Hi everyone,
I’d like to share a project I’ve been working on over the past two months, which is now complete. It builds on the onju voice project — I’ve designed a custom PCB for the Google Home Mini that adds ESP32-based wake word detection and Bluetooth A2DP audio streaming.
Key Features:
Wake word detection is handled locally on the ESP32 (no cloud required).
When a voice command is detected, it’s streamed via Bluetooth.
If no voice interaction is ongoing, the system goes into sleep mode to conserve power.
While idle/sleep, it functions as a Bluetooth A2DP speaker for music or media playback.
The goal was to create a local, low-power, and flexible voice interface — ideal for offline or privacy-focused smart environments.
I’m currently waiting for native support in OpenHAB for ESP32 audio streaming and voice detection, which would allow full integration of this DIY voice assistant (e.g., for triggering rules, controlling items, etc.).
If anyone is exploring similar ideas, has tips, or knows of workarounds — I’d love to connect and exchange thoughts!