• 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

Native Speech / Music / SFX Separation 20261104

Edit name / description Delete
ID Ensemble SDR Music SDR SFX SDR Speech
11083 - 8.7942 10.7707 12.0410
Algorithm info: Offline three-stem audio separation.

Metrics:
Metric sdr for music: 8.7942
Metric si_sdr for music: 7.8498
Metric l1_freq for music: 65.2541
Metric log_wmse for music: 14.8406
Metric aura_stft for music: 6.0628
Metric aura_mrstft for music: 6.5887
Metric bleedless for music: 19.3978
Metric fullness for music: 14.4957
Metric sdr for sfx: 10.7707
Metric si_sdr for sfx: 10.3923
Metric l1_freq for sfx: 64.6365
Metric log_wmse for sfx: 13.9423
Metric aura_stft for sfx: 9.9727
Metric aura_mrstft for sfx: 14.6116
Metric bleedless for sfx: 34.7224
Metric fullness for sfx: 30.4835
Metric sdr for speech: 12.0410
Metric si_sdr for speech: 11.6945
Metric l1_freq for speech: 76.2358
Metric log_wmse for speech: 16.6728
Metric aura_stft for speech: 7.2853
Metric aura_mrstft for speech: 7.3122
Metric bleedless for speech: 39.4395
Metric fullness for speech: 20.5044


Date added: 2026-11-04

Edit submission

Delete submission

MVSEP Logo
Contact support
Advanced features

Quality Checker

Algorithms

Full API Documentation

How the Queue Works

Company

Privacy Policy

Terms & Conditions

Refund Policy

Extra

Help us promote!