• Quality Checker
  • Queue
  • Synth Leaderboard (Full) Synth Leaderboard (Single Models) Synth Leaderboard (Ensemble)
  • Multisong Leaderboard (bass) Multisong Leaderboard (drums) Multisong Leaderboard (other) Multisong Leaderboard (vocals) Multisong Leaderboard (instrum)
  • Other Leaderboards

MDX-Net: kim vocal model fine tuned (old) + UVR-MDX-NET_Main_427 + Demucs: v4 | htdemucs_ft - Ensemble Algorithm: Average/Average

ID Ensemble SDR Instrumental SDR Vocals
123 + 11.1768 11.4584
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: Off

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: Off

Based on the fact Kim's vocal fine tuned model scored highest in non-ensembled.
However, this is an older model (November), than she uploaded at 2022-12-04.
I asked her, but she doesn't have the newer model anymore, since she's still training it.

Time Elapsed: 01:03:06

Date added: 2022-12-25
MVSEP Logo

turbo@mvsep.com

Advanced features

Quality Checker

Algorithms

Full API Documentation

Company

Privacy Policy

Terms & Conditions

Refund Policy

Extra

Help us translate!

Help us promote!