Microphone setup for a room


during building a new house, I want to prepare my setup for voice control. It is pretty easy to fit every room with speakers, that will have enough power. But where I am struggling is how to “record” voice in that room.

Let´s imagine a regular room in dimensions i.e. 5 x 4 meters. What kind of microphone arrays are you guys using here and where are you placing such arrays? Ideally, I want to avoid using non-embedded devices such Google Home or Alexa. My speakers will be built-in into the ceiling and also SBC like Raspberry Pi will be also in the ceiling. So the first choice would be to put something like ReSpeaker also in the middle of the room ceiling. But is it the ideal place? And what is the best microphone array for such a task? ReSpeaker? Something else? On the internet, I found some research, that stated the best mic array is PS 3 Eye. Or my personal experiences from video conferences favoring Jabra Speak 510 USB device.

What also comes to my mind (but is very DIY) is placing the particular microphones from the array all around the room. I.e. for 4 mic array place a mic in every corner of the room. But I haven´t found any processor, that is capable of placing the mics outside of its board.

So any advice here? What kind of HW are you using? Now I am focusing purely on HW.


Years ago I started down the same path and did lots of testing. Ultimately I put heaps of time and effort into getting a system that you had to talk like a robot to get the exact command to work.

My wife bought me a google home mini. I realise you don’t want to use them for whatever reason its totally up to you.

It just works and continues to work. We have a couple around the house and you just ask it to play music from your Spotify account or on the home hub netflix. You can use duo to call them or broadcast a message to a different room.

1 Like

This is it for me. I acknowledge the concerns that some people have about Google and other data collectors, but I’m personally okay with it. Google has had a net-positive impact on my life while never charging me a dime for search, maps, assistant, gmail, and many other services. Nothing in life is free.

Are you also going to put a UPS in every ceiling? If not, you should avoid using SD cards that will quickly get corrupted by power outages.

I would say no. The ceiling will pick up sound from anywhere in the room as it bounces off of walls, but you may get feedback from the speaker audio bouncing off of surfaces in a relatively small space and right back to the mic. That’ll depend somewhat on the furniture, flooring, mic sensitivity, and whatever software you’re using to power all of this.

I would put the speaker and mic next to each other by the main lightswitch. I think that would be more natural than speaking to the ceiling, and make it easier to power and maintain the system. Also, we typically want to talk in the direction of whatever/whomever we’re talking to, and expect the response from the same direction.

A mic in every corner of the room won’t offer much benefit and may have drawbacks. You might also find that some of the mics are blocked by furniture.

I don’t have any advice on hardware, but I think you should keep visibility in mind and have considerations for guests. For example, have a soft LED that’s on when it’s listening, and have a local kill switch to turn off the mic in a room.

If I went to someone’s house and there were microphones everywhere, I wouldn’t feel comfortable at all. I know this is ironic given my professed comfort with Google, but all of my Nest speakers have switches to turn the mics off and are easily unplugged if need be. That’s not quite the same with DIY mics embedded in the walls.

This! Been there, done that, too. Used several ReSpeakers, they do the job but it isn’t about audio HW, it’s ultimately about speech and intent recognition. Tried mycroft, Rhasspy and other stuff.
Ended up with an Alexa and have not regretted since.
Saves you a ton of life time.


Thanks to all for your valuable input.

Got that, but honestly, I would like to avoid sending sound from my house to some cloud. I was looking very closely at Snips project. I really liked its architecture and philosophy. It is a pitty that this project ended. There is a follow-up Project Alice, but haven´t reached stable version yet. In the meantime, you could run Alexa or Google Home on custom hardware.

Not an issue when you are running it in read-only mode and write all changes to RAMdisk. I am already operating multiple devices like that and they´ve never get corrupted.

Can be done by issuing a voice command in a particular room or via the smart switch, that will be deployed in every room.

I will keep searching for a suitable solution for me and post any info if I find anything interesting.

You can try mycroft, it can run local-only on a RPi with a ReSpeaker and has an OH binding (two, actually).

1 Like

Yeah, I am looking in that direction. But another more essential things about the actual house construction have higher priorities. So the plan is to deploy the suitable HW and delay its functionality at a later time. I can “easily” change the entire SW of it, but I definitely don´t want to drill holes in half-year-old walls.

This might be of interest to you.