Audio and video · Independent · 20 Oct 2026

MiniMax-AI/MiniMax-H3

← Audio and video
MiniMax H3: video and audio model. Generates videos and stereo sound from prompts.

In simple English

This tool generates 15-second videos with stereo sound from your text prompts, images, or reference clips. You can run it on your own server or use their online API. It requires massive computer hardware with multiple graphics cards to run locally. The model was built by MiniMax AI to compete with top commercial video generators.

What it is

MiniMax H3 is an omni-modal generative system that outputs 24 frames per second video and 32 kHz stereo audio. It provides open weights for the base model while keeping its context preprocessing and 2K upscaler behind an API.

What it is good for

Solo creators and developers who want to generate video and audio scenes from prompts, animate still images, or combine multi-source reference media.

Stars10k
LicenceNo licence file
LatestNo tagged releasechecked 20 Oct 2026
Last code change2026-08-15
Written inPython
Setup effort
5 / 5
Payoff
4 / 5
Good
  • Generates video and synchronized stereo audio at the same time.
  • Supports text-to-video, image-to-video, and reference-to-video workflows.
  • Provides open weights compatible with standard serving frameworks like SGLang and vLLM.
Watch for
  • Requires multiple high-end graphics cards with massive memory to run locally.
  • The 2K resolution upscaler and context preprocessing modules are closed-source API services.
  • Setup and configuration for distributed inference require advanced technical knowledge.
Replaces

Closed-source video generation platforms and manual video-audio editing

Works well with

SGLang

Similar repositories
  • zai-org/CogVideo13k stars

    Another open model family for text-to-video generation with alternative architecture choices.

  • Wan-Video/Wan2.117k stars

    An open-source video generation model supporting high-resolution outputs and text-to-video tasks.

Also reviewed on Repo Radar
  • motion-canvas/motion-canvas19k stars

    A code-written animation tool with a live editor, for explanatory videos built from shapes and text.

  • paulpacifico/shutter-encoder2.8k stars

    A desktop window over FFmpeg for converting, trimming and processing video, audio and images.

  • htekdev/vidpipe232 stars

    It watches a folder for recordings and produces captions, clips and draft social posts from each one.

Install
hf download MiniMaxAI/MiniMax-H3 --include "model_index.json" "FL2VA/*" --local-dir MiniMax-H3

Independent: reviewed for Repo Radar outside the daily editions. Every independent repository

Get the next edition in your inbox

A dozen repositories, opened and checked. The licence read, the last release dated, and the ones that did not make it named with the reason. It is the half most lists leave out.

No tracking pixels. One click to leave. The archive stays free either way.

We use your address to send the edition and nothing else. Confirm by email, leave in one click. How we handle it.