Look for recommendations for a microphone inputs for Home Assistant voice assistants for those of us who are not super network/programming savvy. Are there any good guides that would make finding a microphone to connect to a raspberry pi for a voice assistant pretty easy?
1 Like
I went with the Jabra Speak 410 USB speaker/microphone combo. Initially, quality of voice recognition was spotty, and I suspected it was the microphone. Switched my back-end speech-to-text engine from CMU pocketsphinx to Kaldi, and accuracy/quality is now about 98-to-99% solid.
SeeedStudios ReSpeaker multi-microphone devices are also quite popular.
For an inexpensive but reasonably accurate choice, the PS3 Eye Microphone is a good choice.
1 Like
I wasn’t able to get the ps3 eye camera to work with ada and home assistant. I would be very curious if anyone has actually got it working with Home Assistant(specifically Hassio.)
I’m using the Seeed Studios USB 4-Microphone array now, and voice recognition is better still. Check out this thread at the Rhasspy forum for details.
Ok, anyone else considering something opensource to replace an Alexa/Siri/Google Home whole house voice recognition approach? I have 6 echos currently, but don’t want them sitting on a counter and would prefer just a “dumb” microphone array in each room.
I’m building a home now, and thinking about future proofing (as best I can). One of the things I’m thinking about is how to pre-wire for such microphones in each room (and there will be a lot of rooms), so that I can try out different voice recognition approaches. Any thoughts? @FredTheFrog, since you’ve worked with a few different options… or anyone else… what would be a good strategy to run the right wires to a spot in the ceiling or wall of every room. Extending USB seems expensive. Most wiring is relatively cheap to do while the walls are open but USB isn’t really feasible.
FYI - I’m going to have a server rack in an equipment closet where my HA server will be. I’ll probably go with Ubuntu running on a pretty beefy machine so I can do things like facial and voice recognition, but that doesn’t mean I want to spend hundreds of dollars per room just to replace an echo dot in each.
1 Like
I’m currently using a pair of Pi3B+ in FLIR aluminum cases as Rhasspy client/satellite systems. If I had my preference, these would be Pi Zero W units instead, simply to reduce the footprint. The upcoming 2.5 release of Rhasspy should make using the Pi Zero W a bit easier. Of course, it still requires a power supply, even if it doesn’t use a Cat-6 Ethernet connection.
As for wiring/placement, I would have done it differently, twenty years ago, if only I had today’s hindsight. I recommend installing an RJ-45 jack on a minimum of two opposing walls in every room ( even the kitchen and dining room, but NOT the bathrooms 
) If you can, drop TWO cables down to every box, and install boxes on all four walls. You never know when a cable may become damaged. Now, whether you place those RJ-45 boxes in the center of a wall, or nearer the corners, is completely up to you.
If your walls are open and accessible right now, definitely wire two cables to each of four boxes in every room. You’ll never regret it. Once the insulation and sheet rock are installed, it’s a LOT more difficult.
2 Likes
To power the Pi’s I would use a PoE splitter like this one, so you only need to run a single cable to each Pi.
2 Likes
Thank you both. One more question… Webby not just run mic wire for fancy conference room mics and connect them to the server in the equipment closet? That Jabra mic you were using @FredTheFrog, that was USB, were you still plugging it into a Pi?
Yes, it’s the 410, the smallest of their conference/speakers, but the USB cable provides plenty of length for convenient placement with the PI out of view.
Every single answer is USB, which means something like a Raspberry Pi with each mic. I know I’m asking for something that doesn’t exist yet, but what is technically stopping someone (bigger company) from coming up with a PoW or powered WiFi microphone that just integrates with Home Assistant? Is there a technical reason this can’t happen? (In before “use an echo”… because I’m not doing the cloud, and there are tons of us like that.”)
I just see the Pi/USB/Mic setup as clunky and cumbersome. If voice is ever going to take off in Home Assistant, that is to say, be widely adopted, there will need to be something like this.
1 Like
Did you ever find a wireless option?
Wireless (analog) as in bluetooth or radio frequency ? Or wireless (digital) as in wi-fi or zigbee ?
Either way, your mic is going to need power as well - especially if it is listening 24/7 - and a certain amount of electronics/logic.
USB mics are an obvious solution because you can buy them off-the-shelf, plug them into any off-the-shelf computer, and Docker will generally pass them through without requiring specific drivers.
I agree that RasPi is expensive and overkill for a voice satellite - but Rhasspy can do wakeword detection on the satellite.
The ESP32-S3-BOX-3 comes pretty close … except key parts of its software are closed source.
I am holding my breath hoping Nabu Casa leverages its ESPHome expertise to bring a ESP32-S3 voice satellite to market at a competitive cost and using the S3’s built-in AI capability to do the listening for wakeword locally.
I think you answered your own question. Amazon & Google have shown that it’s technically possible.
But companies need to make a profit, and Alexa/Google do that by being closed systems locking customers in. Even so, they spent a lot developing their hardware and proprietary software to such a high standard; and have been subsidising their hardware prices to get customers onboard. It is just not economic to compete with companies with such deep pockets.
Maybe they will start charging for their server service to recover running costs; and hopefully Open Source alternatives to the audio processing magic will come to fruition … allowing other options to become economic. I’m sure Nabu Casa are keeping a close watch on developments 
Has anyone seen any cheaper options? I can do just a mic if needed, or a simple speaker mic combo.
This one was too pricy.
Also, to pricey.
I don’t want a camera. 
Thanks in advance!
Kia ora. I’m also wanting to have speakers/mics throughout the house so we can speak to the house and get it to do things that Siri generally does (only we don’t want to use Siri).
Could we use POE SIP speakers like GSC35X6? Ideally these are “always on” with a wake word but the current docs they have for SIP requires you to “call” home assistant. Also I assume the audio quality of something like this is poor. If I’m gonna play music I’ll want something decent.
Another key thing is being able to recognize what room the request comes in from, and only playing back in that room.
For the USB microphone options, do you also need an external compute ( Rasp Pi) or can you plug a Jabra directly into a Home Assistant Yellow?
I have a question. I am just getting started and I used to have a band, so I have a ton of mixers, microphones, good microphones, and speakers. Can I run everything through the single line in and line out on an onboard soundcard?
Another option would be a USB/Wifi repeater. But will hold out for something off-the-shelf that works well. With the mass production of so many smart speakers you’d think it would be easy to get the parts and and replace the interface with ESP Home. I’ve been pulling my hair out getting Alexa to work with Phillips Hue scenes and took about 5 minutes with HA.