Change your singer: a transfer learning generative adversarial framework for song to song conversion

TL;DR

This paper introduces SCM-GAN, a transfer learning-based generative adversarial network that converts songs to sound like a different artist, achieving high similarity and naturalness without needing parallel data.

Contribution

The work presents a novel non-parallel song conversion framework combining transfer learning and GANs, improving voice similarity and naturalness in song conversion tasks.

Findings

Transfer learning improves GV by 35% and MS by 13%.

Achieves 70% similarity to target singer on average.

Generated songs exhibit high naturalness and user satisfaction.

Abstract

Have you ever wondered how a song might sound if performed by a different artist? In this work, we propose SCM-GAN, an end-to-end non-parallel song conversion system powered by generative adversarial and transfer learning that allows users to listen to a selected target singer singing any song. SCM-GAN first separates songs into vocals and instrumental music using a U-Net network, then converts the vocal segments to the target singer using advanced CycleGAN-VC, before merging the converted vocals with their corresponding background music. SCM-GAN is first initialized with feature representations learned from a state-of-the-art voice-to-voice conversion and then trained on a dataset of non-parallel songs. Furthermore, SCM-GAN is evaluated against a set of metrics including global variance GV and modulation spectra MS on the 24 Mel-cepstral coefficients (MCEPs). Transfer learning improves the GV by 35% and the MS by 13% on average. A subjective comparison is conducted to test the user satisfaction with the quality and the naturalness of the conversion. Results show above par similarity between SCM-GAN's output and the target (70\% on average) as well as great naturalness of the converted songs.