MVSEP Logo
  • Home
  • News
  • Plans
  • Demo
  • FAQ
  • Create Account
  • Login

    Ensemble (vocals, instrum)

    Ensemble of best vocal models. Algorithm gives the highest possible quality for vocal and instrumental stems. The latest ensemble consists of BS Roformer, MelBand Roformer and SCNet XL IHF vocal models.

    Quality table

    Algorithm name Multisong dataset Synth dataset MDX23 Leaderboard
    SDR Vocals SDR Instrumental SDR Vocals SDR Instrumental SDR Vocals
    Ensemble (2023.09)
    (UVR-MDX-NET-Voc_FT, Demucs4 Vocals 2023, MDX23C, VitLarge23)
    10.44 16.74 12.76 12.46 11.17
    Ensemble (2024.02)
    (BS Roformer (v1), MDX23C, VitLarge23)
    10.75 17.06 12.72 12.42 ---
    Ensemble (2024.03)
    (BS Roformer (viperx), MDX23C)
    11.06 17.37 13.00 12.70 ---
    Ensemble (2024.04)
    (BS Roformer (finetuned), MDX23C)
    11.33 17.63 13.57 13.27 ---
    Ensemble (2024.08)
    (BS Roformer (finetuned), MelBand Roformer)
    11.50 17.81 13.79 13.50 ---
    Ensemble (2024.12)
    (BS Roformer (finetuned), MelBand Roformer, SCNet XL)
    11.61 17.92 14.09 13.79 ---
    Ensemble (2025.06)
    (BS Roformer (x2), MelBand Roformer (ft), SCNet XL IHF)
    11.93 18.23 14.46 14.17 ---

    Detailed statistics on Multisong dataset:

    Model Vocals fullness Vocals bleedless  Vocals SDR Vocals L1Freq Instrum fullness Instrum bleedless  Instrum SDR Instrum L1Freq
    Ensemble (2025.06) 17.73 36.29 11.93 39.94 28.75 47.64 18.23 40.90
    Ensemble High Vocals Fullness (2025.06) 20.46 32.77 11.69 39.86 --- --- --- ---
    Ensemble High Instrumental Fullness (2025.06) --- --- --- --- 34.79 41.47 17.69 40.51
    🗎 Copy link

    MVSEP Logo

    turbo@mvsep.com

    Advanced features

    Quality Checker

    Algorithms

    Full API Documentation

    Company

    Privacy Policy

    Terms & Conditions

    Refund Policy

    Cookie Notice

    Extra

    Help us translate!

    Help us promote!