A privacy-first skill for macOS providing local speech recognition and synthesis without external API dependencies.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install macos-local-voice
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install macos-local-voice using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is macOS Local Voice?
This skill empowers AI agents to communicate through audio directly on Apple hardware. By leveraging native macOS capabilities like the Speech.framework and the say command, Openclaw Skills users can transcribe audio and generate high-quality voice output entirely offline. It is designed for developers who prioritize privacy and low latency, as it removes the need for cloud-based STT/TTS services.
The toolset includes specialized scripts for transcribing various audio formats and synthesizing speech with smart voice selection. By integrating this into Openclaw Skills, agents gain the ability to handle voice-based workflows including voice note processing and spoken feedback without incurring API costs or sending data to third-party servers.
macOS Local Voice Use Cases
- Transcribing user voice notes into text for agent processing.
- Generating natural-sounding voice responses for multi-modal agent interactions.
- Automating localized speech synthesis across multiple languages including English, Chinese, and Spanish.
- Sending voice notes in messaging platforms using Openclaw Skills integrations.
- Building accessibility tools that interact with macOS native spoken content.
How macOS Local Voice Works
- The agent receives an audio file (ogg, m4a, mp3, etc.) and invokes the STT script using the yap CLI to extract text.
- The system identifies the appropriate locale to ensure high transcription accuracy.
- For outbound communication, the agent selects the best available system voice using the voice management script to check for premium quality levels.
- Text is converted to audio via the native macOS say command.
- If ffmpeg is present, the audio is encoded into an optimized ogg/opus format for efficient storage and transmission.
- The resulting audio path is returned for the agent to use in messaging or further automation tasks.
macOS Local Voice Setup
Ensure you are running on macOS. Apple Silicon is recommended for the best performance. Install the necessary dependencies via Homebrew:
# Install the required yap CLI for speech-to-text
brew install finnvoor/tools/yap
# Install ffmpeg for optimized audio encoding (ogg/opus support)
brew install ffmpeg
You can then verify your voice environment using the included utility scripts:
# List all supported locales
node scripts/stt.mjs --locales
# Find the highest quality voice for a specific locale
node scripts/voices.mjs best en_US
macOS Local Voice Data Schema & Taxonomy
The skill manages audio data and voice metadata with a structured approach:
| Component | Details |
|---|---|
| Input Formats | Supports ogg, m4a, mp3, wav via yap CLI |
| Output Path | Defaults to ~/.openclaw/media/outbound/ with timestamped filenames |
| Output Formats | ogg/opus (if ffmpeg is available) or aiff (native fallback) |
| Quality Levels | 1 (Compact), 2 (Enhanced), 3 (Premium) |
| Locales | Standard BCP-47 tags (e.g., en_US, zh_CN, fr_FR) |
macOS Local Voice Advanced Features
- Smart voice selection that automatically identifies the highest quality premium voice available for a specific language.
- Integrated voice readiness checks to prevent silent fallbacks to low-quality default system voices.
- Multi-language support covering over 15 major locales including Mandarin, Japanese, and Russian.
- Seamless integration with the Openclaw Skills messaging tool for sending generated speech as native voice notes.
- Local processing ensures 100% data residency and offline capability for secure environments.
SKILL.md
Loading
Related Openclaw Skills
Outbound Prospecting
A sophisticated AI agent workflow for conducting deep company intelligence and generating high-conversion outbound email drafts.
Browser Control Tool
A Python-based utility for macOS that enables one-click URL opening in browsers like Safari and Chrome.
SearXNG Web Search
A privacy-focused metasearch skill that enables AI agents to query the web via a self-hosted SearXNG instance without the need for individual engine API keys.
OpenClaw TikTok Downloader
A professional-grade TikTok downloader and metadata extractor powered by yt-dlp for automated content archiving.
Yandex SpeechKit STT
A robust speech-to-text integration that leverages Yandex SpeechKit to transcribe audio files for AI coding agents.
Crypto Portfolio Tracker API
A comprehensive API and CLI tool for monitoring cryptocurrency portfolios with real-time valuation and profit/loss analysis.
METADATA
Featured*
