voice-assistant

What This Skill Does

The OpenClaw Voice Assistant is a sophisticated Windows companion that bridges your microphone with the OpenClaw agent ecosystem. It utilizes high-performance local processing for speech detection and wake-word recognition, ensuring privacy and low latency. By integrating Porcupine for wake-word detection, faster-whisper for accurate speech-to-text, and ElevenLabs for realistic voice synthesis, this skill transforms OpenClaw from a text-based interface into a responsive, conversational entity. It includes advanced features like mic suppression to avoid feedback loops during playback and a system tray interface for easy management, making it an ideal choice for users who want a hands-free interactive experience with their local AI.

Installation

To get started, navigate to the scripts folder within your OpenClaw directory. Initialize a Python virtual environment using python -m venv venv and install the necessary dependencies via pip install -r requirements.txt. Copy the .env.example file to .env and populate it with your Picovoice access key, ElevenLabs API credentials, and your OpenClaw gateway connection details. Once configured, you can launch the application directly or use start.bat to run it as a background process with a system tray icon. Ensure your microphone is correctly configured in your Windows sound settings for optimal detection.

Use Cases

This skill is perfect for users building a custom home-automation dashboard, voice-controlled development assistants, or those who prefer voice dictation for managing complex tasks. It is especially useful for long-form reasoning sessions where you want to "talk" to your computer while working on other physical tasks, or for integrating OpenClaw into a smart home environment where a screen-less, conversational interface is required.

Example Prompts

1. "OpenClaw, what is the status of my local file indexing task?"
2. "Hey OpenClaw, can you summarize the current contents of the technical documentation folder and suggest any missing sections?"
3. "OpenClaw, please draft a code snippet to handle WebSocket connections and add it to my current project directory."

Tips & Limitations

To maximize performance, ensure your silence threshold (SILENCE_TIMEOUT) is tuned to your environment's noise floor; a higher value is better for quiet rooms, while lower is better for active offices. The assistant currently requires an internet connection for ElevenLabs TTS and the OpenClaw gateway. Be aware that running the larger Whisper models will increase memory usage significantly. We recommend testing with the 'base' model before upgrading to 'medium' or 'large' to see if it meets your hardware constraints.