Hello everyone, I wanted to ask if it’s possible to modify my two Alexa Echo devices ( gen 3 and 4) to replace the microphone with the M5 Stack Voice Atom. The idea is to make the Alexa speakers “dumb” and have the Atom handle the voice commands to connect to Home Assistant and manage the devices. This way, I’ll have a local home automation system without relying on the Alexa cloud. Additionally, if my question isn’t related to any of the home’s sensors, I’d like the Gemini Google add-on, previously installed in Home Assistant, to be able to provide an answer. Is what I’m describing possible? Are there any modding diagrams or solutions for what I want to achieve? Thank you very much
Have you heard of HAVPE?
i replaced my alexas with this. It equivalent to a 1st gen Alexa in terms of sound quality.
You can replace alexa internals but in the end your better off buying one of the two devices listed above. You can’t just tack in a esp device and use the Echos speakers and mic. You basically replace all internals, maybe reuse speaker, and get to have an echo case for your hacked together esp device. simply not worth it. projects doing this have 3rd party boards with $70+ USD price tag.
The HAVPE and Waveshare device (and they have other nice units) can utilize HA’s local voice assistant which is slowly improving. The cloud version works pretty well.
Hi tmjpugh,
Yes, I’ve heard about HAVPE, but not about Waveshare. It seems really awesome. So, just for my understanding, can I replace my two Alexas with two Waveshare? Do I need the HAVPE too, or is it possible to flash TTS voice assistant to Waveshare only?
When you give voice commands, Home Assistant manages them internally, especially for home sensors. However, for more technical queries unrelated to home sensors, it leverages Google Gemini instead. Is it possible ?
I don’t want to oversell HAs voice capabilities.
I think it’s similar to 2015/2016 Alexa. Not bad but not great. Announcements are perfect but wake word is hit or miss. Just like when Alexa came out, it’s worth trying for $20USD/each.
Waveshare makes the S3
HAVPE is made by nabu casa
Both are basically esp chip/mic/amp/speaker. That makes a voice device flash able with esphome and useable with HAs voice pipeline/integration
The device connects with HA and HA handles actions: listen / speak / play music / announce and so on
Local only no LLM
Local with your local LLM
Local w/ remote LLM
Nabu remote
Local w/ Nabu
These are possibilities for control. Pretty much any combination
Actually local only is not bad and local with nabu is even better.
Since it’s esphome it can be extended to do other things or process actions on local device seperate from HA.
Music assistant allows music streaming and music groups
I can’t say is better than Alexa but I can’t say it’s worse. It’s different. Truth be told Alexa was bad at recognizing a lot of command as well even in 2025 so I may be harshly judging