search

Anjok07 / ultimatevocalremovergui

GUI for a Vocal Remover that uses Deep Neural Networks.
MIT License
17.28k stars 1.29k forks source link

[Feature Request] BandIt model support #1047

Open iGerman00 opened 8 months ago

iGerman00 commented 8 months ago

BandIt is a DNR model for separating cinematic sounds such as speech, sfx and background music. It is incredibly useful for audio editing and I'd love it to be included in UVR.

https://github.com/karnwatcharasupat/bandit https://zenodo.org/records/10160698 https://ieeexplore.ieee.org/document/10342812

ChrisTG742 commented 8 months ago

+1 Sounds interesting! :) Btw, you can already use 3rd party models by just putting them into the respective subfolder inside of "models" (and if you have, the corresponding yaml-configs in the config folder below "model_data", too). The application should auto-detect the presence of new models and let you choose them from the drop-down menu.

iGerman00 commented 8 months ago

Yea I don't have any experience with setting that kind of thing up. I think the models are in a different architecture anyway, though I'm not sure. I learned about BandIt on mvsep so I thought maybe it's appropriate to suggest this to @Anjok07 since it's probably slightly similar to UVR5, just online.

slayermaster commented 3 months ago

I put it in on mdx_net models folder, and renamed it from chpt to ckpt, and put .yaml file in the subfolder. It just fails, errors out.

HopeAndFaithStudios commented 2 months ago

Late to the party (LOL) but I agree 100% on this!!! I would LOVE a BandIt model for UVR5! I've been using UVR5 for about half a year now, and I'd love to be able to separate the music, vocals and sound effects all into 3 different tracks! I've been trying to find something like this for about as long as I had UVR5, and the BandIt model is pretty much what I'm looking for. I'm also asking (if it's possible) because I'm still feeling a little iffy about using MVSep.