🎙️ VoiceClaw

OpenClaw voice assistant plugin — wake word detection, speech-to-text, text-to-speech, all as a standard OpenClaw plugin.

Features

• 🗣️ Wake word — configurable trigger (default: "미르야")
• 🎤 STT — SenseVoice (primary), faster-whisper, OpenAI Whisper
• 🔊 TTS — Edge TTS, Piper, macOS say
• 🧠 Agent integration — Voice input goes directly to your OpenClaw agent
• ⚙️ Standard plugin — openclaw.plugin.json manifest, config via openclaw.yaml

Architecture

Node.js (OpenClaw plugin)  ◄── JSON-RPC stdio ──►  Python (audio/ML)
├── index.ts (entry)                                 ├── vad.py (Silero VAD)
├── bridge.ts (IPC)                                  ├── stt.py (SenseVoice/Whisper)
│                                                    ├── tts.py (Edge TTS)
│                                                    └── audio.py (PyAudio)

Installation

# Install as OpenClaw plugin
cd ~/voiceclaw
openclaw plugins install .

# Install Python dependencies
bash scripts/install-python.sh

Usage

# Via chat command
/voiceclaw start
/voiceclaw stop
/voiceclaw status

# Via Gateway RPC
# voiceclaw.status
# voiceclaw.config {action: "get"}

Configuration (openclaw.yaml)

plugins:
  entries:
    voiceclaw:
      enabled: true
      config:
        stt:
          engine: sensevoice
          language: ko
        tts:
          engine: edge-tts
          voice: ko-KR-SunHiNeural
        wakeWord: "미르야"
        vad:
          threshold: 0.5
        autoStart: false
        agentId: main

Troubleshooting

Issue	Solution
No audio input detected	Check System Settings → Privacy → Microphone permissions
PyAudio install fails	Run brew install portaudio first, then retry
Wake word not triggering	Lower vad.threshold (e.g., 0.3) in config
TTS no sound output	Verify audio output device; try say "test" in terminal
Python bridge won't start	Ensure python3 -m pip list shows all deps from requirements.txt

Development

npm install
npm run build    # TypeScript → dist/
npm run dev      # Watch mode

License

MIT © MUIN