I now have 3 assist satelites and was wondering what the roadmap looks like. SInce they are preview I am not sure if I should invest more into the next pieces I need (to replace my Echos) or wait for the final release.
Will there be hardware changes? Is there a timetable for that next release? Any place I can follow those developments?
The hardware probably isn’t preview. There is likely demand for more expensive hardware with better noise reduction. But I doubt there will be better hardware at $50 in the next year or two.
What improves STT substantially is using a large whisper model. Whisper does not improve wake word reliability, of course. But I feel that many people are blaming the Voice device for what is mostly a problem with using a too weak whisper variant.
Wolfgang I suggest you (re)read the announcement of the Preview Edition, where Mike did explain the reason for the name.
My understanding is pretty similar to dzmiller - that the current Preview Edition is acceptably close to the big boys hardware; and for now focussing on the software is better use of resources.
There are several highly desirable software techniques we could use - but all the current implementations are proprietary. Mike is constantly looking out for Open Source projects which can be included in Voice Assist, and the Micro Wakeword is an excellent example of how much difference the right extra resource can make.
As dkmiller states, there is also effort being put into improving the models our current software is using; particularly broadening Voice Assist’s usability for non-english languages.