DTT-BSR: GAN-based DTTNet with RoPE Transformer Enhancement for Music Source Restoration

TL;DR

This paper introduces DTT-BSR, a GAN-based model with RoPE transformer and dual-path RNN for music source restoration, achieving top leaderboard rankings with high fidelity and semantic accuracy in a compact form.

Contribution

It presents a novel hybrid GAN architecture combining RoPE transformer and dual-path RNN for improved music source restoration.

Findings

Achieved 3rd place on ICASSP 2026 MSR Challenge objective leaderboard.

Achieved 4th place on subjective leaderboard.

Model has a compact size of 7.1 million parameters.

Abstract

Music source restoration (MSR) aims to recover unprocessed stems from mixed and mastered recordings. The challenge lies in both separating overlapping sources and reconstructing signals degraded by production effects such as compression and reverberation. We therefore propose DTT-BSR, a hybrid generative adversarial network (GAN) combining rotary positional embeddings (RoPE) transformer for long-term temporal modeling with dual-path band-split recurrent neural network (RNN) for multi-resolution spectral processing. Our model achieved 3rd place on the objective leaderboard and 4th place on the subjective leaderboard on the ICASSP 2026 MSR Challenge, demonstrating exceptional generation fidelity and semantic alignment with a compact size of 7.1M parameters.