Input
Full-band songs or dense mixes split into vocals, drums, bass, guitar, piano, and other WAV stems
Supported formats
Output
Six separated WAV stems: vocals, drums, bass, guitar, piano, and other.
Use case
Fixed vocals, drums, bass, guitar, piano, and other stems from one mix
Choose by input and output needs, then comparable tests, published evaluations and capabilities. Missing scores do not imply lower quality; more parameters or a newer release do not prove superiority. This is a suggested usage order, not an overall leaderboard.
Cost, time, formats and parameters depend on the selected model.
AvailableSix-Stem Separation splits a mix into six WAV stems: vocals, drums, bass, guitar, piano, and other.
BS-RoFormer SW Fixed uses BS-Rofo-SW-Fixed.ckpt plus BS-Rofo-SW-Fixed.yaml to split a full mix into six WAV stems.
Separates a full mix into four standard MUSDB-style stems.
XLANCE-MSR extends a six-stem BS-RoFormer system into eight-class source restoration with separation, denoising, and dereverberation/restoration. The MIT checkpoint weights are public.
Open-weight BS-RoFormer checkpoint for MVSep Mega 53 Stems. It targets many instrument classes and filters detected outputs rather than replacing the six-stem quality default.
BS-RoFormer SW Fixed uses BS-Rofo-SW-Fixed.ckpt plus BS-Rofo-SW-Fixed.yaml to split a full mix into six WAV stems.
Scores apply only to the cited dataset, metric and model version. Leading this table does not mean a current world record or a result from this service. Capabilities and tests can justify prioritizing models without comparable scores.
The table includes only models with public weights or configurations that can run locally.
This tool uses BS-RoFormer SW Fixed for vocals, drums, bass, guitar, piano, and other.
Public sources checked: 2026-09-10
| Test setting | Model | Vocals SDR (dB) | Detailed metrics |
|---|---|---|---|
| MVSep entry 8374 | BS-RoFormer SW Fixed 6-stemTelkNetMVSep entry 8374 This tool uses BS-RoFormer SW Fixed for vocals, drums, bass, guitar, piano, and other. | 11.3019 | Instrumental SDR (dB)17.5091Bass SDR (dB)14.6235Drums SDR (dB)14.1129Other SDR (dB)8.7169 |
| Rank | Model | Piano SDR (dB) | Detailed metrics |
|---|---|---|---|
| MVSep entry 9578 | BS-RoFormer SW FixedTelkNetMVSep entry 9578 | 7.8031Best | Other SDR (dB)19.5824Best |
| MVSep entry 7396 | Ensemble (SCNet + Mel)MVSep entry 7396 | 6.2054 | Other SDR (dB)18.4851 |
| MVSep entry 7395 | SCNet Large (2024.09)MVSep entry 7395 | 5.8912 | Other SDR (dB)18.1698 |
| MVSep entry 7394 | MelRoformer (viperx)MVSep entry 7394 | 5.7118 | Other SDR (dB)17.9917 |
| MVSep entry 7392 | mdx23c (2024.09)MVSep entry 7392 | 5.5996 | Other SDR (dB)17.8796 |
| MVSep entry 7390 | mdx23c (2023.08)MVSep entry 7390 | 4.7969 | Other SDR (dB)17.0766 |
BS-RoFormer SW Fixed · MVSep guitar SDR
Public sources checked: 2026-09-10
| Rank | Model | Guitar SDR (dB) | Detailed metrics |
|---|---|---|---|
| MVSep entry 9245 | BS-RoFormer SW FixedTelkNetMVSep entry 9245 | 9.0120Best | Other SDR (dB)15.9828Best |
| MVSep entry 7149 | Ensemble (BSRoformer + MelRoformer)MVSep entry 7149 | 7.5112 | Other SDR (dB)14.4820 |
| MVSep entry 6741 | Ensemble (mdx23 + MelRoformer)MVSep entry 6741 | 7.1835 | Other SDR (dB)14.1542 |
| MVSep entry 6776 | BSRoformer (viperx)MVSep entry 6776 | 7.1266 | Other SDR (dB)13.9775 |
| MVSep entry 6720 | MelRoformer (2024.06)MVSep entry 6720 | 7.0211 | Other SDR (dB)13.9919 |
| MVSep entry 6740 | mdx23c (2024.06)MVSep entry 6740 | 6.3422 | Other SDR (dB)13.3130 |
Huge-SCNet-4stems V1.2 · MVSep Multisong
Public sources checked: 2026-09-10
| Rank | Model | Average SDR (dB) | Detailed metrics |
|---|---|---|---|
| MVSep entry 9648 | Huge-SCNet-4stems V1.2MVSep entry 9648 | 10.0155Best | Vocals SDR (dB)9.6073Bass SDR (dB)12.0639BestDrums SDR (dB)11.7422BestOther SDR (dB)6.6485Best |
| MVSep entry 8446 | SCNet XL IHF 4-stemMVSep entry 8446 | 9.9261 | Vocals SDR (dB)9.6896BestBass SDR (dB)11.9436Drums SDR (dB)11.5839Other SDR (dB)6.4873 |
Official results rank XLANCE-MSR first on objective metrics and listening tests; these are eight-class source-restoration scores, not six-stem SDR scores.
