Learning Interpretable Representation for Controllable Polyphonic Music Generation

TL;DR

This paper introduces a VAE-based model that learns interpretable latent factors of polyphonic music, enabling controllable generation such as style transfer and texture variation.

Contribution

It proposes a novel architecture for disentangling chord and texture in polyphonic music within a VAE framework, enhancing controllability.

Findings

Successful disentanglement of chord and texture factors

High-quality controllable music generation demonstrated

Applications include style transfer and accompaniment arrangement

Abstract

While deep generative models have become the leading methods for algorithmic composition, it remains a challenging problem to control the generation process because the latent variables of most deep-learning models lack good interpretability. Inspired by the content-style disentanglement idea, we design a novel architecture, under the VAE framework, that effectively learns two interpretable latent factors of polyphonic music: chord and texture. The current model focuses on learning 8-beat long piano composition segments. We show that such chord-texture disentanglement provides a controllable generation pathway leading to a wide spectrum of applications, including compositional style transfer, texture variation, and accompaniment arrangement. Both objective and subjective evaluations show that our method achieves a successful disentanglement and high quality controlled music generation.