Hello! I would like to develop a ‘Music Separation’ plugin for Audacity. I have the underlying logic complete, it uses demucs_v4 deep learning model to separate a stereo track into 4 new tracks – vocals, bass guitar, drums, and other instruments.
I was wondering if there was a plugin framework (LADSPA, LV2, Audacity module, etc.) that is recommended to accomplish this? As my ‘standalone’ application is C/C++ based, I’d definitely like to stick with that if possible.
Thanks. I did see that project, and even downloaded / built the PR branch that was mentioned on the website. I was starting to use it’s ‘mod-deep-learning’ module as a reference for creating my own (non-torch based) module on Audacity’s 3.2.5 branch – but I find that many of the base classes (e.g. Effect, etc.) have changed or been shuffled quite a lot since then.
More generically, if one wants to create a plugin that accepts 1 stereo track, and outputs 4 new stereo tracks – is there a recommended approach for that?