cs.SDMar 4, 2026

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

Authors: Tobias Morocutti, Emmanouil Karystinaios, Jonathan Greif, Gerhard Widmer

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

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.

Figures & tables

Explore similar work

CardsList
  1. Restore, Separate, Restore: A Modular Framework for Music Source Restoration

    Oct 6, 2026Tobias Morocutti, Emmanouil Karystinaios, Gerhard WidmerAudio Source SeparationMusic Source Separation

  2. Music Source Restoration with Ensemble Separation and Targeted Reconstruction

    Mar 13, 2026Xinlong Deng, Yu Xia, Jie JiangAudio Source SeparationMusic Source Separation

  3. DTT-BSR+: A Generative-Regression Cascade for Music Source Restoration

    Jun 23, 2026Youran Ni, Shihong Tan, Yuzhu Wang +1Audio Source SeparationMusic Source Separation