adjust the performance or try different audio, regenerate.

OmniHuman 2 works well on photos where the subject is clearly visible and facing somewhat toward the camera. Photos don't need to be professional; phone photos and artistic renders both work.

OmniHuman 2 pricing

OmniHuman 2 is priced by the length of your audio at $0.111 per second of audio. A 30-second clip costs half as much as a 60-second one. Pricing scales linearly with audio duration. Calculate exact costs on this page.

Photo source and video format don't affect the price. Only the duration of the audio matters.

OmniHuman 2 compared to other avatar models

Kling Avatar 2 Pro and Kling Avatar 2 offer simpler animation for shorter videos. VEED Fabric 1.0 is a low-cost option that mostly moves the head and shoulders. OmniHuman 2 excels when you need expressive gesture, cinematic quality, or a long take in one clip.

Who OmniHuman 2 is for

Film and narrative production

Create digital humans for shorts, web series, and indie films. Expressive gesture and prompt-steered camera work give characters presence in a scene, not just an animated face.

Commercial and advertising

Animate brand mascots or spokesperson characters with nuanced performance. Gesture and expression create engagement that static avatars don't achieve.

Podcast and interview video production

Podcasters and audio creators can turn a full segment into video in one generation. Five minutes of audio per clip covers most segments without cutting.

Educational and training at scale

Create instructor avatars for online lessons. A full lesson section renders as one take, with gesture that keeps the video engaging to watch.

Interactive and immersive experiences

Game developers and XR creators can use OmniHuman 2 to generate NPC performances or interactive characters with natural dialogue and gesture.

Accessible performance for performers

Actors, musicians, and performers can create video content from audio without needing cameras or multiple takes. One good audio performance becomes video with nuanced gesture and expression.