H3 StudioMiniMax video & sound

Give your scene a soundtrack.

Describe a moment. H3 creates the motion and the sound together.

MiniMax-H3 Community demo

Your scene

Include action, camera movement, and sound.0 / 6000
Guide with images Optional

Add a beginning, an ending, or both. Frames are cropped to your selected aspect ratio before upload.

JPG, PNG, or WebP. Up to 10 MB each.

Generation settings

Longer clips, larger frames, and more steps use more GPU quota.

Prompts and selected images are sent to multimodalart’s H3 demo and its conditioner. Shared ZeroGPU limits apply.

Scene monitor

960 × 544 / 24 fps
Sample output · Snow & silence
Play the sample with sound, or generate your own scene.

Start with a scene

Load a prompt, then make it yours

Write the sound into the scene. Footsteps, dialogue, ambience, or music: describe what you want to hear alongside what you want to see.

Connection options & quota

Anonymous requests use the shared backend’s limits. If generation is blocked by quota, try a shorter clip or open the inference demo and sign in there. You can also supply your own Hugging Face read token for API access. A token does not guarantee available GPU quota.

Kept only in this page’s memory until cleared or reloaded. Used only with Hugging Face and the inference Space; never included in the public source. Create a token with minimal read permissions in Hugging Face settings.