Music Style Transfer with GANs: Cross-Domain Discriminators for Enhanced Realism - ISMIR 2021

This paper proposes a novel approach to music style transfer using Generative Adversarial Networks (GANs) with cross-domain discriminators. The proposed method effectively transfers the style of one music piece to another, achieving high realism and preserving the content of the source music. The key innovation lies in the introduction of cross-domain discriminators, which enable the generator to learn the intricate relationships between different musical domains, leading to more realistic and musically coherent style transfers. The paper presents extensive experiments demonstrating the effectiveness of the proposed method, achieving state-of-the-art results in terms of both realism and musical coherence.