audio-reply
由 matrixy 提供
Speech & Transcription简介
Generate audio replies using TTS.
描述
该技能的功能
Converts Claude responses to spoken audio using the MLX Audio chatterbox-turbo TTS model running locally on macOS. Accepts a public URL to fetch and read aloud, or generates a conversational spoken reply for any topic. Trigger phrases include "read it to me", "talk to me", "voice reply", and "say it".
Runs TTS locally on macOS with no API key or cloud service, keeping audio generation private and offline.
何时使用
- Listening to a blog post or article while away from the screen
- Hands-free answers during cooking or exercise
- Getting an audio summary of a long documentation page
- Hearing a conversational explanation of a technical topic
- Reading out a webpage in the background during a commute
示例工作流
User asks: read it to me https://blog.example.com/new-feature
- 1
Validate URL against safety rules, rejecting local or private targets
- 2
Fetch page content with WebFetch and extract main article text
- 3
Summarize content if it exceeds 500 words
- 4
Run mlx_audio.tts.generate with chatterbox-turbo to produce and auto-play audio
- 5
Delete the temporary .wav file after playback
Audio plays through system speakers; agent confirms "Done reading the article to you."
前置要求
- uv installed (brew install uv)
安装
clawhub install matrixy/audio-reply-skill运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。