Well I got my S3-BOX-3, spent ages configuring it and added some custom sentences. not once does it recognize them
I’m using faster-whisper, piper (en-GB) and openwakework OK nabu
The cloud stuff works great, but the whole point of doing this was to use local processing
What am i missing?
Official HA hardware is coming on the 19th
Better be good cuz I deleted Amazon when voice came out and really need replacement hardware
Is your hardware up to the task?
I don’t currently use voice but I am planning on spending the next couple of months getting voice up and running. However I expect mediocre at best results from my own hardware because this stuff is so dam GPU intensive and my N100 based mini PC uses an integrated GPU. The people who are getting this to work fully locally are investing in multi GPU servers that handle voice and only voice.
And are you absolutely possitive that your set up is good? Have you shown your set up to anyone and asked if they can spot where you might be going wrong with your configuration?
No problem here. The S3 box works great within the limits of assist. But you are scares on details about what you do, how you do it and what your expectations are in the first place.
You say you did a lot of configuration but if you just install the voice image on it it will work straight away. Especially in English.
Have you tried the assist trouble shoot tools to see where it fails and what the intents are it recognise
Someone mentioned on this forum that there are S3-box clones with a lousy mic.
There’s not much to the setup
Configure it in voice assistant section (very few choices there)
S3 Box is garbage IMO
Recently added a Jabra speaker phone to the usb port on server (N100 16GB) and it’s considerably better
Hopefully the xmos stuff the HA hardware will be a big help (design is awful IMO: it actually looks like a 20 year old iPod)
It just took a Long time (relatively speaking) to learn and setup all the voice assistant parts and configure the S3 box. Wasted a LOT of time trying to install software on the S3 (turned out to be a bad usb cable)
I’m using recommended hardware S3 box 3. And it’s just terrible, slow inaccurate and terrible speaker
I confirmed a bunch of custom sentences and it barely recognizes any of them (they work fine when I type the sentence)
Can’t wait to get the HA hardware with the xmos, better speaker and ability to add external speakers.