Singer-Informed Vocal Source Separation for Multi-Singer Music Mixtures
IWAENC - 2026
Extract the singer matching a separate enrollment recording with the official Concat dual λ=0.1 or FiLM layer-varying checkpoint.
Loading...
Separate overlapping singers and choir parts, listen to each and select parts for further editing.
Input
Supports inputs in formats such as mp3 / wav / flac / ogg / m4a / aac / aif / aiff / mp4 / mov / mkv
Supported formats
Output
WAV
Use case
Remix, practice with accompaniment and edit separate parts.
Choose a model for your material and purpose, and review the examples before submitting.
Cost, time, formats and parameters depend on the selected model.
Include the tool name, steps and the problem in your feedback. Screenshots can help us locate it.
Concat · dual lambda=0.1: Remix, practice with accompaniment and edit separate parts.
FiLM · Layer-varying: Remix, practice with accompaniment and edit separate parts.
Separate overlapping male and female vocals. Similar vocal ranges and choirs may leave voices in the other track.
Separate overlapping male and female vocals. Similar vocal ranges and choirs may leave voices in the other track.
IWAENC - 2026
Extract the singer matching a separate enrollment recording with the official Concat dual λ=0.1 or FiLM layer-varying checkpoint.
ICASSP - 2026
Blindly separate two or more concurrent singing voices into numbered singer stems with UNMIXX.
arXiv - 2025
Separate alto, bass, finger snap, lead vocal, soprano, tenor and vocal percussion with SepACap.
2 output categories
Only records with verified public inference implementations and matching weights are shown; algorithms requiring no weights are also eligible. Closed services, unreleased weights and unmatched experiment versions are excluded. This is not an all-model leaderboard.
Each set of weights appears once per benchmark, using its best published result for the selected metric. An overall score comes from one evaluation. Expand to compare evaluations and settings.
Public weights do not establish a fully open-source license. Expand the data for declared weight licenses and access conditions. Undeclared licenses are marked separately.
Markers identify models or components used here. Scores and settings come from the cited tests, not a new evaluation of the complete site pipeline.
Choose a dataset to view all metrics for each model. Rows follow the official rank or selected metric. Bold marks each column’s best comparable value, including ties. Different evaluation protocols are ranked separately; missing scores are shown as -.
A complete public inference package for these experiments has not been verified. Other weights with the same model name do not qualify.
Concat layer-varying · Duet target singer · Baseline · Duet target singer
Scores come from the linked base checkpoint, not a new measurement of the model running here.
| Metric rank | Model | Test setting | SI-SDR (dB) ↑ | Both singers · SI-SDR (dB) ↑ | Background · SI-SDR (dB) ↑ | Interfering singer + background · SI-SDR (dB) ↑ |
|---|---|---|---|---|---|---|
| 1 | Concat dual λ=0.1Used on this sitePublic weights · Inference code | Duet target singer | 5.58 | 0.59 | -5.91 | 7.98 |
| 2 | FiLM layer-varyingPublic weights · Inference code | Duet target singer | 3.39 | 0.17 | -10.14 | -2.99 |
Markers identify models or components used here. Scores and settings come from the cited tests, not a new evaluation of the complete site pipeline.
Datasets, versions and conditions remain separate, including sources with no eligible ranking rows.
Deployable entries 2 / source rows 4
A complete public inference package for these experiments has not been verified. Other weights with the same model name do not qualify.
Concat layer-varying · Duet target singer · Baseline · Duet target singer
Scores apply to the listed dataset, metric and model version; results from different tests are not directly comparable.
The table includes only models with public weights or configurations that can run locally.
Choose a dataset to view all metrics for each model. Rows follow the official rank or selected metric. Bold marks each column’s best comparable value, including ties. Different evaluation protocols are ranked separately; missing scores are shown as -.
Scores come from the linked base checkpoint, not a new measurement of the model running here.
| Metric rank | Model | Test setting | SI-SDR (dB) ↑ | Both singers · SI-SDR (dB) ↑ | Background · SI-SDR (dB) ↑ | Interfering singer + background · SI-SDR (dB) ↑ |
|---|---|---|---|---|---|---|
| 1 | Concat dual λ=0.1Public weights · Inference code | Duet target singer | 5.58 | 0.59 | -5.91 | 7.98 |
| 2 | FiLM layer-varyingPublic weights · Inference code | Duet target singer | 3.39 | 0.17 | -10.14 | -2.99 |