I had some similar issues with all of my voice recognition / responses being very slow... I am running on home assistant supervised on top of Proxmox. What solved the issue for me was changing the CPU type to 'host' in Proxmox. I guess the recognition may rely on some CPU extensions being available.
this post was submitted on 09 Nov 2023
34 points (97.2% liked)
homeassistant
14211 readers
4 users here now
Home Assistant is open source home automation that puts local control and privacy first. Powered by a worldwide community of tinkerers and DIY enthusiasts. Perfect to run on a Raspberry Pi or a local server. Available for free at home-assistant.io
founded 2 years ago
MODERATORS
Good to know!I'm running HAOS on proxmox, too. I'll look into CPU types.
I picked up an Atom for testing as well. I do get a delay, but definitely not 28 seconds... Maybe more like 4-5?
It's totally dependent on the speed of the hardware you are running HA on though. Since the microphone is just streaming sound to the server, which then processes it. I'm running on fairly beefy server.
Just for a test, you could also try toggling off the wake word in the Atom Device settings in HA. Then you can send a command by pushing the physical button. Could at least narrow the delay down to wake word processing vs normal speech processing that way?
Excellent troubleshooting tip. Thanks.
I changed the VM's CPU type in Proxmox and gave the VM more resources (most of the hosts's RAM and CPU cores) and the delays cut in half to around 16 seconds. So I know what's causing my delay (or probably most of it). I guess I need a beefier box.
Holy crap been looking everywhere for a smart speaker to replace Google Home/Nest speakers. This is the first time hearing Atom Echo and its integration. Definitely going to check this out. Thank you!
Check HA Devs videos called "year of the voice" to know everything you have missed. Also atom echo isn't really a smart speaker, more a micro with a possible always listening mode
super cool! I just really wish the android assistant could use wake words.
I got an atom echo 2 months ago and was so happy when the wale word was added as before having to press the button and having it tethered defeated the purpose in my opinion.
Using the echo I make sure it's in the same room as me (on a USB portable charger) and I get maybe a second or two delay. My issue has been setting up custom sentences to control a media player volume, followed their example in the docs for a media player but no errors and no action.
wondering if all this investment they're doing on voice is going to have diminishing returns, like, the whole pitch for HA cloud integration is that you could see and use your HA devices with your cloud enabled voice assistant services, right? Nabu Casa makes money for the HA project. is getting rid of the middle man worth it? I needed some first hand experience to convince myself about getting into it or just simply keep paying for NC
(not first hand experience but) Nabu casa is a way to support HA development, having cloud assistant is just a bonus, you also get easy external access and another bonus is that you don't need to setup the different addons to be able to use assist, while it is doable on something like raspberry 4, problem will be latency compare to nabu casa or you will need a more powerful device to be able to run them smooth. If you didn't bother to setup cloud assistant by yourself to avoid fees, chances are you will stick to nabu casa and it is a good thing as you are supporting home assistant development.
Just bought an Atom Echo. Can’t wait for it to arrive so that I can give Janet commands.