SkyNotSilent/awesome-MiniMax-H3-cases

MiniMax H3 Cases & Guides – What It Is

MiniMax H3 Cases & Guides is an open‑source, community‑maintained catalog of real‑world outputs produced with the MiniMax H3 video‑generation model (also known as Hailuo H3). The repository stores metadata for thousands of playable videos, the exact prompts that created them (when the original author has published the full text), and a set of step‑by‑step tutorials that show how to reproduce or extend those results on various hardware setups.

The project does not host the video files themselves – they live in external storage – but it provides a searchable web front‑end (hosted at https://h3‑field‑notes‑production.up.railway.app) that lets anyone watch the cases, inspect verified prompts, and follow vetted tutorials without needing a login.

---

Core components (as described in the README)

Component	What it contains	How it’s used
cases.json	Metadata for 2 052 cataloged video cases (source, creator, timestamps, duration, style, etc.)	The front‑end reads this to build the searchable gallery; developers can import it to build custom tools.
project‑stats.json	Live counters for cases, complete prompts, tutorials, and creator rankings	Shown in the badge shields on the README and used by CI to verify that the site and README stay in sync.
tutorial‑guides.json & tutorials.json	Structured, bilingual (English/Chinese) guides that map a goal (e.g., “first generated video”) to a concrete deployment route (cloud template, NVIDIA ComfyUI, Apple‑Silicon C + Metal, low‑VRAM quantized, etc.)	Users click a tutorial card, get a “Copy for AI” package, and run the exact commands on their machine.
creators.json	Aggregated profiles of 1 043 source‑attributed authors, with rankings based on community‑verified contributions	Enables discovery of prolific creators and their full set of cases and tutorials.
Agent Skills	Three npm‑installable skills for Codex and Claude‑Code agents (minimax-h3-prompt-library, minimax-h3-tutorial-guide, minimax-h3-submission-helper)	Allow AI assistants to query the catalog, suggest a tutorial matching a hardware/goal combo, or prepare a ready‑to‑submit issue for new cases.

---

What you can do with it

1. Browse & get inspired – Watch any of the 2 k+ videos directly in the web gallery, filter by date, duration, style, or whether a full prompt is available.
2. Study proven prompts – When a case includes a complete, verbatim prompt, the library shows the exact text and a link back to the original source, so you can see exactly how the result was generated.
3. Reproduce results – Pick a tutorial that matches your hardware (NVIDIA GPU, Apple Silicon, low‑VRAM, etc.) and follow the step‑by‑step guide; the “Copy for AI” button creates a ready‑to‑run command bundle.
4. Contribute – Submit your own video case or tutorial via the provided GitHub issue templates; accepted submissions are credited, added to the catalog, and can appear in a rotating spotlight.
5. Automate with agents – Install the provided Skills so a code‑generation assistant can fetch prompts, recommend tutorials, or scaffold a submission without hallucinating any missing information.

---

Who it’s for

• AI creators & hobbyists who experiment with MiniMax H3 and want a reliable source of real examples.
• Researchers or educators looking for concrete, source‑verified prompts and video outputs to illustrate generative‑video techniques.
• Developers building tooling around MiniMax H3 (e.g., search interfaces, recommendation engines) who need a clean, version‑controlled dataset.
• AI‑assistant developers who want ready‑made Skills to let large‑language‑model agents interact with a curated video‑generation knowledge base.

---

Getting started (quick‑run)

# Clone the repo
git clone https://github.com/SkyNotSilent/awesome-minimax-h3-cases.git
cd awesome-minimax-h3-cases

# Install dependencies (Node 22+ recommended)
npm ci

# Run the local development server – the catalog works offline
npm run dev

The site will be available at http://localhost:3000. No API keys are required for browsing cases; only the optional media playback needs you to configure compatible object‑storage if you want to host the videos yourself.

---

License & provenance

• The code (website, data‑processing scripts, Skills, etc.) is released under the MIT license.
• The videos, prompts, creator names, and other collected content remain under the terms of their original platforms (X, GitHub, YouTube, etc.). The project’s CONTENT_POLICY.md explains how it respects those rights.
• The catalog is not affiliated with MiniMax; it is a community effort that mirrors the output of the MiniMax H3 model while keeping strict provenance (no reconstructed or reverse‑engineered prompts).

---

TL;DR

MiniMax H3 Cases & Guides is a meticulously curated, source‑attributed library of MiniMax H3 video outputs, complete prompts, and reproducible tutorials. It offers a searchable web UI, machine‑readable JSON data, and ready‑to‑install AI‑agent Skills, making it a practical resource for anyone who wants to learn from or build upon real MiniMax H3 creations.