Public sources checked: 2026-09-10
| Test setting | Model | MMSNR | Detailed metrics |
|---|---|---|---|
| Official challenge result | xlancelabOfficial challenge results | 4.4623 | Zimt0.0137FAD0.1988Separation MOS4.2358Restoration MOS3.3892Overall MOS3.4665 |
| Official challenge result | CUPAudioGroupOfficial challenge results | 2.3405 | Zimt0.0164FAD0.2253Separation MOS3.8360Restoration MOS2.9173Overall MOS2.9253 |
| Official challenge result | HachimiOfficial challenge results | 2.0016 | Zimt0.0183FAD0.2939Separation MOS3.5814Restoration MOS2.6331Overall MOS2.7235 |
| Official challenge result | 1.4520 | Zimt0.0182FAD0.2907Separation MOS3.5425Restoration MOS2.4768Overall MOS2.5412 | |
| Official challenge result | 0.8329 | Zimt0.0189FAD0.3814Separation MOS3.5510Restoration MOS2.0838Overall MOS2.1414 |
These are the official per-stem SDR values published on the MVSep Mega 53 page. They are displayed as scores, not as a product ranking.
Public sources checked: 2026-09-10
| Stem | Model | Open-source SDR | Detailed metrics |
|---|---|---|---|
| Accordion | MVSep Mega 53 Stems | 6.2494 | New model SDR6.6498Delta SDR0.4004 |
| Acoustic guitar | MVSep Mega 53 Stems | 5.0024 | New model SDR5.0797Delta SDR0.0773 |
| Back vocal | MVSep Mega 53 Stems | 6.4179 | New model SDR6.5262Delta SDR0.1083 |
| Banjo | MVSep Mega 53 Stems | 3.1593 | New model SDR3.7532Delta SDR0.5939 |
| Bass | MVSep Mega 53 Stems | 11.1680 | New model SDR11.2886Delta SDR0.1206 |
| Bassoon | MVSep Mega 53 Stems | 4.6595 | New model SDR5.1663Delta SDR0.5068 |
| Bells | MVSep Mega 53 Stems | 1.1190 | New model SDR4.8040Delta SDR3.6850 |
| Bowed strings | MVSep Mega 53 Stems | 12.4486 | New model SDR12.4486Delta SDR0.0000 |
| Brass | MVSep Mega 53 Stems | 6.7042 | New model SDR6.8487Delta SDR0.1445 |
| Cello | MVSep Mega 53 Stems | 5.0364 | New model SDR5.2257Delta SDR0.1893 |
| Clarinet | MVSep Mega 53 Stems | 5.0505 | New model SDR5.8690Delta SDR0.8185 |
| Congas | MVSep Mega 53 Stems | 9.1747 | New model SDR9.5946Delta SDR0.4199 |
| Digital piano | MVSep Mega 53 Stems | 7.9634 | New model SDR9.0179Delta SDR1.0545 |
| Dobro | MVSep Mega 53 Stems | 7.6562 | New model SDR8.2290Delta SDR0.5728 |
| Double bass | MVSep Mega 53 Stems | 14.1731 | New model SDR15.6032Delta SDR1.4301 |
| Drums | MVSep Mega 53 Stems | 9.2502 | New model SDR11.4520Delta SDR2.2018 |
| Electric guitar | MVSep Mega 53 Stems | 8.1543 | New model SDR8.1856Delta SDR0.0313 |
| Flute | MVSep Mega 53 Stems | 6.2134 | New model SDR6.8557Delta SDR0.6423 |
| French horn | MVSep Mega 53 Stems | 5.2635 | New model SDR5.5136Delta SDR0.2501 |
| Glockenspiel | MVSep Mega 53 Stems | 3.6621 | New model SDR8.1523Delta SDR4.4902 |
| Guitar | MVSep Mega 53 Stems | 2.5661 | New model SDR2.6164Delta SDR0.0503 |
| Harmonica | MVSep Mega 53 Stems | 10.9265 | New model SDR11.8575Delta SDR0.9310 |
| Harp | MVSep Mega 53 Stems | 6.3767 | New model SDR7.6523Delta SDR1.2756 |
| Harpsichord | MVSep Mega 53 Stems | 1.6090 | New model SDR1.9524Delta SDR0.3434 |
| Hi-hat | MVSep Mega 53 Stems | 2.4681 | New model SDR3.2904Delta SDR0.8223 |
| Keys | MVSep Mega 53 Stems | 9.3032 | New model SDR9.3106Delta SDR0.0074 |
