Hi!
Thanks a lot for this amazing repository, you did a brilliant job and I am astonished by the performances of your model. Your work at deezer research is so cool!

I would just have a feature request: right now I'd like to run your model on many different files but not all of them necessarily have the same extension or are in the same folder, etc.

Would it be possible to add an option to the CLI to provide a text file with a list of audio files to iterate over? Looking at the structure of your repo it would just overwrite the audio_files variable in your detect_key function.

Thanks a lot for your work and I hope I have the honor to meet you in person at ISMIR 2025!