Xiaomi MiMo-V2-TTS Skill

Text-to-speech using Xiaomi's MiMo-V2-TTS model. Supports emotional style control, Chinese dialects (Northeastern/Sichuan/Cantonese/Taiwanese), role-playing voices, and singing synthesis.

When to Use

• User asks to convert text to speech / audio
• User mentions "read aloud", "TTS", "voice synthesis", "narrate"
• User wants specific voice styles, emotions, or dialects

API Details

• Platform: https://platform.xiaomimimo.com
• Base URL: https://api.xiaomimimo.com/v1
• Endpoint: /v1/chat/completions (NOT /audio/speech)
• Model: mimo-v2-tts
• Auth: Bearer Token via MIMO_API_KEY env var

Important: API Format

MiMo TTS uses the Chat Completions endpoint with special requirements:

• ❌ No system role allowed (returns error)
• ✅ Must include assistant role message (the text to synthesize)
• user message = style/voice instructions
• assistant message = text to be spoken
• Response: choices[0].message.audio.data contains base64-encoded audio

Usage

python3 <skill_dir>/scripts/mimo_tts.py \
  --text "Hello, world!" \
  --output /tmp/openclaw/tts_output.mp3 \
  [--style "cheerful tone"] \
  [--speed 1.0] \
  [--format mp3] \
  [--api-key YOUR_KEY]

Set MIMO_API_KEY environment variable or pass --api-key.

Parameters

Style Control Examples

• Dialects: --style "speak in Cantonese" / "Sichuan dialect" / "Taiwanese accent"
• Emotions: --style "happy and excited" / "sad and gentle" / "start happy then turn melancholic"
• Characters: --style "news anchor" / "gentle older sister"
• Singing: --style "sing it"
• Combined: --style "Northeastern dialect, enthusiastic and bold"

Notes

• Pricing: Free during launch period (March 2026), may charge later
• Supports Chinese and English text
• Best results with Chinese text and style descriptions in Chinese
• Get API key at https://platform.xiaomimimo.com