MiniMax H3 is a multimodal generation model designed for coherent video, native audio, and richer prompt following.

Can I use images as references?

Yes. Add one or more visual references to guide identity, composition, material, and style.

Can H3 generate audio?

Audio controls are available when native sound is supported by the configured generation provider.

Eligible failed tasks return their reserved credits automatically and idempotently.

Text to video turns a production brief into expressive motion and connected visual beats.
Image to video animates a reference while preserving the intended subject and composition.
Multimodal prompt control combines written direction and visual references in one generation workflow.
Unified audio and video lets you plan sound and image as parts of the same scene.
Longer narrative beats let you develop a complete beat rather than a single disconnected movement.
Reference-driven direction uses reference images to guide identity, materials, composition, and style.

Runs in a web browser.
Audio controls are available only when native sound is supported by the configured generation provider.
Eligible failed tasks return their reserved credits automatically.

What is MiniMax H3?: MiniMax H3 is a multimodal generation model designed for coherent video, native audio, and richer prompt following.
Can I use images as references?: Yes. Add one or more visual references to guide identity, composition, material, and style.
Can H3 generate audio?: Audio controls are available when native sound is supported by the configured generation provider.
Are failed generations refunded?: Eligible failed tasks return their reserved credits automatically and idempotently.

Create expressive AI video from text and images, with coherent motion, native sound

Unified audio and video