mimo-tts

name	mimo-tts
description	Convert text to speech using Xiaomi MiMo TTS — preset voices, voice design from text descriptions, and voice cloning from audio samples. Supports style control via natural language and audio tags, dialects, and singing mode. Use this skill when the user asks to: generate speech from text, create a voiceover, clone a voice, design a custom voice, read text aloud, convert text to audio, or generate singing vocals. Also trigger on "TTS", "text to speech", "speech synthesis", "voice clone", "voice design", "voiceover", "read aloud".
license	MIT
metadata	{"version":"2.2","category":"ai-multimodal","sources":["https://platform.xiaomimimo.com/docs/zh-CN/usage-guide/speech-synthesis-v2.5"]}

MiMo TTS

Prerequisites

Requires mcp__mimo-multimodal__tts tool. If not available, read references/setup.md and help the user configure the MCP server — you will need to ask for their API plan type (token plan vs standard API) and credentials.

Model Selection

Choose the right model based on the user's need:

User wants TTS
  │
  ├─ Has an audio sample to clone? ──→ mimo-v2.5-tts-voiceclone
  │   (provide mp3/wav sample, max 10MB Base64)
  │
  ├─ Describes a voice in words? ──→ mimo-v2.5-tts-voicedesign
  │   (e.g. "young woman, warm and confident, slow pace")
  │
  └─ Wants a ready-made voice? ──→ mimo-v2.5-tts
      (pick from preset voice list below)

Preset Voices (mimo-v2.5-tts only)

Voice	ID	Language	Gender
MiMo Default	mimo_default	Auto	—
冰糖	冰糖	Chinese	Female
茉莉	茉莉	Chinese	Female
苏打	苏打	Chinese	Male
白桦	白桦	Chinese	Male
Mia	Mia	English	Female
Chloe	Chloe	English	Female
Milo	Milo	English	Male
Dean	Dean	English	Male

Style Control

Two methods, placed in different message locations:

Natural language (in user message) — describe the desired style:

"Bright, bouncy, slightly sing-song tone — like you are bursting with good news. Fast pace, rising pitch at the end."

Director Mode — for high-quality performance, describe: Character (identity, personality), Scene (what's happening, emotion), Direction (speed, breath, pauses, resonance, texture).

Audio tags (in assistant message text):

Opening style tags at the beginning: (style1 style2)text to speak

Category	Examples
Emotions	开心/悲伤/愤怒/兴奋/平静/冷漠, 怅然/欣慰/无奈/愧疚
Tone	温柔/高冷/活泼/严肃/慵懒/俏皮/深沉
Voice	磁性/醇厚/清亮/空灵/稚嫩/甜美/沙哑
Character	夹子音/御姐音/正太音/大叔音/台湾腔
Dialects	东北话/四川话/河南话/粤语
Singing	唱歌 — place at very beginning: (唱歌)lyrics

Inline tags anywhere in text: [吸气] [叹气] [笑] [大笑] [抽泣] [颤抖] [变调] [破音] [气声]

Workflow

Preset Voice TTS

1. User provides text and optionally voice/style preference
2. Call tts tool with text, voice ID, and optional style description
3. Return the generated audio

Voice Design TTS

1. User describes the desired voice (gender, age, timbre, emotion, speed)
2. Call tts tool with model mimo-v2.5-tts-voicedesign, voice description in user message
3. Optional: optimize_text_preview: true to polish the text automatically

Voice Clone TTS

1. User provides an audio sample (mp3/wav) and text to speak
2. Call tts tool with model mimo-v2.5-tts-voiceclone, sample as Base64 in voice field

Output Formats

Format	Use Case
wav	Complete audio file (non-streaming)
pcm16	Streaming — 24kHz PCM16LE mono

Important Notes

• The text to synthesize goes in role: assistant content, NOT user
• The user message is for style instructions only — its content will NOT be spoken
• For mimo-v2.5-tts-voicedesign, the user message is required
• TTS is currently free (limited-time promotional pricing)

Error Handling

• 401 Unauthorized: API key is invalid or expired — ask user to check their key
• 413 Payload Too Large: Voice clone sample exceeds 10MB Base64
• Connection timeout: Check if MIMO_API_BASE URL is correct for the user's plan
• Empty audio response: Check that text is in role: assistant (not user)

For direct API calls (without MCP), see references/api-examples.md.