minimax

SKILL.md

Troubleshooting

If requests fail, run zero doctor check-connector --env-name MINIMAX_TOKEN or zero doctor check-connector --url https://api.minimax.io/v1/text/chatcompletion_v2 --method POST

How to Use

All examples below assume you have MINIMAX_TOKEN set.

Authentication uses Bearer token in the Authorization header.

1. Basic Chat Completion

Send a chat message:

Write to /tmp/minimax_request.json:
{
  "model": "MiniMax-Text-01",
  "messages": [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Hello, who are you?"}
  ]
}

Then run:

curl -s "https://api.minimax.io/v1/text/chatcompletion_v2" -X POST -H "Authorization: Bearer $MINIMAX_TOKEN" -H "Content-Type: application/json" -d @/tmp/minimax_request.json | jq '.choices[0].message.content'

Available models:

MiniMax-M2: Reasoning model (best quality)

MiniMax-M1: Reasoning model (balanced)

MiniMax-Text-01: Standard model (fastest)

2. Chat with Temperature Control

Adjust creativity:

Write to /tmp/minimax_request.json:

{
  "model": "MiniMax-Text-01",
  "messages": [
    {"role": "user", "content": "Write a short poem about AI."}
  ],
  "temperature": 0.7,
  "max_tokens": 200
}

Then run:

curl -s "https://api.minimax.io/v1/text/chatcompletion_v2" -X POST -H "Authorization: Bearer $MINIMAX_TOKEN" -H "Content-Type: application/json" -d @/tmp/minimax_request.json | jq '.choices[0].message.content'

Parameters:

temperature (0-1): Higher = more creative

top_p (0-1, default 0.95): Sampling diversity

max_tokens: Maximum output tokens

3. Streaming Response

Get real-time output:

Write to /tmp/minimax_request.json:

{
  "model": "MiniMax-M1",
  "messages": [
    {"role": "user", "content": "Explain quantum computing."}
  ],
  "stream": true
}

Then run:

curl -s "https://api.minimax.io/v1/text/chatcompletion_v2" -X POST -H "Authorization: Bearer $MINIMAX_TOKEN" -H "Content-Type: application/json" -d @/tmp/minimax_request.json

Streaming is recommended for reasoning models (M1/M2).

4. Reasoning Model (M1/M2)

Use reasoning models for complex tasks:

Write to /tmp/minimax_request.json:

{
  "model": "MiniMax-M1",
  "messages": [
    {"role": "user", "content": "Solve step by step: A train travels 120km in 2 hours. What is its average speed in m/s?"}
  ],
  "stream": true
}

Then run:

curl -s "https://api.minimax.io/v1/text/chatcompletion_v2" -X POST -H "Authorization: Bearer $MINIMAX_TOKEN" -H "Content-Type: application/json" -d @/tmp/minimax_request.json

Response includes reasoning_content field with thought process.

5. Text-to-Speech (Basic)

Convert text to speech:

Write to /tmp/minimax_request.json:

{
  "model": "speech-02-hd",
  "text": "Hello, this is a test of MiniMax text to speech.",
  "voice_id": "male-qn-qingse",
  "speed": 1.0,
  "format": "mp3"
}

Then run:

curl -s "https://api.minimax.io/v1/t2a_v2" -X POST -H "Authorization: Bearer $MINIMAX_TOKEN" -H "Content-Type: application/json" -d @/tmp/minimax_request.json --output speech.mp3

6. TTS with Emotion

Add emotion to speech (speech-02 models):

Write to /tmp/minimax_request.json:

{
  "model": "speech-02-hd",
  "text": "I am so happy to meet you today!",
  "voice_id": "female-shaonv",
  "emotion": "happy",
  "speed": 1.0,
  "format": "mp3"
}

Then run:

curl -s "https://api.minimax.io/v1/t2a_v2" -X POST -H "Authorization: Bearer $MINIMAX_TOKEN" -H "Content-Type: application/json" -d @/tmp/minimax_request.json --output happy_speech.mp3

Emotion options: happy, sad, angry, fearful, disgusted, surprised, neutral

7. TTS with Audio Settings

Fine-tune audio output:

Write to /tmp/minimax_request.json:

{
  "model": "speech-02-hd",
  "text": "High quality audio test.",
  "voice_id": "male-qn-qingse",
  "speed": 1.0,
  "vol": 1.0,
  "pitch": 0,
  "audio_sample_rate": 32000,
  "bitrate": 128000,
  "format": "mp3"
}

