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

• URL: http://arxiv.org/abs/2603.04032v1
• Date: Wed, 04 Mar 2026 13:10:39 GMT
• Title: Multi-Stage Music Source Restoration with BandSplit-RoFormer Separation and HiFi++ GAN
• Abstract summary: Music Source Restoration (MSR) targets recovery of original, unprocessed instrument stems from fully mixed and mastered audio.<n>This technical report presents the CP-JKU team's system for the MSR ICASSP Challenge 2025.
• Score: 8.647179779228761
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: 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.

Related papers

• DTT-BSR+: A Generative-Regression Cascade for Music Source Restoration [13.053766595316224]
  DTT-BSR+ is a two-stage cascade MSR system that decouples distribution fitting from signal reconstruction into separate stages.<n>A generative DTT-BSR separator in the first stage produces stems matching the prior of clean sources.<n>A modified Demucs network in the second stage enhances the first stage output using time-domain and multi-resolution spectral losses.
  arXiv  Detail & Related papers  (2026-06-23T04:22:20Z)
• ResMerge: Residual-based Spectral Merging of Large Language Models [52.65011413018957]
  We propose ResMerge, a residual-based spectral merging framework for RL experts.<n>Based on this observation, we propose ResMerge, a residual-based spectral merging framework for RL experts.
  arXiv  Detail & Related papers  (2026-06-01T13:42:45Z)
• Music Source Restoration with Ensemble Separation and Targeted Reconstruction [6.528712884168779]
  The Inaugural Music Source Restoration (MSR) Challenge targets the recovery of original, unprocessed stems from fully mixed and mastered music.<n>To address MSR, we propose a two-stage system. First, an ensemble of pre-trained separation models produces preliminary source estimates.<n>Then a set of pre-trained BSRNN-based restoration models performs targeted reconstruction to refine these estimates.
  arXiv  Detail & Related papers  (2026-03-13T07:40:42Z)
• Summary of The Inaugural Music Source Restoration Challenge [52.405241781545634]
  Music Source Restoration (MSR) aims to recover original, unprocessed instrument stems from professionally mixed audio.<n>MSR Challenge features objective evaluation on studio-produced mixtures using Multi-Mel-SNR, Zimtohrli, and FAD-CLAP.<n>Winning system achieved 4.46 dB Multi-Mel-SNR and 3.47 MOS-Overall, corresponding to relative improvements of 91% and 18% over the second-place system, respectively.
  arXiv  Detail & Related papers  (2026-01-07T19:31:06Z)
• SAM Audio: Segment Anything in Audio [55.50609519820557]
  General audio source separation is a key capability for multimodal AI systems.<n>We present SAM Audio, a foundation model for general audio separation.<n>It unifies text, visual, and temporal span prompting within a single framework.
  arXiv  Detail & Related papers  (2025-12-19T22:14:23Z)
• High-Quality Sound Separation Across Diverse Categories via Visually-Guided Generative Modeling [65.02357548201188]
  We propose DAVIS, a Diffusion-based Audio-VIsual Separation framework that solves the audio-visual sound source separation task through generative learning.<n>Our framework operates by synthesizing the desired separated sound spectrograms directly from a noise distribution, conditioned concurrently on the mixed audio input and associated visual information.
  arXiv  Detail & Related papers  (2025-09-26T08:46:00Z)
• Back to Ear: Perceptually Driven High Fidelity Music Reconstruction [4.380428073231143]
  epsilonar-VAE is an open-source music signal reconstruction model that rethinks and optimize the Variational Autoencoders (VAEs)<n> Experiments show epsilonar-VAE at 44.1kHz substantially outperforms leading open-source models across diverse metrics.
  arXiv  Detail & Related papers  (2025-09-18T12:41:34Z)
• Music Source Restoration [20.814486236405823]
  We introduce Music Source Restoration (MSR), a novel task addressing the gap between idealized source separation and real-world music production.<n>MSR models mixtures as degraded sums of individually degraded sources, with the goal of recovering original, undegraded signals.<n>Due to the lack of data for MSR, we present RawStems, a dataset annotation of 578 songs with unprocessed source signals organized into 8 primary and 17 secondary instrument groups, totaling 354.13 hours.
  arXiv  Detail & Related papers  (2025-05-27T23:27:31Z)
• Separate This, and All of these Things Around It: Music Source Separation via Hyperellipsoidal Queries [53.30852012059025]
  Music source separation is an audio-to-audio retrieval task.<n>Recent work in music source separation has begun to challenge the fixed-stem paradigm.<n>We propose the use of hyperellipsoidal regions as queries to allow for an intuitive yet easily parametrizable approach to specifying both the target (location) and its spread.
  arXiv  Detail & Related papers  (2025-01-27T16:13:50Z)
• A Stem-Agnostic Single-Decoder System for Music Source Separation Beyond Four Stems [53.30852012059025]
  Banquet is a system that allows source separation of multiple stems using just one decoder. A bandsplit source separation model is extended to work in a query-based setup in tandem with a music instrument recognition PaSST model.
  arXiv  Detail & Related papers  (2024-06-26T20:25:53Z)
• The CHiME-7 DASR Challenge: Distant Meeting Transcription with Multiple Devices in Diverse Scenarios [61.74042680711718]
  We introduce the CHiME-7 distant ASR (DASR) task, within the 7th CHiME challenge. This task comprises joint ASR and diarization in far-field settings with multiple, and possibly heterogeneous, recording devices. The goal is for participants to devise a single system that can generalize across different array geometries.
  arXiv  Detail & Related papers  (2023-06-23T18:49:20Z)
• Music Separation Enhancement with Generative Modeling [11.545349346125743]
  We propose a post-processing model (the Make it Sound Good) to enhance the output of music source separation systems. Crowdsourced subjective evaluations demonstrate that human listeners prefer source estimates of bass and drums that have been post-processed by MSG.
  arXiv  Detail & Related papers  (2022-08-26T00:44:37Z)
• Wavesplit: End-to-End Speech Separation by Speaker Clustering [35.573006591664864]
  We introduce Wavesplit, an end-to-end source separation system. From a single mixture, the model infers a representation for each source and then estimates each source signal. Wavesplit is also applicable to other domains, by separating fetal and maternal heart rates from a single abdominal electrocardiogram.
  arXiv  Detail & Related papers  (2020-02-20T18:30:36Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.