MiniMax API

Use the MiniMax API via direct curl calls for AI chat completion, text-to-speech, and video generation.

Official docs: https://platform.minimax.io/docs

---

When to Use

Use this skill when you need to:

• Chat completion with Chinese-optimized LLM (MiniMax-M1/M2)
• Text-to-speech with natural voices and emotion control
• Video generation from text prompts (T2V)
• Image-to-video conversion (I2V)

---

Prerequisites

1. Sign up at MiniMax Platform
2. Go to Account Management > API Keys to create an API key
3. Note: Global users should use api.minimaxi.chat (with extra "i")

export MINIMAX_API_KEY="your-api-key"

API Hosts

Region	Base URL
China	https://api.minimax.io
Global	https://api.minimaxi.chat

---

How to Use

All examples below assume you have MINIMAX_API_KEY 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_API_KEY" -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_API_KEY" -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_API_KEY" -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_API_KEY" -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_API_KEY" -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_API_KEY" -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_API_KEY" -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_API_KEY" -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_API_KEY" -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-H