(Replying to PARENT post)
1) Their native Home Assistant voice stuff combined with either their ESPHome project or their "voice satellite" software implementations that run on a Raspberry PI (or whatever) with any microphone attached. Either of these provide wildly different results. It turns out you can't just slap a microphone on something and get quality voice and speech recognition from 15 feet away (or even much less) under typical conditions. I learned this the hard way many many years ago.
2) My project Willow[0].
Our approaches could not be more different in terms of implementation and each has advantages and disadvantages. That said, between my 20 years of experience with the intersection of voice, networks, etc and thousands of users using Willow in the real world for six months I certainly believe our approach is substantially better in terms of a user experience competitive with the commercial big tech stuff.
HA voice is essentially at initial release and they will undoubtedly be making improvements but my experience (confirmed by initial user feedback and my own testing) tells me they will probably need a complete re-architecture at some point. Basically, I took my 20 years of experience with this stuff and combined it with "inspiration" from the big tech implementations and made it local and private.
I love Home Assistant and have very deep respect for the team but they clearly don't have much real-world experience with voice (which is a very unique animal). Their architecture and the end result demonstrates that.
I want to be very careful to not sound overly critical or harsh but decades of experience and the resulting accumulated knowledge makes a huge difference. They only just started learning voice and it's very painful process.
[0] - https://heywillow.io/
(Replying to PARENT post)
The first thing you should do is learn to organize your thoughts into coherent questions.
(Replying to PARENT post)
And can I talk to it like Alexa? Half the site seems to say yes, the other half no.
I'm probably going to hold out for something more friendly that inevitably will come out just as Alexa is being lowered into the grave.