Graph-based Polyphonic Multitrack Music Generation

TL;DR

This paper introduces a novel graph-based deep learning model for polyphonic multitrack music generation, enabling hierarchical structure-content separation and conditional instrument control, resulting in realistic and musically coherent compositions.

Contribution

It proposes a new graph representation for music and a hierarchical Variational Autoencoder that models structure and content separately, enhancing controllability and interpretability in music generation.

Findings

Generated music is tonally and rhythmically consistent.

Model can interpolate between musical sequences realistically.

Latent space organizes according to musical concepts.

Abstract

Graphs can be leveraged to model polyphonic multitrack symbolic music, where notes, chords and entire sections may be linked at different levels of the musical hierarchy by tonal and rhythmic relationships. Nonetheless, there is a lack of works that consider graph representations in the context of deep learning systems for music generation. This paper bridges this gap by introducing a novel graph representation for music and a deep Variational Autoencoder that generates the structure and the content of musical graphs separately, one after the other, with a hierarchical architecture that matches the structural priors of music. By separating the structure and content of musical graphs, it is possible to condition generation by specifying which instruments are played at certain times. This opens the door to a new form of human-computer interaction in the context of music co-creation. After training the model on existing MIDI datasets, the experiments show that the model is able to generate appealing short and long musical sequences and to realistically interpolate between them, producing music that is tonally and rhythmically consistent. Finally, the visualization of the embeddings shows that the model is able to organize its latent space in accordance with known musical concepts.