A self-contained, offline-first voice assistant stack for macOS featuring local speech-to-text and a cyberpunk dashboard.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install maylo-voice-assistant
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 maylo-voice-assistant 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 Maylo Voice Assistant?
Maylo Voice Assistant is a comprehensive, privacy-focused stack designed to transform a macOS device into an intelligent, always-on AI companion. By prioritizing offline-first processing, this skill utilizes local Whisper ASR and openWakeWord detection to ensure that all voice interactions remain on-device. It serves as a prime example of the power of Openclaw Skills by bridging hardware-level audio recording with intelligent agent orchestration.
The system includes a cyberpunk-inspired dashboard that provides visual feedback via waveforms and supports push-to-talk functionality over WebSockets. This allows users to interact with their Mac through voice commands or a remote mobile interface, all while leveraging the native macOS speech synthesis for natural-sounding responses without cloud latency.
Maylo Voice Assistant Use Cases
- Deploying a private, home-hosted AI assistant on a Mac mini.
- Creating hands-free workflows using the "Hey Jarvis" wake word via Openclaw Skills.
- Interacting with local AI agents via a mobile browser on the same Wi-Fi network.
- Building custom voice-controlled automations that function entirely without an internet connection.
How Maylo Voice Assistant Works
- The system initiates an always-on listener using openWakeWord to monitor for the specific activation phrase.
- When the wake word is detected, VAD (Voice Activity Detection) triggers a recording of the user's spoken command.
- The local faster-whisper engine processes the audio to generate a high-accuracy text transcription.
- This transcription is routed through the Openclaw Skills responder bridge to the local agent for processing.
- The agent's text response is converted to audio using the native macOS say command, supporting various system voices.
- The Web UI maintains a real-time connection via WebSocket to stream audio and provide visual status updates.
Maylo Voice Assistant Setup
To get started, first copy the application assets to your preferred working directory:
cp -R assets/app ~/maylo-voice-assistant
Next, initialize the virtual environment and install all necessary dependencies:
scripts/install.sh ~/maylo-voice-assistant
Launch the assistant to begin listening for the wake word:
scripts/run_assistant.sh ~/maylo-voice-assistant
For mobile or remote access, start the HTTPS web interface:
scripts/run_web_https.sh ~/maylo-voice-assistant --host 0.0.0.0 --port 8443
Maylo Voice Assistant Data Schema & Taxonomy
The skill organizes its operations and troubleshooting data within a structured directory format:
| Data Component | Description |
|---|---|
recordings/assistant.log |
Captures wake-word detection events and VAD status. |
recordings/responder.log |
Logs text transcriptions and agent processing logic. |
recordings/web.log |
Monitors HTTPS server requests and WebSocket stream health. |
assets/app/ |
Contains the core binaries for Whisper, wake-word models, and the web frontend. |
Maylo Voice Assistant Advanced Features
- Custom wake-word support for personalized activation phrases.
- Integrated feedback loop mitigation using the MAYLO_POST_SAY_INHIBIT_SEC setting to prevent the assistant from hearing itself.
- Support for self-signed certificates to enable secure microphone access from mobile browsers.
- Local-only execution path ensuring maximum data privacy within the Openclaw Skills ecosystem.
SKILL.md
Loading
Related Openclaw Skills
genviral
A comprehensive API automation tool for creating, scheduling, and analyzing AI-driven video and slideshow content across multiple social platforms.
Genviral Partner API Automation
A comprehensive CLI-driven automation suite for the Genviral Partner API to create, schedule, and analyze multi-platform social media content.
Moltext
A specialized compiler that transforms human-centric web documentation into high-density, AI-native memory context for agents.
Streaming Buddy
An AI-powered personal assistant that tracks your viewing progress and suggests movies or shows based on your unique taste and available streaming services.
Agent Confessions
A social and community-driven skill allowing AI agents to share anonymous, humorous observations about human behavior.
Agent Shield
A decentralized security protocol designed to protect AI agents and their users from malicious skill patterns and supply chain attacks.
Featured*
