Installing the Hass.io add-on from the synesthesiam repo should automatically download/install the profiles for all of the supported languages during the Docker build process. Every language should have its own profile directory under profiles, like profiles/en, profiles/nl, etc.
All of the files you mentioned except espeak_phonemes.txt and custom_words.txt are generated during the training process. custom_words.txt is where you add your own pronunciations outside of base_dictionary.txt.
Now espeak_phonemes.txt is where I really need help. This is very language specific. It maps the pocketsphinx phonemes (from the table) to what eSpeak uses for pronouncing words. This is only useful when adding custom words, since it helps you hear how Rhasspy expects you to pronounce a word. We can figure this out later â after we get past this initial hump of getting everything installed and running.
I did it that way but only had the en profile available in the webinterface. Also when I changed the configuration to default_profile": "nl" I got errors that some things were not found and the add-on crashed.
I read somewhere on Github to clone the repo locally, so I assumed this was the way to do it for languages other then English.
Are you telling me to install docker or the docker client version on my Pi?
âI think the easiest way going forward for the moment would be to install Docker on your Pi, but then use run-venv.sh as is.â
But then you say the docker client version is actually not going to work on the Pi?
âDocker client image of Rhasspy is only supported on amd64 machinesâ
Am I still supposed to use the âcreate_venv.shâ-script to install anything?
I meant to just install the Docker program itself. Rhasspy needs some external tools to run, which can either be installed or run as individual little Docker images.
I just updated the create-venv.sh script to be a little smarter. If it finds docker, it goes ahead and uses the Docker images for all of the external tools. Otherwise, it goes out to Internet and installs all of the tools locally. Please give it a try and let me know how it goes.
Looks like youâre close to getting this working! In your Settings, I see that I must have left in my Rhasspy client/server configuration accidentally (pumpkin.lan is my home server). You should probably set it to âDo speech recognition on this deviceâ (also set âDo intent recognition on this deviceâ).
For the aplay error, please try the Download button instead of Pronounce. It will give you a WAV file instead of trying to play it out of the speakers. Docker seems very hard to configure for playing/recording sound. Thatâs what the /dev/snd and privileged settings are for, but they donât seem to work so well with PulseAudio (only ALSA).
Command â[âaplayâ, â-tâ, âwavâ, â/tmp/tmpkndjpi4j.wavâ]â returned non-zero exit status 1.
I still have this error
and not for equipment selection
Yesterday I also saw rhasspy reading and speaking
OK, my guess is that this is an issue with Docker and /dev/snd. If you can, please take a look at the log in the Rhasspy Docker container and see if thereâs more information around the aplay error.
Can you post the docker-compose configuration you got to work also?
Lastly, can you verify that running aplay some-file.wav works in a terminal on your NUC? Replace some-file.wav with the name of any WAV file you have.