| Kick | MVSep Mega 53 Stems | 11.5485 | New model SDR11.5485Delta SDR0.0000 |
| Lead vocal | MVSep Mega 53 Stems | 5.4663 | New model SDR5.4663Delta SDR0.0000 |
| Mandolin | MVSep Mega 53 Stems | 4.4256 | New model SDR4.7735Delta SDR0.3479 |
| Marimba | MVSep Mega 53 Stems | 4.4821 | New model SDR4.8830Delta SDR0.4009 |
| Oboe | MVSep Mega 53 Stems | 3.3616 | New model SDR4.5555Delta SDR1.1939 |
| Organ | MVSep Mega 53 Stems | 10.3684 | New model SDR10.8244Delta SDR0.4560 |
| Percussion | MVSep Mega 53 Stems | 2.5008 | New model SDR2.8897Delta SDR0.3889 |
| Piano | MVSep Mega 53 Stems | 6.7787 | New model SDR6.8080Delta SDR0.0293 |
| Saxophone | MVSep Mega 53 Stems | 8.8875 | New model SDR9.4985Delta SDR0.6110 |
| Sitar | MVSep Mega 53 Stems | 4.6529 | New model SDR5.0434Delta SDR0.3905 |
| Snare | MVSep Mega 53 Stems | 6.1338 | New model SDR6.7784Delta SDR0.6446 |
| Strings | MVSep Mega 53 Stems | 8.8151 | New model SDR8.8151Delta SDR0.0000 |
| Synth | MVSep Mega 53 Stems | 2.0539 | New model SDR2.0539Delta SDR0.0000 |
| Tambourine | MVSep Mega 53 Stems | 3.0589 | New model SDR3.5544Delta SDR0.4955 |
| Timpani | MVSep Mega 53 Stems | 4.6423 | New model SDR4.9779Delta SDR0.3356 |
| Toms | MVSep Mega 53 Stems | -2.0607 | New model SDR-1.0708Delta SDR0.9899 |
| Triangle | MVSep Mega 53 Stems | 5.9274 | New model SDR6.0197Delta SDR0.0923 |
| Trombone | MVSep Mega 53 Stems | 2.6949 | New model SDR3.0751Delta SDR0.3802 |
| Trumpet | MVSep Mega 53 Stems | 4.9658 | New model SDR5.8668Delta SDR0.9010 |
| Tuba | MVSep Mega 53 Stems | 7.1957 | New model SDR7.5229Delta SDR0.3272 |
| Ukulele | MVSep Mega 53 Stems | 6.7869 | New model SDR6.9721Delta SDR0.1852 |
| Viola | MVSep Mega 53 Stems | 1.8581 | New model SDR1.8581Delta SDR0.0000 |
| Violin | MVSep Mega 53 Stems | 3.1018 | New model SDR3.3285Delta SDR0.2267 |
| Vocal | MVSep Mega 53 Stems | 11.6590 | New model SDR11.6590Delta SDR0.0000 |
| Wind | MVSep Mega 53 Stems | 8.6317 | New model SDR8.6632Delta SDR0.0315 |
| Wind chimes | MVSep Mega 53 Stems | 4.6440 | New model SDR6.7529Delta SDR2.1089 |
| Woodwind | MVSep Mega 53 Stems | 3.3123 | New model SDR3.3224Delta SDR0.0101 |
SDX23 music separation track - 2023
BS-RoFormer uses band-split spectrogram features plus RoPE Transformers; the SDX23 system ranked first in the music source separation track and a smaller version reported 9.80 dB average SDR on MUSDB18HQ.
ICASSP - 2023
HT Demucs replaces core Demucs layers with cross-domain Transformer attention and reports 9.20 dB SDR with extra data and fine-tuning.
ICASSP - 2025
Introduces the subband sparse-compression architecture behind the deployed Huge-SCNet-4stems checkpoint: the mixture spectrogram is split into bands compressed at different rates for efficient high-quality separation.
arXiv - 2026
Technical report for the XLANCE-MSR system submitted to the first Music Source Restoration Challenge.
arXiv - 2026
Challenge summary paper describing the MSR task, datasets, objective metrics, subjective evaluation, and final ranking tables.
Working on a similar task? These guides list the matching tools, model boundaries, sources, and FAQs for each scenario.
Include the tool name, steps and the problem in your feedback. Screenshots can help us locate it.