MiniMax H3 Max is now the first-choice video model

MiniMax H3 Max Video Generator

Start with a prompt, animate a frame, or direct a clip with visual and audio references. The tool opens with MiniMax H3 Max selected.

Ctrl/⌘ + Enter to generate. 0 / 20000

Create a video from a text prompt, without reference uploads.

25

Enter at least 3 characters to continue.

What you can create with MiniMax H3 Max

Use text-to-video for a new scene, image-to-video when the opening composition matters, and reference-to-video for visual direction. Recast is a separate workflow for replacing a person in existing footage.

When to use this tool

Choose MiniMax H3 Max for a new 5–15 second scene with native audio, starting from a written brief, an opening image, or visual and audio references.

New-clip workflows offer 480P and 768P; use MiniMax H3 when a 2K output is needed. To change a person inside existing footage while retaining its timing and soundtrack, open MiniMax H3 Recast.

A prompt to adapt

A rainy night market, a vendor lifts a steaming bowl, slow push-in, warm lantern light, footsteps and quiet conversation.

Tool specifications

Model
MiniMax H3 Max
Workflow
Text to Video
Output resolution
480P / 768P
Video duration
5–15 seconds

How to use MiniMax H3 Max

  1. 01

    Prepare your brief

    Describe the intended result and add the files required by the selected workflow.

  2. 02

    Choose and review settings

    The matching model or Agent is already selected. Review its settings, inputs and credit estimate.

  3. 03

    Generate and review

    Sign in to submit the task, follow its progress and review the output or storyboard in your workspace.

Check the quote before generating

The live estimate depends on the model, settings and input materials. Sign in and use generation credits to submit a task; review the quote in the tool before you start.

For MiniMax H3 Max text- or image-to-video, output length and resolution set the estimate. Reference-to-video also accounts for reference image dimensions and video or audio length.

View plans and credits

Questions about this tool

How does H3 Max generation differ from Recast?

Text, image and reference workflows create a new 5–15 second clip. Recast uses a 5–30 second source video and person reference photos to replace a character while retaining the source motion and audio.

Do I need an account?

You can explore the tool and its settings on this page. Sign in to upload task materials, submit generation requests and save results to your workspace.

How are generation credits calculated?

Credits depend on the selected model, duration, resolution and input materials. Review the current estimate before submitting. Agent projects may include several generation tasks, each with its own credit usage. For MiniMax H3 Max text- or image-to-video, output length and resolution set the estimate. Reference-to-video also accounts for reference image dimensions and video or audio length.

Where can I find and download my results?

Completed results are available in your workspace. Open the task to preview or download its files. For an Agent project, review the storyboard and the assembled video in the project.

All AI tools