Then run:

curl -s "https://api.minimax.io/v1/t2a_v2" -X POST -H "Authorization: Bearer $MINIMAX_TOKEN" -H "Content-Type: application/json" -d @/tmp/minimax_request.json --output hq_speech.mp3

TTS models:

speech-02-hd: High definition (best quality)

speech-02-turbo: Fast generation

speech-01-hd: Previous gen HD

speech-01-turbo: Previous gen fast

8. Text-to-Video (T2V)

Generate video from text prompt:

Write to /tmp/minimax_request.json:

{
  "model": "T2V-01-Director",
  "prompt": "A cat playing with a ball of yarn [Static shot].",
  "duration": 6,
  "resolution": "1080P"
}

Then run:

curl -s "https://api.minimax.io/v1/video_generation" -X POST -H "Authorization: Bearer $MINIMAX_TOKEN" -H "Content-Type: application/json" -d @/tmp/minimax_request.json | jq '.task_id'

Video generation is async - returns a task ID to poll for completion.

9. T2V with Camera Control

Control camera movement in videos:

Write to /tmp/minimax_request.json:

{
  "model": "MiniMax-Hailuo-2.3",
  "prompt": "A person walking through a forest [Tracking shot], then stops to look at a bird [Push in].",
  "duration": 6,
  "resolution": "1080P"
}

Then run:

curl -s "https://api.minimax.io/v1/video_generation" -X POST -H "Authorization: Bearer $MINIMAX_TOKEN" -H "Content-Type: application/json" -d @/tmp/minimax_request.json | jq '.task_id'

Camera commands (in brackets):

Movement: Truck left/right, Pan left/right, Push in/Pull out

Vertical: Pedestal up/down, Tilt up/down

Zoom: Zoom in/out

Special: Shake, Tracking shot, Static shot

Combine with [Pan left, Pedestal up] (max 3 simultaneous).

10. Image-to-Video (I2V)

Generate video from an image:

Note: For I2V, use MiniMax-Hailuo-2.3 or S2V-01 model which supports first_frame_image. The T2V-01-Director model is text-to-video only.

Write to /tmp/minimax_request.json:

{
  "model": "MiniMax-Hailuo-2.3",
  "prompt": "The scene comes to life with gentle movement [Static shot].",
  "first_frame_image": "https://example.com/image.jpg",
  "duration": 6,
  "resolution": "1080P"
}

Then run:

curl -s "https://api.minimax.io/v1/video_generation" -X POST -H "Authorization: Bearer $MINIMAX_TOKEN" -H "Content-Type: application/json" -d @/tmp/minimax_request.json | jq '.task_id'

Provide first_frame_image as URL or base64-encoded image.

11. Function Calling (Tools)

Use tools with chat:

Write to /tmp/minimax_request.json:

{
  "model": "MiniMax-Text-01",
  "messages": [
    {"role": "user", "content": "What is the weather in Beijing?"}
  ],
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "get_weather",
        "description": "Get weather for a location",
        "parameters": {
          "type": "object",
          "properties": {
            "location": {"type": "string", "description": "City name"}
          },
          "required": ["location"]
        }
      }
    }
  ],
  "tool_choice": "auto"
}

Then run:

curl -s "https://api.minimax.io/v1/text/chatcompletion_v2" -X POST -H "Authorization: Bearer $MINIMAX_TOKEN" -H "Content-Type: application/json" -d @/tmp/minimax_request.json | jq '.choices[0]'

Response Format

Chat Completion

{
  "id": "string",
  "choices": [{
  "message": {
  "role": "assistant",
  "content": "Response text",
  "reasoning_content": "Thought process (M1/M2 only)"
  },
  "finish_reason": "stop"
  }],
  "usage": {
  "prompt_tokens": 10,
  "completion_tokens": 50,
  "total_tokens": 60
  }
}

Guidelines

Use correct host: China uses api.minimax.io, global uses api.minimaxi.chat

Streaming for reasoning: M1/M2 models work best with stream: true

Camera syntax: Video commands go in [brackets] within prompts

Emotion in TTS: Only works with speech-02-* and speech-01-* models

Async video: Video generation returns task ID - poll for completion

Chinese optimized: MiniMax excels at Chinese language tasks