Mycroft https://mycroft.ai/
I’ve been waiting to get the v2 model, but it’s taking so long I might make my own with a raspberry pi and their software.
Seems promising, I’ll give it a try on my Raspberry Pi
It still uses proprietary recognition servers…
I thought they wrote their own STT engine? Guess it was a TTS though.
After looking it up I do see what you’re talking about though. They say they proxy everything so your IP doesn’t touch the Google servers. But I also see you can run your own Mozilla DeepSpeech & use that instead, which seems like a decent option. It’s just not the default.
I also see you can run your own Mozilla DeepSpeech & use that instead, which seems like a decent option
Yes, it somewhat works, but it’s super fiddly and you have to provide your own pretrained models etc.
I’ve messed about with it but it doesnt recognise some words with my voice well :P
It seems like Snips was the voice control of choice with HA but was bought up by Sonos.
Rhasspy - https://github.com/synesthesiam/rhasspy - might work (https://dev.to/jeikabu/home-assistant-voice-recognition-with-rhasspy-1jb2).
All depends on you using Home Assistant of course, although looking quickly at the docs Rhasspy seems to work with other systems.
https://jasperproject.github.io/ might also be worth a look.
hehe rhASSpy
Upvoting for the meme
Honestly, I’ve never had a use for a home assistant.
What’s the context behind this poster?
big companies listen to your every word and sell your data to the highest bidder, including intelligence agencies. The posters are from WW2 where allied forces were warned Nazis sent their spies everywhere. In a sense, many modern conglomerates are organised like fascists, so the comparison fits like a glove. Thanks for reading my Ted talk.
there is ProjectAlice, which is based on snips,
and of course Rhasspy
rhasspy plays nicely with i.e. HomeAssistant and/or nodeRed, it works completely offline.
