#Setting up my new voice preview

1 messages · Page 1 of 1 (latest)

tranquil compass
#

My home assistant voice preview just arrived! I’m following the installation instructions, but the home assistant app is stuck at “scanning for servers”. The installation instructions don’t mention this. Do I need to setup a server on my home network first? I thought the device was self-contained, so I wouldn’t need a raspberry pi or anything alongside it.

plain scroll
#

The device is basically only a microphone and speaker - you will need a dedicated device for home assistant

paper tusk
#

And most likely some sort of LLM, either paid or locally hosted, if you want more than basic controls.

tranquil compass
#

Got it. Perhaps I read the announcement too quickly, as I got the impression it was a plug-and-play replacement for commercial voice assistants. I’ll do some more research to look for recommended specs and buy a device for a server. Thanks for confirming.

paper tusk
tranquil compass
paper tusk
tranquil compass
#

Okay thanks, that's helpful

#

And if I do graduate to hardware, it looks like the green is pretty comparable with the yellow; the main difference being that the yellow could allow me to upgrade the memory and storage, and is matter-ready, whereas the green requires a usb dongle add-on if I need matter compatibility. Have I got that right, or are there other advantages to the yellow?

dry wave
#

The green is supposed to be the easiest entry into the homeassistant eco system offering a low power solution that is very easy to setup. The yellow offers a couple more customization options with slightly more powerful hardware and additional radios.

#

Neither is a great solution if you intend to use the voice pe with local processing as they arent all that powerful. But the yellow would be the better choice of them both for this usecase in terms of performance.

tranquil compass
#

Interesting. I don’t suppose there are minimum specs listed anywhere for local processing?

dry wave
#

The website calls it "powerful" hardware which is probably a rather accurate description. Theoretically the tools for speech to text and text to speech support smallish models that can run somewhat on a pi 4 with reasonable memory size (eq to the yellow).

#

Personally if you were to go with the yellow, I would recommend picking the compute model 5 (pi 5).
It would be better to pick some more capable compute like a mini pc with an N100 cpu or similar or at best something with a dedicated nvidia gpu. But that is a lot more pricy

tranquil compass
#

Ah okay, yeah I assumed a gpu would be needed for any reasonable performance

#

But all right, that’s really helpful, thank you

#

I’ got it going on my (extremely noisy) old MacBook last night, and wired up ChatGPT to it (which I appreciate is very much not local processing), and it was amazing how quickly it blew Alexa out of the water with just that feature! Looking forward to playing around more with it.

dry wave
#

It is quite nice yea

#

Have a look at customizing the entities exposed to assistant as that can be used to set up proper aliases and limit those to only the ones important to you. Makes it better and faster

tranquil compass
#

I’ve got to actually buy some smart home devices first 😬

dry wave
#

smart plugs and bulbs bring all the fun of power monitoring and cozy vibes

#

everyone loves a good graph ^^

tranquil compass
#

I’ve got a couple smart plugs but that’s about it. Mainly wanted an excuse to get Amazon’s microphone out of my flat!

#

But yeah will have to look around - sounds like there are a ton of non-silly sensors and other home IoT things that have popped up in the last few ones

dry wave
#

there is a lot you can dig through with homeassistant

#

(the voice things can be used to trigger custom automations which can do pretty much anything so it can have endless customization options ^^)