MiniMax H3 Documentation

MiniMax H3 is a multimodal AI video model that can combine text, images, video, and audio references in one generation workflow. This documentation explains how to use the independent MiniMax H3 online workspace at minimaxh3.pro—from a first prompt through reviewing saved results.

Choose a workflow

Start with text

Use Getting Started for a complete orientation, then follow the Text to Video guide to turn a written scene description into a video. These pages explain supported controls, credit estimates, prompt structure, and what happens after generation begins.

Control the opening and ending

Use First and Last Frame when composition at both ends of the shot matters. This workflow is useful for product reveals, transitions, before-and-after sequences, and shots that must arrive at a defined final state.

Direct a scene with references

Use Multimodal Reference when images, clips, or audio should guide identity, motion, framing, atmosphere, or sound. The guide explains how to assign each reference a clear purpose instead of overloading the model with conflicting signals.

Write better MiniMax H3 prompts

The Prompting guide provides a reusable structure for subject, action, camera, look, sound, and constraints. It also explains how the homepage Try This controls copy an example-specific prompt into the Playground, so you can study a working direction before adapting it.

Understand credits and saved results

• Credits and Pricing explains the credit estimate shown before generation, monthly plans, one-time credit packages, and why 2K generations require more credits than 768P output.
• Video History explains where completed generations appear, what each status means, and how saved R2 links help keep a result available after an upstream temporary URL expires.

Recommended path for new users

1. Read Getting Started.
2. Create a simple shot with Text to Video.
3. Improve the direction using the Prompting guide.
4. Move to First and Last Frame or Multimodal Reference when you need more control.
5. Review the estimate in Credits and Pricing.
6. Find completed outputs in Video History.

What this documentation covers

These docs focus on the workflows available on this website. They do not present the site as the official MiniMax API documentation, and they do not promise that every capability discussed by the broader open-source community is available in the hosted Playground. For research on local deployment, ComfyUI, VRAM, GGUF, LoRA, Hugging Face, native audio, cost, and model comparisons, visit the MiniMax H3 Blog.

Need help?

If a generation stays pending, a completed link does not appear, credits look incorrect, or checkout fails, contact support@minimaxh3.pro with the email address on your account and the generation ID when available. Do not send API keys, passwords, payment card details, or other secrets.