Voice PE stops listening very quickly after activation, have to speak in one quick unbroken sentence

Hi all; I just set up my Voice PE and I’m having trouble using it, because although it wakes correctly every time to the wake word, it stops listening almost immediately. So if I say “Hey Mycroft” I basically have to say the command as one continuous string of sound, because if I so much as pause it hangs up. What’s terrible about this is that I have to keep talking then through the wake chime, during which it hears somewhat poorly and so almost always mishears what I’m saying, if it hears it at all. It does hear me though, if I speak continuously, so it’s not like it isn’t working at all.

Given there’s supposed to be a “never mind” command to make it stop listening, I feel like this has to be something configured incorrectly, yet I don’t see a configuration setting for it anywhere. Help?

edit: I have tried changing the “finish speaking detection”; it seems to do nothing.

Ok, I narrowed this down some; it’s actually specifically a problem with the “hey mycroft” activation word. It seems like it wakes and begins listening immediately, but still catches some part of the word “mycroft” as the beginning of the command, causing it to stop listening very quickly. If I change it to “hey jarvis” the problem goes away immediately, and everything works great. It’s also far better at understanding me, I’m guessing since it’s no longer picking up part of what it thinks is a word at the beginning of the sentence.

It is doubtful that this is a wake word sound. After it, an audible signal is triggered, followed by the listening phase.
You can obtain accurate information about what is happening by recording samples. Troubleshooting Assist - Home Assistant

Also, if you are still using the stable version of the firmware, this may be related to this PR. Repeated false activation terminates the session.
Stay tuned for updates.

Or install the latest version manually https://support.nabucasa.com/hc/en-us/articles/25800241218717-Reinstalling-the-firmware

1 Like

Thanks for the reply; I am using the stable version, yes. The description in the PR you linked sounds like exactly the problem I was describing, although it’s strange I can work around it by speaking more quickly after the wake word; since it was an issue with just that one wake word though, I’m simply using a different one for now.

Is it recommended to use the unstable firmware? The phrase “still using” implies to me a bit that many do not.

I just want to add, I’m experiencing this too & my wake work is also Hey Mycroft. I believe it started for me after the last update. My pipeline is configured to use HA Cloud for Speech-to-text, Text-to-speech and Home Assistant for the Conversation agent.