OS: macOS Monterey 12.4 for development, Windows 10 for data collection
PsychoPy version: v2022.2.4
What are you trying to achieve?: Install the German language model for pocketsphinx located here.
What did you try to make it work?:
- Downloaded files
- Extracted cmusphinx-de-voxforge-5.2.tar.gz
- Copied files in “model_parameters/voxforge.cd_cont_6000” to “/Applications/PsychoPy.app/Contents/Resources/lib/python3.8/pocketsphinx/model/de-de”
- Copied “etc/voxforge.lm.bin” and “etc/voxforge.dic” to “/Applications/PsychoPy.app/Contents/Resources/lib/python3.8/pocketsphinx/model”
- Renamed “voxforge.lm.bin” to “de-de.lm.bin” to match the corresponding en-US file name
- Renamed “voxforge.dic” to “cmudict-de-de.dict” to match the corresponding en-US file name
- Changed language parameter in call to mic.bank() from “en-US” to “de-DE”
- Ran test experiment with microphone response component and built-in transcription enabled
What specifically went wrong when you tried that?:
Audio files are saved with the recorded input (mic access is not an issue), but no transcription output in the data file. No error occurs, although I suspect this might be related to the fix described in this post.
I’m not sure if I copied the correct files, if how I renamed the files might be an issue, or if using “de-DE” actually switches to the new model. This doc is the best advice I’ve found so far, but it’s not specific to psychopy.
Everything works as expected if “en-US” is used as the language… Has anyone succeeded in installing another language for pocketsphinx that worked with psychopy? Many thanks for any help!