I own a ha yellow and a home assistant voice. (preview edition)
Only it's so slow, even a question: what is the time takes 6-8 seconds for a response.
lights on for the livingroom takes for ever, (it feels for ever) So everyone is walking to the lightswitch.
If i see some demo;s on youtube the respons is < 1 second.
Is there something i can do or change to make it faster?
#very slow voice response
1 messages · Page 1 of 1 (latest)
For the local experience you will need some significantly more powerful hardware (or pick a lighter model if available). Alternatively you can use the cloud variant with a nabu subscription
so it is a hardware problem, moving to a mini pc with a intel i7 will do the trick?
just to be sure, is the:
Beelink EQ14 Pro Mini PC, Intel Prozessor N150
a good candidate as a home assistant pc or do you recommend something else?
or:
Beelink Mini PC S12 PRO Windows 11 Pro, Intel N100
Yes, it's better in almost everything, except the voice part. If you would like to run it locally, you'd have to get a decent gpu and lots of ram as far as i know.
The best option for current hardware (also for beelink etc.) would be the advice from Horstexplorer above, especially the nabu casa cloud part > voice does not need hardware power and responses within a second or two. (at least for me) You might try nabu casa for 1 month, it's free to try.
i have nabu casa, and i must say that it is slow in resonse and often wrong.
what is the weather tomorrow? - there is no device called tomorror or something like that
i just asked again:
okay nabu what's the weather tomorrow
- i can.t give you the answer, look at a weather website or use a weather app <-- well that is usefull info 🙁
thats another thing. Turn on light (locally processed) takes a second..
it just an example, i wish it takes a second
local automtion, about 18 seconds orso?
checked network setup? including wifi channel etc?
everything is working as it should (and after: the automation excuted, well, nothing happened) automation according to the log, was fired 3 hours ago
Can't really complain about either the voice experience on nabu nor local. Both work quick and reliable for me, even tho I don't enjoy the available voices for my language
The included actions do not support everything you could ask it for. If you need more details you gotta either use an LLM with additional data sources, create custom automations or if you need better detection of intents you will need to extend them with your preferred speech patterns
The thing done by either nabu or local services is converting your speech to text and text to speech. Evaluation of what should be done is what your instance does