SubRoformer: Subband Time-Domain Representation for Music Source Separation

Chunyat Wu; Chuang Ding; Paris Smaragdis; Qiuqiang Kong

eess.ASNov 5, 2026

SubRoformer: Subband Time-Domain Representation for Music Source Separation

Authors: Chunyat Wu, Chuang Ding, Paris Smaragdis, Qiuqiang Kong

Organizations: The Chinese University of Hong Kong · Moonstep AI · Massachusetts Institute of Technology

Abstract

Music source separation extracts vocals and instruments from a mixed recording, enabling remixing, content production, and music analysis. Existing systems commonly operate on waveforms or full-band spectrograms. Waveform inputs contain many samples even in short segments, whereas a single spectrogram resolution imposes a shared time-frequency resolution trade-off across frequency ranges. We propose SubRoFormer, a Transformer-based system comprising a subband module and a residual RoFormer module. The subband module decomposes the mixture into frequency-localized, lower-rate signals for separation. It then reconstructs the target waveforms from the estimated subbands. The residual RoFormer module models temporal and cross-band relationships on temporally compressed features. It combines this context with fine-rate features through gated residual fusion to estimate source-specific complex masks. On the MUSDB18-HQ test set, training only on MUSDB18-HQ yields an average signal-to-distortion ratio (SDR) of 10.63 dB across four stems using target-specific models, achieving state-of-the-art performance among the compared systems without extra training data. With 500 additional training songs, a single model jointly separating all four stems achieves an average SDR of 11.59 dB, establishing state-of-the-art performance among the compared single-model multi-stem systems trained with extra data. Code and checkpoints are released at https://github.com/ArchiMickey/SubRoformer.

Figures & tables

Fig. 1: Overview of SubRoFormer . The fixed front end converts the analytic mixture into demodulated, decimated subbands; local complex features are processed by fine-rate boundary blocks and a temporally patched axial RoFormer; the inverse path interpolates, remodulates, and sums the estimated subbands. $C$ , $L$ , and $K$ denote channels, time samples, and realized subbands, respectively.

Fig. 2: Schematic target magnitude responses of the triangular filter bank, with illustrative frequency spacing. The highlighted interior band $A_{k}(f)$ has lower edge $f_{k,\ell}$ , peak $f_{k,p}$ , and upper edge $f_{k,u}$ . Neighboring responses overlap along their slopes; the first and last bands are one-sided. The dashed line shows their unit sum over $0\leq f\leq F_{s}/2$ . These are ideal targets, rather than the realized finite-length FIR responses.

Fig. 3: Per-band analysis and rate conversion: (a) the input $x[n]$ ; (b) the band-pass filtered signal $x_{k}[n]$ ; (c) the analytic signal $x_{k}^{\mathrm{a}}[n]$ ; (d) the complex baseband signal $b_{k}[n]$ ; and (e) the decimated sequence $z_{k}[m]$ . Uppercase symbols denote the corresponding spectra. The example uses $F_{s}=48$ kHz, $f_{k}=10.4$ kHz, approximately 800-Hz band support, and $D=60$ , giving a reduced sampling rate of $F_{s}/D=800$ Hz.

Transform	Rate	Mixture	Vocals	Bass	Drums	Other	LSD (dB) $\downarrow$	NME $\downarrow$
		Reconstruction SDR (dB) $\uparrow$
STFT [ 19 ]	4.271	138.58	138.47	138.75	138.64	138.63	0.0000	0.0000
SubRoFormer mel–linear	4.267	28.04	26.80	47.78	26.58	29.03	3.3375	0.0264
SubRoFormer ERB–linear	4.267	28.04	26.80	48.26	26.56	29.11	3.3318	0.0265
Auditory ERB [ 46 ]	0.125	0.54	0.05	1.00	0.88	0.15	113.0755	0.9353
PQMF [ 18 ]	1.000	11.30	10.25	13.41	12.03	10.38	3.5195	0.2770

TABLE I: Transform reconstruction quality on the MUSDB18-HQ test set. SDR is delay-aligned and summarized by the median over 50 songs per signal; LSD and NME are medians over 250 song–signal pairs. Rate counts real scalar coefficients per sample per channel, including padding. Higher SDR and lower LSD and NME indicate better fidelity.

Single-stem models
Model	Params.	$N$	Voc. $\uparrow$	Bass $\uparrow$	Dr. $\uparrow$	Oth. $\uparrow$	Avg. $\uparrow$
BSRNN [ 14 ]	37.6M	4	10.01	7.22	9.01	6.70	8.24
BS-RoFormer [ 15 , 35 ]	82.8M	4	11.02	11.58	9.66	7.80	10.02
Mel-RoFormer [ 35 ]	94.8M	4	11.60	–	9.34	7.93	–
SubRoFormer-S (Ours)	22.52M	4	11.34	9.18	10.60	8.16	9.82
SubRoFormer-B (Ours)	97.17M	4	12.63	10.20	10.40	9.27	10.63

TABLE II: MUSDB18-HQ metrics for systems trained without extra data. Metrics for research baselines are taken from their original papers.

Fig. 4: Mean test SDR (dB; higher is better) on MUSDB18-HQ versus total deployed parameters for selected systems trained without extra data. Single-stem systems, including both SubRoFormer sizes, use four networks; multi-stem baselines use one. The plotted BS-RoFormer and Mel-RoFormer points are six-layer configurations. Baseline results are drawn from [ 11 , 14 , 15 , 35 , 13 , 62 , 36 ] .

