mlsanityData to 13 Oct 2026

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

arXiv 2603.04032 ↗First posted 4 Mar 2026cs.SD cs.LG eess.AScited by 0

Not on the map. Why?

Abstract (arXiv)

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.

Key facts. No citing papers in the map yet (data through 13 Oct 2026). Builds on 8 of its 12 references that are in the map. First posted on arXiv 4 Mar 2026.

Builds on

8 of 12 references.

References · 8

  1. 2019Cutting Music Source Separation Some Slakh: A Dataset to Study the Impact of Training Data Quality and Quantitycited by 86
  2. 2023Moisesdb: A dataset for source separation beyond 4-stemscited by 39
  3. 2023Music Source Separation with Band-Split RoPE Transformercited by 37
  4. 2022A Two-Stage U-Net for High-Fidelity Denoising of Historical Recordingscited by 17
  5. 2024FINALLY: fast and universal speech enhancement with studio-like qualitycited by 7
  6. 2025SonicMaster: Towards Controllable All-in-One Music Restoration and Masteringcited by 6
  7. 2025Music Source Restorationcited by 5
  8. 2025MSRBench: A Benchmarking Dataset for Music Source Restorationcited by 4

Built on by

Not cited yet.