I got some Atom Echos, configured them, and they work! I even customized my own wakeword and it worked on the first try. Thanks, Home Assistant team, for such an awesome product as Home Assistant and for fantastic documentation.
Though the Echos and voice recognition works, I’m waiting about 28 seconds between speaking and having Home Assistant respond. “OK Nabu, do the thing”…then I wait ~28 seconds and then at the same time I hear the Echo say “Done” and Home Assistant responds.
Is the delay due to the Echos being small/cheap/slow processors? They react instantly to the wakeword, but perhaps that requires less processing power because it’s trained. Is the delay due to forwarding the audio content of my spoken word over the network to Home Assistant so Whisper can process it? I’m able to transfer other content over my network very quickly, and I doubt the data size of a few spoken words is very large. Is the delay in Whisper processing my spoken command?
What has your experience been with the Echos and openwakeword?
Do you get that kind of delay when you talk to the assistant on your phone or through a browser? It might be whisper taking a while to process it?
What is your home assistant running on?
Also, what’s the range/mic/audio quality like on those atom echos? I’m thinking of looking into something like that now that a decent voice assistant is a possibility.
Not the OP, but commenting on the Atoms. They are good for testing, but not sure I’d want to use them for a full setup.
For one, speakers are pretty rough in them 😂 That’s kinda to be expected though, it’s intended as a cheap dev device.
However, the bigger thing for me is to wait and see what hardware HA will support when they implement on-device wake-word processing. I’d definitely prefer no continuous audio streaming over the network, until after I have said the wake word.
Good questions. I haven’t talked to the assistant through the browser or phone yet -that’s a good way to help narrow down what process might be causing delay.
I’m running HAOS in proxmox on a mini PC with a celeron. A couple people have said they’re using beefy hardware, so I might need a new box.
I don’t yet know the range of these Echoes, but they seem to do a great job listening. They also have a speaker but it sounds super wuiet, not really useful. If I want a verbal response I’ll have to push it through other speakers via an automation.