Description

Onju Voice - AI assistant replacement to Google Nest Mini

READ THIS FIRST:
This isn't a finished product, nor will I be making it one. I'm sharing this as many were interested in hacking on this further - if you're not comfortable tinkering on software, firmware & electronics but expect a final product, best to pass on this. Before thinking of ordering make sure you're comfortable running the server code, tweaking the firmware, and with the fact that there may be unseen issues. That said, I have validated a batch from PCBWay using the files included here, and this is the most mature version. The design files are on github and you're free to order from somewhere else, but for a datapoint every other build I did (from that other PCB company) had issues that needed debugging and re-work.
----

An ESP32-S3 based board, designed to be a drop-in replacement to the PCB in a Google Nest Mini (2nd gen), or used in one's own custom enclosure, to create a powerful, privacy-focused and expressive home voice assistant.

All functionality of the Nest Mini is replicated in hardware, such as Wifi, Bluetooth, capacitive touch pads, mute switch and 6 RGB LED's. It has a debug ribbon cable that can pull out of the device, as well as a USB so that anyone with a Arduino IDE can program it. (this only uses 2 Mic's whereas the original uses 3, as the ESP32-S3's audio framework currently only supports 2 for BSS)

I chose the Google Nest Mini as a starting point due to the relatively low cost of a beautifully designed shell with custom colors, and a quality speaker. From my survey's, this was the most owned (and unused) smart home assistant that was sent out free to many users, is well stocked on eBay and selling for <$40 at some big-box retailers. (cheaper than an "ESP32 Box" yet much better design)

While the ESP32-S3 has powerful audio framework capabilities, the ideal use case is VAD and sending the audio to a local server that has your credentials, logins, logic, "wakewords", even your own LLM such as LLaMa2 or TTS model if you desire, else calls to OpenAI & ElevenLabs etc. This offloads the need for custom logic on each device and centralizes your logic and credentials to a single secure device on your home network.