Home assistant is working on a voice assistant but it’s not done yet. They’re supposed to announce progress on a YouTube live stream on the 12th.
October 12th? Wow that is great news!
From the release notes for October:
Even more exciting news, as this is Home Assistant’s year of the voice: Chapter 4 of this endeavor has been announced and will be streamed live on YouTube on 12 October 2023, at 12:00 PDT / 21:00 CEST! Be sure to tune in! I’ve seen some early demos… All I’m going to say is: It is mind-blowing! 🤯
Live stream link is here: https://www.youtube.com/watch?v=YzgYYkOrnhQ
Here is an alternative Piped link(s):
https://www.piped.video/watch?v=YzgYYkOrnhQ
Piped is a privacy-respecting open-source alternative frontend to YouTube.
I’m open-source; check me out at GitHub.
Will be curious to see if they ever add transient timers for commands. Google will happily turn off the lights in 5 minutes as a command. In Home Assistant I’d have to go in, create a timer helper, add the light to turn off…
I can imagine eventually but I don’t think that’ll be their prioritized focus for a bit, cause it doesn’t seem like it’d have a massive useage case
Mycroft is the closest thing I can think off.
☹️ Mycroft ran out of money and closed down. I think you can still order a Mark2, but the cloud components they were hosting were taken down which essentially breaks the software. There is another company that picked up the ball and took over support but it’s a more limited experience at the moment and the whole ecosystem is in a bit of a rut. The OpenVoiceOS project that spun out of Mycroft looks like the best bet long-term but they haven’t finished a stable release yet. I bought a Mark2, which I still think is the best hardware available for an open source voice assistant, but finding and stitching together the software to make it work is a chore atm.
OVOS & Neon are MyCroft’s successor. They work well, and can even plug into LLMs
deleted by creator
As others have commented, the open source home assistant project can take voice commands and perform smart home functions like turning lights on and off, reading off the forecast, taking down notes, etc etc. But it does have limits, you will have to script any kind of complex commands, like pulling headlines from an RSS feed, or playing spotify playlists, or really anything that requires fetching info from an API, it won’t do those kinds of things out of the box.
The other factor which others have called out is that it doesn’t currently handle wake word functionality, though that’s been on their road map this year and the Oct update might fix that. That being said, running a dedicated wake word app to fill in that gap is very much possible. See my thread here for more info: https://community.home-assistant.io/t/setting-up-a-100-local-smart-speaker-on-an-android-tablet-using-tasker-and-snowboy-to-handle-wake-word-detection/611435
https://github.com/toverainc/willow
They even have a dedicated piece of hardware for it.
