
Mimo V2 5 Tts
Converts text to speech using Xiaomi's MiMo V2.5 models with three modes: preset voices, voice design from text descriptions, and voice cloning from audio samples. The preset mode supports singing and comes with eight built-in voices like 冰糖 (lively girl) and 白桦 (mature male). You can control emotion and style either through natural language prompts or inline tags like (紧张,深呼吸) for nervous breathing. The voice design mode is interesting because it lets you write character descriptions ("middle-aged male, auctioneer style, rapid rhythm") and generates matching voices. Supports Chinese dialects, mixed emotions, and director mode for detailed performance control. Requires a MiMo API key and Python with openai installed. The documentation is surprisingly thorough on how to write effective voice descriptions and when to add emotional tags versus full context.
- skills.sh installs
- 664
- GitHub stars
- 103
- last updated
- Aug 21, 2026
Install
$ npx -y skills add xiaomimimo/mimo-skills --skill mimo-v2-5-tts --agent claude-codeInstalls into .claude/skills of the current project.
Files
MiMo V2.5 TTS
使用小米 MiMo V2.5 TTS 系列模型生成语音。支持中英文、预置音色、音色设计、音色克隆、情绪风格、方言、唱歌。
脚本目录:$SKILLS_PATH/mimo-v2-5-tts/scripts/
$SKILLS_PATH说明: skills 目录路径,因部署环境而异。
模型选择
V2.5 系列提供三种模型,根据使用场景选择:
| 模型 ID | 用途 | 音色来源 | 特殊能力 |
|---|---|---|---|
mimo-v2.5-tts | 预置音色语音合成 | 内置精品音色 | 支持唱歌 |
mimo-v2.5-tts-voicedesign | 文本描述定制音色 | 文本描述生成 | — |
mimo-v2.5-tts-voiceclone | 音频样本复刻音色 | 音频样本 | — |
选择建议:
- 需要快速生成语音、需要唱歌功能 →
mimo-v2.5-tts(预置音色) - 需要独特音色 →
mimo-v2.5-tts-voicedesign(文本描述生成) - 需要模仿特定声音 →
mimo-v2.5-tts-voiceclone(音频样本复刻)
注意: TTS 有随机性,同样输入的效果可能不同,用户有需要时可以多生成几次以供挑选。
环境依赖
| 环境变量 | 说明 | 必需 |
|---|---|---|
MIMO_API_KEY | MiMo API 密钥(MiMo 开放平台获取) | 是 |
Sponsors
Details
- Repository
- github.com/xiaomimimo/mimo-skills
- Category
- Generative Media
- First seen
- Jun 3, 2026