How Do I Call MiniMax H3 Text-to-Video Through an API?

Overview

Calling text-to-video is a flow, not a single request, so think in steps rather than one long code block. You submit a generation task with your prompt and parameters, receive a task ID, poll for status until it completes, then retrieve the output. Confirm the exact endpoint, payload, and parameter names in the current MiniMax H3 documentation before you wire it up.

Source note: Verified 2026-08-06 against the MiniMax official H3 blog, MiniMax Video Generation API docs, and Hugging Face MiniMax-H3 model page.

Walk the path deliberately. In the create call you send the text prompt plus settings like duration and resolution, formatted to the documented schema. Because video is asynchronous, that call returns a task ID quickly rather than the finished clip. You then query the status endpoint on a sensible interval, branching on terminal states such as completed or failed instead of assuming success. On completion you fetch the output URL and, ideally, copy the file to your own storage so you are not depending on a link that may expire.

Wrap each step in its own function and keep the task ID as the connecting thread. That structure makes retries, logging, and later maintenance far easier than a single tangled request.

Once this pattern works for text-to-video, the same submit-poll-retrieve shape carries over to image-to-video and reference modes with only the input payload changing.