Wyoming Voice Pipeline

For everyone

Enables hands-free, private voice assistant control throughout the space. Speak wake words like 'Hey Jarvis' or 'OK Nabu' to voice satellites to trigger automations, adjust lights, or ask questions with instant, natural-sounding voice replies—100% private, offline, and with zero cloud dependency.

What you can do

• •Process wake words ('Hey Jarvis', 'OK Nabu') locally with real-time edge detection
• •Transcribe voice commands using Whisper STT (Faster-Whisper small-int8)
• •Synthesize natural-sounding voice responses with Piper TTS in milliseconds
• •Interface with ESP32 voice satellites (e.g. Home Assistant Voice PE, Atom Echo)
• •Trigger Home Assistant routines, ambient lighting scenes, and smart home scripts hands-free
• •100% offline audio processing ensures no spoken audio or telemetry ever leaves the local network

For admins

Orchestrates three dedicated Wyoming microservices interfacing directly with Home Assistant: Wyoming Whisper for speech-to-text (STT) running the small-int8 model, Wyoming Piper for neural text-to-speech (TTS) synthesis using en_US-lessac-medium, and Wyoming OpenWakeWord for continuous edge wake word detection without streaming audio to external servers.