Voice control via HA server on windows tablet

I’ve installed home assistant on an old Microsoft surface 3 windows tablet using virtualbox
can I use it as a home assistant voice preview edition ?
I mean it has built in mic and speaker so it should be possible

I tried installing assist microphone add-on and configed it but it doesn’t do anything when for example I say ok nabu turn off aqua light
I should mention I can turn on /off that same light with my android phone so I assume needed add-ons are working fine
this is the debug for my android phone command

with assist microphone addon when I say ok nabu turn off aqua light nothing happens and it doesn’t even make an assistant debug

please guide me what could be the problem
and tell me if there are any other way to use my tablet as voice assistant
oh, I also tried GLaSSIST but it’s so laggy and 90% of times I get “assistant did not respond”

On ha voice there is a service in background continuously listening for wake word

HA UI does not do this

You would need to code

Listen for wakeword
Wakeword detect >> start assist

EDIT

It’s possible to do this but not practical for anyone who doesn’t code as part of daily life

M5stack Atom Echo is under $19USD or so shipped and would give equal result

2 Likes

thanks, I’m new to this world, could you please guide me how I can do that?

No you misunderstood… Short version. It’s not practical… The equivalent of Glassist IS what you’d need to code. Not exactly practical. Use Glassist or find a different solution.

2 Likes

So it sometimes work? I will make this “turn off animations and all useless shit consuming resources” button until weekend propably

1 Like

I saw this youtube video, they guy did that using a usb mic conected to the raspi server, it seemed simple so I thought I could do the same with my virtualbox on a windows tablet

I live in iran and it’s not available here and because of sanctions I can’t order from international online shop

same guy has video on how to make Voice Assistant w/ On-Device Wake Word Detection using ESP32 S3
please check the video out and tell me if it’s doable
I can buy the parts he used

do you know any alternatives to Glassist ?

Didn’t look at video but yes

Esp + mic + speaker and you have voice assistant. That’s Atom Echo

HA voice PE adds dual mic and some processor to improve voice processing

Again
Didn’t look at video but yes. You can make anything into voice assistant with HA if you have time/effort

In your case it may be worth time/effort so if you wanna give it a go please do as others may benefit from your hard work

EDIT

That’s because wake word detection is hard with simple mic. Audio processing hardware is added to most “real” voice assistants to improve this and they are still imperfect.

1 Like

it worked afew times on tablet, I’m not sure if I changed anything but now I keep getting assistant did not respond. I unintstalled and installed 2-3 different versions but with no luck

debug assist on tablet:


another atempt:

and same version works great on my main pc. (HA server is on the tablet

I’m confused, you mean on raspi + usb mic and like a few click in 10mins it’s possible to have voice assistant with HA but on my windows tablet I need to know expert codeing and It’s not practical?

I’m going to try if I don’t find a simpler solution

No.

Time will vary and I don’t do large coding project.

I’m saying

I would build esp first since it is simple and maybe 20 minute project

For RasPi, Linux, or windows. My next go to would be look at existing wake word project and utilize them for wake word detection. My expectation is HA has API for Voice Assistant and I only need to detect wake word and start HA voice assistant on the desired device. If wake word is detected I expect I can utilize existing HA services to complete remainder. You can also consider start with button or other method that may be more reliable and easier to create.

It already possible to speak command to Homeassistant by pushing mic button in UI. This works well and without issue. My goal would be create function to push that button, not create complete voice chain.

On a PC. No.

There are several ways to do this. For example, you can run the latest go2rtc build, start the server for openwakeword, and use the configuration from the example. go2rtc/internal/wyoming at master · AlexxIT/go2rtc · GitHub
It works, but it is in the early stages of development.
But the best option would be a separate device on esp32S3.

2 Likes