Skip to main content
Not on Enterprise yet? Get your enterprise plan at ModelsLab
MiniMax H3 runs directly on your own enterprise container — requests are posted straight at your dedicated server instead of going through the shared queue. The model exposes three capabilities across two generation endpoints: Text to video has an endpoint to itself, so it takes no model_id. Both image-to-video modes share one endpoint and nothing in the payload separates them, so model_id is required there.

​
Available Endpoints

Text to Video

This endpoint generates a video from a text prompt using the MiniMax H3 model.

Image to Video

This endpoint animates between a first and a last frame, or generates a video from image, video and audio references.

Fetch Queued Videos

This endpoint fetches already generated or queued videos by ID.

Clear Server Cache

This endpoint clears the server cache.

Clear Server Queue

This endpoint clears the server queue.

Restart Server

This endpoint restarts the server.

System Details

This endpoint retrieves system and server details.

Update Server

This endpoint updates the server to the latest version.