I just wanted to know about the scale of external data used for UMX-L. It just says private compressed stems, but could you tell us how much extra data it was trained on?
I also don’t know the details - only what @faroit wrote here: FaRo on X: "We also present a `umxl` a new open-unmix 4-stem model, that was pre-trained on 50x more stems data compared to MUSDB18, thus greatly improving generalization. To try it out, just run ``` pip install openunmix umx input.wav --model umxl ``` code: https://t.co/YTkFnJ04Z5" / X
@faroit Do you remember how many hours the dataset corresponded to?
Kind regards
Stefan