A complete offline voice-to-voice AI assistant for OpenClaw using Whisper.cpp and Pocket-TTS for 100% local processing.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install local-voice-agent
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 local-voice-agent 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 Local Voice Agent?
The Local Voice Agent is a robust, privacy-focused solution for users wanting a voice-to-voice interface within the Openclaw Skills ecosystem. By combining Whisper.cpp for high-performance speech-to-text (STT) and Pocket-TTS for text-to-speech (TTS), this skill enables a fully local processing pipeline without reliance on cloud APIs or external costs.
This architecture ensures maximum data privacy and eliminates latency issues associated with internet connectivity. It allows developers and power users to interact with their AI coding agents through natural language commands, making it an essential addition to any developer's Openclaw Skills toolkit for hands-free operations.
Local Voice Agent Use Cases
- Hands-free coding and terminal operation via voice commands.
- Automated daily morning briefings and schedule readouts.
- Creating voice-activated reminders and task notes without typing.
- Improving accessibility for users with mobility or visual constraints.
- Converting existing audio or video files into text transcriptions locally.
How Local Voice Agent Works
- User voice input is captured via the microphone or a provided audio file.
- Whisper.cpp processes the audio to generate an accurate text transcription locally.
- The transcribed text is sent to the OpenClaw AI engine for logic processing and response generation.
- The resulting AI text response is forwarded to the Pocket-TTS server.
- Pocket-TTS converts the text back into natural-sounding speech for the user to hear.
Local Voice Agent Setup
To integrate this into your collection of Openclaw Skills, follow these installation steps:
1. Install Whisper.cpp (STT)
git clone https://github.com/ggerganov/whisper.cpp ~/.local/whisper.cpp
cd ~/.local/whisper.cpp
make -j4
bash ./models/download-ggml-model.sh tiny
2. Configure Pocket-TTS (TTS)
# Start your local server
python3 -m app.main --host 0.0.0.0 --port 5000
export POCKET_TTS_URL="http://localhost:5000"
3. Install Dependencies
sudo apt-get install -y ffmpeg
Local Voice Agent Data Schema & Taxonomy
The skill manages audio processing and configurations through a structured YAML system and API responses:
| Component | Configuration File | Key Parameters |
|---|---|---|
| STT | config/voices.yaml |
model (tiny, small, medium), language |
| TTS | config/voices.yaml |
url, voice selection, output format (wav, mp3) |
| Performance | config/voices.yaml |
CPU threads, realtime processing toggle |
API responses for voice commands return a JSON object containing the transcription, the raw response text, and the file path to the generated audio response.
Local Voice Agent Advanced Features
- Interactive mode for continuous, fluid voice-to-voice conversations.
- Custom voice cloning and selection via the Pocket-TTS integration.
- Multi-format audio support (WAV, OGG, MP4) with automatic FFmpeg conversion.
- Performance optimization using adjustable CPU thread allocation for faster transcription.
- REST API endpoints for triggering voice commands via external webhooks or scripts.
SKILL.md
Loading
Related Openclaw Skills
PySpark to Excel
Effortlessly converts ASCII-style PySpark table outputs into clean Tab-separated text for direct pasting into spreadsheets.
Sedentary Reminder
An AI skill for designing intelligent sedentary reminder systems that prioritize helpful timing and context-awareness over mechanical repetition.
NVIDIA NeMo Agent Toolkit (NAT)
A versatile framework for building, evaluating, and deploying enterprise-grade AI agents with seamless tool and data integration.
Alibaba Cloud WAF Quick Showcase
A specialized skill for automating the deployment of Alibaba Cloud Web Application Firewall (WAF) to protect web applications running on ECS instances.
volc-docs: Volcengine Documentation Agent
A specialized tool for AI agents to search and retrieve comprehensive official documentation, APIs, and SDK guides from Volcengine.
Excel Power Tool
A comprehensive Excel management and AI-powered analysis tool designed for OpenClaw agents.
Featured*
