MiniMax Hub — Codex Skill

Integrate MiniMax's AI creation capabilities into Codex CLI through the local MiniMax Hub Gateway.

English | 中文

Features

Category	Capabilities
🖼️ Image Generation	Text-to-image, image editing, style transfer, batch (up to 8 concurrent)
🎬 Video Generation	DAG async workflow for long-running tasks (5-30 min)
🎤 Speech Synthesis	594+ preset voices (CN/EN/JP/KR/FR/DE...), voice cloning
🎵 Music Generation	Lyrics→song, instrumental BGM, cover songs
✂️ Audio/Video Editing	Concatenation, lip-sync, super resolution, ASR, audio separation

Prerequisites

1. MiniMax Hub installed and running
   • Download: MiniMax Official Site
   • Gateway listens on http://localhost:8001 by default
2. Python 3.8+ (optional, for helper script)

Installation

Via npx skills (recommended)

npx skills add zzj2004/minimax-hub

Manual installation

git clone https://github.com/zzj2004/minimax-hub.git

# Windows:
copy minimax-hub %USERPROFILE%\.agents\skills\minimax-hub\

# macOS / Linux:
cp -r minimax-hub ~/.agents/skills/

Restart Codex after installation.

Usage Examples

Just tell Codex:

"Generate an image of a cat using MiniMax" "Convert this text to speech with MiniMax Hub" "Create a piano BGM with MiniMax" "Generate a video of a sunset beach using MiniMax"

CLI Helper

# Health check
python scripts/minimax.py health

# List available voices
python scripts/minimax.py voices

# Generate image
python scripts/minimax.py image "sunset beach, oil painting style"

# TTS (Japanese voice)
python scripts/minimax.py tts Japanese_Whisper_Belle "Hello, welcome to MiniMax Hub"

# Submit video DAG task
python scripts/minimax.py dag run video_gen '{"prompt":"A cat running on grass"}'

API Endpoints

All requests go through http://localhost:8001. See references/gateway-api.md for full docs.

Image Generation

curl -X POST http://localhost:8001/api/generate/image \
  -H "Content-Type: application/json" \
  -d '{
    "backend": "nano_banana",
    "model_id": "banana_2",
    "prompt": "A cute orange cat, photorealistic",
    "filename": "cat_output"
  }'

Speech Synthesis

curl -X POST http://localhost:8001/api/generate/speech \
  -H "Content-Type: application/json" \
  -d '{
    "backend": "minimax_tts",
    "prompt": "Hello world",
    "filename": "hello",
    "params": {"voice_id": "Friendly_Person", "speed": "1.0"}
  }'

Video Generation

curl -X POST http://localhost:8001/api/dag/run \
  -H "Content-Type: application/json" \
  -d '{
    "dag_id": "video_workflow",
    "inputs": {"prompt": "video description"},
    "asset_keys": ["video"]
  }'

Configuration

Env Variable	Default	Description
MINIMAX_GATEWAY	http://localhost:8001	Gateway URL
MINIMAX_TIMEOUT	300	Request timeout (seconds)

Project Structure

minimax-hub/
├── SKILL.md                  # Main skill file (triggers Codex)
├── README.md                 # English docs
├── README.zh-CN.md           # Chinese docs
├── LICENSE                   # MIT License
├── agents/
│   └── openai.yaml           # UI metadata
├── scripts/
│   └── minimax.py            # Cross-platform CLI helper (Python 3.8+)
└── references/
    └── gateway-api.md        # Gateway API reference

Compatibility

Item	Requirement
OS	Windows / macOS / Linux
Codex	Any version
Python	3.8+ (helper script)
MiniMax Hub	2025+
Network	Localhost only (no internet required)

FAQ

Q: Gateway connection failed? A: Make sure MiniMax Hub is running. Check http://localhost:8001/api/dag/run.

Q: "Backend not registered" error? A: Wait a few seconds and retry — Gateway may still be initializing.

Q: How to update this skill? A: Re-run npx skills add or git pull, then restart Codex.

License

MIT