MiniMax-MCP MCP — Cas d'usage, Installer & Démo en direct

Pourquoi l'utiliser

Fonctionnalités clés

text_to_audio with multiple voices
voice_clone from audio samples
voice_design — generate a voice from a text description
text_to_image and generate_video (async with query_video_generation)
music_generation with lyrics input

Démo en direct

Aperçu en pratique

minimax.replay ▶ prêt

0/0

Installer

Choisissez votre client

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "minimax": {
      "command": "uvx",
      "args": [
        "MiniMax-MCP"
      ],
      "_inferred": true
    }
  }
}

Ouvrez Claude Desktop → Settings → Developer → Edit Config. Redémarrez après avoir enregistré.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "minimax": {
      "command": "uvx",
      "args": [
        "MiniMax-MCP"
      ],
      "_inferred": true
    }
  }
}

Cursor utilise le même schéma mcpServers que Claude Desktop. La config projet l'emporte sur la globale.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "minimax": {
      "command": "uvx",
      "args": [
        "MiniMax-MCP"
      ],
      "_inferred": true
    }
  }
}

Cliquez sur l'icône MCP Servers dans la barre latérale Cline, puis "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "minimax": {
      "command": "uvx",
      "args": [
        "MiniMax-MCP"
      ],
      "_inferred": true
    }
  }
}

Même format que Claude Desktop. Redémarrez Windsurf pour appliquer.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "minimax",
      "command": "uvx",
      "args": [
        "MiniMax-MCP"
      ]
    }
  ]
}

Continue utilise un tableau d'objets serveur plutôt qu'une map.

~/.config/zed/settings.json

{
  "context_servers": {
    "minimax": {
      "command": {
        "path": "uvx",
        "args": [
          "MiniMax-MCP"
        ]
      }
    }
  }
}

Ajoutez dans context_servers. Zed recharge à chaud à la sauvegarde.

claude mcp add minimax -- uvx MiniMax-MCP

Une seule ligne. Vérifiez avec claude mcp list. Supprimez avec claude mcp remove.

Cas d'usage

Usages concrets : MiniMax-MCP

How to turn a blog post into narration audio

👤 Creators, podcasters, accessibility teams ⏱ ~15 min beginner

Quand l'utiliser : You want a listenable version of written content.

Prérequis

MiniMax account + API key — Global: minimax.io; Mainland: minimaxi.com. Set MINIMAX_API_KEY + MINIMAX_API_HOST matching region
uvx installed — brew install uv

Déroulement

List voices

list_voices — show all available voices in English.✓ Copié

→ Voice catalog with previews
Render audio

text_to_audio: [paste article] with voice 'female-calm-en'. Save to /audio/article.mp3.✓ Copié

→ MP3 file saved, URL returned

Résultat : Listenable audio version of your content.

Pièges

Long articles hit per-call character limits — Split into paragraphs, render, and concatenate
Pronunciation of proper nouns is off — Use SSML-style phonetic hints or the voice_design tool for specialized voices

Combiner avec : filesystem

How to clone your own voice for narration

👤 Creators wanting a consistent narrator ⏱ ~10 min intermediate

Quand l'utiliser : You want AI narration in your own voice.

Prérequis

10-60 second voice sample (clean audio, your voice) — Record via your phone; export as mp3/wav

Déroulement

Clone

voice_clone with sample /recordings/me.wav, name 'my-voice'.✓ Copié

→ Voice id returned
Use it

text_to_audio with voice='my-voice' on [paste text].✓ Copié

→ Output in your voice

Résultat : Personal narration without a mic every time.

Pièges

Using someone else's voice without consent — Don't — legal and ethical issue; use voice_design instead

How to generate an image and turn it into a short video

👤 Content creators, marketers ⏱ ~15 min intermediate

Quand l'utiliser : Social content, product previews.

Déroulement

Generate image

text_to_image: 'minimal studio product shot of a ceramic mug, natural light'.✓ Copié

→ Image URL
Animate

generate_video with prompt 'slow rotating mug on a wooden table'. Poll status with query_video_generation.✓ Copié

→ Video URL when ready

Résultat : Short marketing clip.

Pièges

Video generation is async and can take minutes — Always use query_video_generation to poll — don't block a chat waiting

Combiner avec : google-workspace

Combinaisons

Associez-le à d'autres MCPs pour un effet X10

minimax + filesystem

Save generated assets into a structured content folder

Generate narration for each chapter in /manuscript/ and save MP3s to /audio/chapters/.✓ Copié

minimax + google-workspace

Produce audio from a Google Doc, then attach as a link in the doc

Read Doc 'Launch Post', generate narration, upload to Drive, append the link as a footer.✓ Copié

Outils

Ce que ce MCP expose

Outil	Entrées	Quand appeler	Coût
text_to_audio	text: str, voice: str, format?: 'mp3'\|'wav', output_path?	Narrate text	Per MiniMax TTS pricing
list_voices	language?: str	Find a voice	free
voice_clone	sample_path: str, name: str	Custom voice	Per MiniMax pricing
voice_design	description: str	Generate a voice from prose	Per MiniMax pricing
text_to_image	prompt: str, aspect_ratio?	Image gen	Per MiniMax image pricing
generate_video	prompt: str, duration_s?	Kick off video gen	Per MiniMax video pricing
query_video_generation	task_id	Poll video job	free
music_generation	lyrics: str, genre?: str	Create a short track	Per MiniMax music pricing

Coût et limites

Coût d'exécution

Quota d'API: Per MiniMax plan — check current pricing at minimax.io or minimaxi.com
Tokens par appel: Modest text input; outputs are audio/image/video files
Monétaire: Pay-per-call; cheapest to start with Flash/Hailuo-Mini equivalents
Astuce: Batch short TTS into fewer longer calls when possible; video is the most expensive — cache results.

Sécurité

Permissions, secrets, portée

Portées minimales : MINIMAX_API_KEY + MINIMAX_API_HOST matched to region

Stockage des identifiants : Env vars

Sortie de données : Your text/audio goes to MiniMax's servers in the selected region

Ne jamais accorder : Don't mix global and mainland keys — causes 'Invalid API key'

Cloned voices are powerful; only clone voices you have explicit consent to use.
Generated content may be subject to platform TOS — review before commercial use.

Dépannage

Erreurs courantes et correctifs

Invalid API key

Key and host region must match. Set MINIMAX_API_HOST=https://api.minimax.io for global or https://api.minimaxi.com for mainland.

Vérifier : env | grep MINIMAX

video generation never completes

Poll with query_video_generation up to 10 min; longer-than-expected usually means server load. Retry with a new task if stuck.

voice_clone quality poor

Sample audio must be 10-60s, clean (no music/noise), single speaker.

Rate limit exceeded

MiniMax enforces per-minute and per-day caps. Back off; upgrade plan for higher limits.

Alternatives

MiniMax-MCP vs autres

Alternative	Quand l'utiliser	Compromis
ElevenLabs MCP	You want Western-market-focused TTS with dedicated English voices	No video/image/music in one bundle
OpenAI/Whisper for TTS	You already have OpenAI credits	Fewer voice options than MiniMax

Plus

Ressources

📖 Lire le README officiel sur GitHub

🐙 Voir les issues ouvertes

🔍 Parcourir les 400+ serveurs MCP et Skills