Algorithm info: Submitted by: Bas Curtiz
Version: UVR v5.5.0/Patch_12_22_22_23_44/Win
Advanced MDX-Net:
Chunks: Auto
Chunk Margin: 44100
Volume Compensation: 1.035
Denoise Output: On
Spectral Inversion: On
Advanced Demucs:
Segments: Default
Shifts: 1
Overlap: 0.25
Chunks: Auto
Chunk Margin: 44100
Enable Chunks: Off
Split Mode: Off
Combine Stems: Off
Spectral Inversion: On
Based on CyberWaifu's suggestion at Audio Separation Discord:
"BREAKING NEWS: In terms of Vocal SDR, UVR-MDX-NET_Main_438 is actually NOT the best vocal model.
Currently, that goes to UVR-MDX-NET_Main_427.
SDR Vocals: 11.012639
SDR Instrumental: 10.387514"
Date added: 2022-12-24 |
