I started working with the Voice Assistant PE over Christmas and ran into the following two issues that I would like to report:
- Voice Assistant PE becomes unusable, if you disable all Piper services and then re-activate them. You then need to restart it, so that it will work again.
- I am german and noticed that I have to define aliases between “grosses” and “großes” or else some of my entities will not work (I used
ss
in their name, but Whisper (correctly) detectsß
).ß
andss
can be treated equal for all intents and purposes (though they matter for pronounciation) and mapping them to each other should make things more robust.
Regarding the second point and slightly off topic: Where can I find in-depth technical information on what the Voice Assistant processing pipeline currently looks like? (The issue of ß
vs. ss
suggests that it is doing string-matching, which would explain, why it is so brittle… so I want to understand what is being done.)