For Home Assistant Assist, I am looking for a replacement for my Google speaker. I already have the hardware ready on the shelf, including the options of RPI1, RPI2, RPI Zero, and ESP32. I also have a USB conference speaker with a microphone, which, in my opinion, is suitable for this purpose.
My question is, which underlying software should I use for this, and which hardware is the best choice? Currently, I have Whisper and Piper running as add-ons on Home Assistant, and it works reasonably well via my phone.
I am proficient with Linux and can potentially solder a button or something to one of the hardware devices. However, the ideal scenario would be to use a wake word.
ESPHome is our firmware to allow users to easily create devices for their smart home. In Year of the Voice - Chapter 2, we’ve added support for ESPHome to accept voice commands when a user pushes a button.
Today, that support is extended to allow any ESP32 device with an i2s microphone to become a voice satellite for Home Assistant.
Re which software … I have happily used RasPi 3A, 3B and Zero with reSpeaker HATs with Rhasspy and HA almost 2 years - and I am swapping over to HA Voice Assist. Now that WakeWord has been added it does everything I want. If any drawback, is that it’s still fairly new and under active development.
Re: hardware … there are many correct answers to this, and yet also none which is clearly “best”. Currently none match Google’s and Alexa’s devices for cost and sound quality.
I hesitate to recommend people to spend money to buy any of the current hardware options (though the M5 ATOM Echo isn’t too much for someone to later decide to shelve); however since you already have some hardware, I do encourage you to give it a try. Year of the Voice - Chapter 4: Wake words contains links to several options:
plugging your conference speaker directly into your HA box (assuming it is in a convenient location);
Since you already have some RasPi’s you could add USB Mic and speakers (or your conference speaker) with homeassistant-satellite; and HA Voice Assist (Whisper, Piper and openWakeWord) on the Home Assistant server, to get started easily.
IMHO the biggest issue with all these hardware options is that Open Source software for the audio processing is lagging a fair way behind the proprietary developed by the big names. Filtering out background noise (eg from a TV in the same room), Echo Cancellation (subtracting the audio it is playing from the sound the microphone hears, to get what the person is saying), beamforming (using multiple microphones to enhance sound from a particular direction), and a whole bunch of acromyns
As for spending money to buy satellite devices … I personally am waiting to see what comes from the ESP32-S3 direction; with its greater memory and AI instructions. The ESP32-S3-BOX-3 comes close (a really impressive demo) … though it may need Nabu Casa to launch their own Voice Assist Satellite device.
That is such a cool project!!! Unfortunatly he said " This is not being actively maintained, but I’ve released all source code and design files for anyone else to pick up from here."
I’d happily buy these if someone built them and they worked with Assist.
