2.5 pre-release: I can't get wake word working with german profile

Help
1

Hi,
I’m running 2.5 pre-release with german profile in docker on a raspberry pi 3B+ with a PS3 eye. I’ve a working setup for recording, sound output, speech to text, text 2 speech. The problem that I’m having right now is to get a wake word working. I’ve tried snowboy (with various models), porcupine and pocketsphinx with no success at all.

Am I missing something obvious?

Best regards
Mirko

2

I´ve got the same problem with german profile and installation via docker
Does your log show any errors?

3

I even thought, “snowboy” wasn’t working for me, but when I increase the sensitivity, it is working.
Perhaps try “snowboy” again and for testing set “sensitivity” to 0.9 instead of 0.5.
Then you can decrease the sensitivity in small steps, until you get the best false/positive results.
Seems as if it depends on the microfone, you are using

4

No errors in the log.

5

No luck with the higher sensitivity either.

6

Maybe try the English profile for example? I mean, it´s a satellite, so you have the STT and so on on a server and for wake word doesn’t it matter, if it’s English or German.

7

Thank you for your post. I was not working with a master/satellite setup, but now I am :slight_smile:

After setting the profile to english on the satellite, the different wake-word-engines are working as expected.