Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Version 0.0.3 - Added more usage triggers (e.g. "Say <text>", "Reply with voice message ...") for broader activation. - Expanded documentation with agent instructions for handling and sending the generated audio files as voice messages. - Updated wording and formatting in documentation for clarity and consistency. - Removed the version field from SKILL.md metadata for simplified versioning.
What This Skill Does
Runs text-to-speech on Apple Silicon using MLX and open-source models (default QWen3-TTS). Generates audio files locally with no API key or external server needed. Works only on macOS with Apple Silicon.
Runs entirely on-device with no API costs or rate limits, unlike cloud TTS services.
When to use it
- Converting written notes to audio for hands-free review
- Generating voice messages to send in chat
- Listening to long text without reading
- Creating audio drafts from written scripts
- Narrating content during prototyping or demos
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Say hello world
- 1Receives text input from the user
- 2Runs mlx-tts.sh with the text as argument
- 3Receives the output file path from the script
- 4Sends the audio file to the user as a voice message
An audio file delivered as a voice message in the conversation
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.