Hi Community,
I would like to slowly start change my voice assistant at home from Alexa to Home Assistant.
But I have a problem with the settings. I can choose a text to speech engine but no speech to text engine
#Having problems with voice assistant. Can’t choose speech-to-text.
nova silo
glass glen
think you have to install whisper for that
might try nabu casa cloud 1 month for free, think you'd like the voice part 😉
nova silo
I have installed whisper. But I can’t choose it. There is no option. It is only showing „none“.
glass glen
done a reboot of haos?
nova silo
Yes
Have I configured it correctly?
This is what is showed when I try to choose whisper.
nova silo
wyoming?
glass glen
yup
nova silo
Okay I will try it
glass glen
find it in integrations
it will ask for host and port > ip of haos and 10300
or host is local .. long time ago
nova silo
Does not work with haos ip and port 10300.
I had a look at the whisper settings. There is an option „show deactivated ports“
Than 10300 is showed. But I didn’t deactivated any port.
nova silo
I have installed Wyoming with piper month ago. I remember it as it shows up in my integrations. But I can’t add whisper to Wyoming.
nova silo
back in a moment and a beer
Cheers
I had a quick look at the piper integration that is discovered and configured correctly on port 10200. There is the „deactivated ports“ also.
This is Piper
glass glen
have to admit, i havent used those in years, since i got nabu casa all of that was not needed anymore
Maybe other users can hop in here
nova silo
Thanks anyway for your help.
I got it working. Found a thread where „core-whisper“ as hostname was recommended.
And that worked for me!
The only thing is now that the voice isn’t recognised. I have an iPhone and the home assistant app. I granted the microphone permission.
Have to figure out what’s wrong.
glass glen
make sure your languages are setup right. Also it can be that voice-commands not working because of http usage. Without secure connection (https) mic is blocked as far as i know.
nova silo
I got it working now.
But speech to text takes about 7 sec to process my voice command.
It’s really slow 😖
I host my home assistant instance on a proxmox server running on an minisforum hm80
I also get this warning when I start whisper server.
[07:59:35] WARNING: Your CPU does not support Advanced Vector Extensions (AVX). Whisper will run slower than normal.
nova silo
I changed to „host“ cpu settings. Now it’s about 3 seconds. Can I speed up voice recognition anyhow
glass glen
use nabu casa cloud 😉
left tendon
I changed to „host“ cpu settings. Now it’s about 3 seconds. Can I speed up voice...
as @glass glen stated. NC cloud is not a bad option.
if you have a system with a GPU you could run whisper with GPU acceleration on that system instead. and get HA to use it across the network.
I use Onnx-ASR with the parakeet models instead of whisper but this only really works with english.
nova silo
Think I will try NC Cloud
left tendon
Think I will try NC Cloud
NC cloud has the other advantage of supporting the project too 🙂
glass glen
and encrypted online backups, so if the fan drops in the sh*t, you're saved