h3 max is a developer-facing API gateway for H3 Max video generation. Instead of placing a provider credential in client code, h3 max issues a site key that starts with h3_live_, validates each request at the edge, and then submits approved jobs to H3 Max text-to-video or image-to-video routes from the server. The goal is a small, practical bridge between model documentation and a real integration test: one page for creating a key, sending a prompt, checking queue state, and copying the exact request shape your backend can use.

H3 Max is presented by fal as a post-trained variant of MiniMax H3 for higher-quality video generation. MiniMax H3 itself is a multimodal model family described by MiniMax as supporting text, image, video, and audio context with video output that can include synchronized stereo audio. This site keeps those source boundaries visible. It does not claim to be the model creator, it does not expose upstream provider credentials, and it does not treat open-weight MiniMax H3 references as the same thing as the hosted H3 Max API routes. For teams evaluating the h3 max api, the useful first step is to validate the request contract, queue behavior, resolution and duration choices, safety settings, and result payload before building billing, teams, or automation around it.