Picture published by the creator on Hugging Face.
MiniMax-H3: Make short videos with synced sound from text or images
Added · Updated
Made by MiniMax
View profile on Hugging Face (opens in a new tab)Generate 4–15 second videos with built-in stereo sound from text, images, video or audio, through an online API, web app or desktop app.
Model details
Original model MiniMaxAI/MiniMax-H3
Input Unknown Output Unknown
Ways to use it Browser demo · Run locally · Use through an API
What it does
MiniMax H3 is a general-purpose model that understands text, images, video and audio together, and generates video with built-in stereo sound. The full system has three parts: Context IR, which refines your instructions into a form the model understands; H3 Base, which makes the audio and video at 768p; and H3 Regenerate 2K, which redoes the result at higher 2K resolution. Dialogue is stably supported in 11 languages.
How to use it
Input → action → output: provide a text description, optionally with images, video clips or audio clips, then generate through the online API, web app or desktop app. The output is a video of 4–15 seconds with stereo audio. In first-and-last-frame mode you give zero, one or two images; in reference mode you give images, video and audio as references.
Common use cases
- Describe a scene in text with no image and get a short video with sound.
- Give a first image and a last image and get a video that moves between them.
- Supply reference images, video clips and audio clips and get a new video that follows them.
Can I use this?
Requirements The source lists an online API, a web app and a desktop app, with global and China links on the model page. The license is the MiniMax H3 Community License Agreement; its terms are not summarized here; read them on the model page. Requirements for running the model on your own hardware are not documented there either.
Caveat The source strongly recommends putting Context IR into your pipeline or following its Prompting Guidance, because it matters for output quality. Output is 768 pixels on the shorter side by default; 2K needs H3 Regenerate 2K.
Worth knowing Quality depends on the Context IR step, which the source strongly recommends using. The default output is 768 pixels on the shorter side; 2K needs the separate Regenerate 2K step.
Why it's here
Not trending this week
Last seen on the Hugging Face trending list on 2 November 2026. It stays on BuildTube so you can still find it.
- 6,654 people have liked it on Hugging Face.
- It was downloaded 6,429,606 times in the last 30 days.
Numbers checked on 8 November 2026; not refreshed since.
Behind it
View model pageBuildTube has not run or verified this project. Everything above is written from what the creator published.
Made this?
Quick questions
- What is MiniMax-H3?
- MiniMax H3 is a general-purpose model that understands text, images, video and audio together, and generates video with built-in stereo sound.
- What can I use MiniMax-H3 for?
- Describe a scene in text with no image and get a short video with sound; Give a first image and a last image and get a video that moves between them; Supply reference images, video clips and audio clips and get a new video that follows them
- Can I run MiniMax-H3 on my computer?
- It runs on your own hardware with a model runtime. Requirements for running the model on your own hardware are not documented there either.
- What do I need to use MiniMax-H3?
- Access to the online API, web app or desktop app linked on the model page.
- What license does MiniMax-H3 use?
- The license is the MiniMax H3 Community License Agreement; its terms are not summarized here; read them on the model page.
More from this week
- P-H3-Image-Comfy Make and edit 8-megapixel images in ComfyUI
Turn short text requests into 8-megapixel images and edit pictures on your own graphics card.
- Idol Turn one picture of an object into a 3D Gaussian splat
Idol turns one photo, render or generated image of an object into a 3D Gaussian splat file.
- agent-desktop Let AI agents read and click desktop apps
For developers building computer-use agents: agent-desktop lets an AI agent read and operate desktop apps on Mac, Windows or Linux from a command line.
Explore by topic
Post this
Download share imageFor makers
Made this? Show it’s on BuildTube — add the badge to your README or embed the card on your site.
README badge (Markdown)
[](https://buildtube.app/p/hf-minimaxai-minimax-h3/)Embed card (HTML)
<iframe src="https://buildtube.app/embed/hf-minimaxai-minimax-h3/" width="560" height="240" style="border:0;max-width:100%" loading="lazy" title="MiniMax-H3 on BuildTube"></iframe>