Dear participants,
Presenting a new open-unmix 4-stem model, umxl: Pre-trained on 50x more stems data compared to MUSDB18, thus greatly improving generalization.
We hope this helps more participants to quickly get better scores!
1 Like