I noticed immediately that Voice PE has an issue with ambient sound (for me radio playing).
I am really enthusiastic about Voice PE, I can ask questions that are feeling so natural I am amazed (I use Google gemini) that it works out of the box. From closing covers to setting lights (without remembering the right device names and stuff, is truly amazing. I notice that I even ask follow up questions as it feels natural ti just keep talking, which do not work obviously. The problem is apparently ambient sound, I noticed that response times were getting longer and I could not explain why. It turns out to be the difference as the radio playing. In my debug log I see that my question is perfectly understood but than it picks up people talking on the radio and it goes haywire. In general it still does what I ask from it but it takes a lot of time to process the command and strip it fromm the radio bs that is playing.
Yes, I’m having the same problem. Voice is basically unusable when the radio is on or other people are having a conversation in the same room (both using faster-whisper and the Nabu Casa Cloud stt).
I imagine some logic that stops voice recognition when the volume difference is big or (best case) when a different person is speaking would be the “real” solution.
A low - tech approach could be to have a stop-word. Something as simple as “okay nabu /command/ thanks /discard everything here/” maybe?
Yes, I’m having the same problem. In my case it’s the TV. When the TV volume is at its normal volume (for my family), the PE only hears what’s on the TV…kinda funny looking at the log to see what it heard, no wonder it can’t process it! I have to mute the TV or turn the volume down really low before the PE will work. But, when it does, it works very well.