	Total (M)	Representation + mask (M)	Separator (M)
BS-RoFormer L6	72.19	50.87 (70.47%)	21.32 (29.53%)
Mel-RoFormer L6	84.19	62.87 (74.68%)	21.32 (25.32%)
SubRoFormer L4+2	24.08	2.76 (11.45%)	21.32 (88.55%)

TABLE III: Parameter distribution of single-output variants with matched separators.

Variant	SDR $\uparrow$	$\Delta$ SDR $\uparrow$
SubRoFormer-S (400k updates)	10.88	$0.00$
BS-RoFormer L6 ∗ [ 15 ]	10.18	–
Mel-RoFormer L6 ∗ [ 35 ]	10.08	–
A1: Front-end representation
Global-STFT band-split	9.43	$-1.45$
Mel band-split	8.88	$-2.00$

TABLE IV: Vocals-only ablation results on MUSDB18-HQ.

Model	Params.	$N$	Extra	Voc. $\uparrow$	Bass $\uparrow$	Dr. $\uparrow$	Oth. $\uparrow$	Avg. $\uparrow$
Single-stem models
HT Demucs (FT) [ 13 ]	41.4M	4	800	9.20	10.39	10.08	6.32	9.00
BSRNN [ 14 ]	37.6M	4	1750	10.47	8.16	10.15	7.08	8.97
BS-RoFormer L12 [ 15 ]	93.4M	3	450	12.72	13.32	12.91	9.01	11.99
Multi-stem models
Hybrid Demucs [ 12 , 13 ]	83.64M	1	800	8.75	9.13	9.31	6.18	8.34

TABLE V: MUSDB18-HQ metrics for systems trained with extra data and a contextual system with unknown training data. Metrics for research baselines are taken from their original papers.

Music Source Separation

See all 21 papers

Mar 4, 2026

Multi-Stage Music Source Restoration with BandSplit-RoFormer Separation and HiFi++ GAN

Music Source Restoration (MSR) targets recovery of original, unprocessed instrument stems from fully mixed and mastered audio, where production effects and distribution artifacts violate common linear-mixture assumptions. This technical report presents the CP-JKU team's system for the MSR ICASSP Challenge 2025. Our approach decomposes MSR into separation and restoration. First, a single BandSplit-RoFormer separator predicts eight stems plus an auxiliary other stem, and is trained with a three-stage curriculum that progresses from 4-stem warm-start fine-tuning (with LoRA) to 8-stem extension via head expansion. Second, we apply a HiFi++ GAN waveform restorer trained as a generalist and then specialized into eight instrument-specific experts.

Tobias Morocutti, Emmanouil Karystinaios, Jonathan Greif +1

Institute of Computational Perception (CP-JKU) Johannes Kepler University Linz, Austria

Music Source Separation Audio Source Separation

Ask PDF

Oct 6, 2026

Restore, Separate, Restore: A Modular Framework for Music Source Restoration

Music Source Restoration (MSR) seeks to recover original, unprocessed instrument stems from mixed, mastered, and possibly degraded recordings. Unlike conventional source separation, which treats the mixture as a linear sum of clean sources, MSR must additionally invert nonlinear production effects, such as equalization and compression, and transmission-related degradations, such as codec artifacts. We propose a three-stage framework built around this distinction: (1) a mixture restoration model that addresses degradation before separation, (2) a single model that separates the restored mixture into eight target stems (vocals, guitars, keyboards, synthesizers, bass, drums, percussion, and orchestra), and (3) stem-specific restoration experts fine-tuned on the separator's own residual artifacts. Each stage improves restoration quality over the previous one on the MSR Challenge test set. We release code and models to support future research in MSR at https://github.com/theMoro/music_source_restoration.

Tobias Morocutti, Emmanouil Karystinaios, Gerhard Widmer

Institute of Computational Perception, Johannes Kepler University Linz, Austria · LIT Artificial Intelligence Lab, Linz, Austria

Audio Source Separation Music Source Separation

Ask PDF

Mar 13, 2026

Music Source Restoration with Ensemble Separation and Targeted Reconstruction

The Inaugural Music Source Restoration (MSR) Challenge targets the recovery of original, unprocessed stems from fully mixed and mastered music. Unlike conventional music source separation, MSR requires reversing complex production processes such as equalization, compression, reverberation, and other real-world degradations. To address MSR, we propose a two-stage system. First, an ensemble of pre-trained separation models produces preliminary source estimates. Then a set of pre-trained BSRNN-based restoration models performs targeted reconstruction to refine these estimates. On the official MSR benchmark, our system surpasses the baselines on all metrics, ranking second among all submissions. The code is available at https://github.com/xinghour/Music-source-restoration-CUPAudioGroup

Xinlong Deng, Yu Xia, Jie Jiang

College of Artificial Intelligence, China University of Petroleum, Beijing

Audio Source Separation Music Source Separation

Ask PDF

Abstract

Music Source Separation

Explore similar work

Multi-Stage Music Source Restoration with BandSplit-RoFormer Separation and HiFi++ GAN

Restore, Separate, Restore: A Modular Framework for Music Source Restoration

Music Source Restoration with Ensemble Separation and Targeted Reconstruction