Redistribution of pre-trained models in open source project?

[RyanMetcalfeInt8]

Hello! Awesome project -- I would love to add some of the pre-trained models that you give links to in top-level README as selectable models / features in our AI plugins for Audacity project ( https://github.com/intel/openvino-plugins-ai-audacity ).

Is this okay? I didn't see any license information posted for the linked models. Would it be possible to add an open source license to your project (like MIT, etc.)?

Thanks!
Ryan

[ZFTurbo]

Hello, not all models were posted by me. While I can add some open license on repo I think I can't really decide on each model. Are you interested in some particular models to add?

[RyanMetcalfeInt8]

Thanks!

Are you interested in some particular models to add?

Yes, at least initially I am interested in these ones:

• MelBand Roformer (KimberleyJensen edition) for Vocal Extraction
• MelBand Roformer (anvuew edition) for Reverb Removal

And also potentially these ones.. still evaluating them:

• MelBand Roformer Denoise (by aufr33)
• MelBand Roformer Denoise Aggressive (by aufr33)

And outside of the model weights themselves, I'll need to port some of the python bits to C++ (various pieces from https://github.com/ZFTurbo/Music-Source-Separation-Training/blob/main/utils.py, https://github.com/ZFTurbo/Music-Source-Separation-Training/blob/main/models/bs_roformer/mel_band_roformer.py, etc). And this is because I distribute Audacity plugins as pure-C++ compiled DLLs, which makes distribution a bit easier than python (IMO).

Thanks!
Ryan