assistant-x-openclaw-jarvis-voice-assistant
Installation
SKILL.md
Assistant-X-OpenClaw: JARVIS Voice Assistant
Skill by ara.so — Hermes Skills collection.
A multi-role AI voice assistant inspired by Iron Man's JARVIS. Runs local ASR (Automatic Speech Recognition) and TTS (Text-to-Speech) using sherpa-onnx, connects to LLMs via OpenClaw Gateway, supports multiple character roles, voice wake words, continuous dialogue, speaker verification, and Flutter-based HUD visual effects.
What It Does
- Multi-role support: Built-in Jarvis and Lin Meimei characters with independent wake words, TTS voices, sound effects, and HUD animations
- Local ASR/TTS: SenseVoice/Zipformer for speech recognition, Piper/VITS/MeloTTS for synthesis
- OpenClaw integration: Each assistant role maps to an OpenClaw Agent via WebSocket device pairing
- Wake word detection: sherpa-onnx KWS with custom wake words per role
- Continuous dialogue: Supports multi-turn conversations with 30s idle timeout
- Interruption: Wake word can interrupt ongoing TTS playback
- Speaker verification: Gradual embedding updates during conversation, forced verification on wake
- HUD overlay: Flutter transparent window with rotating ring animations and audio level visualization
- Remote control: HTTP API for status and shutdown