SYMPLEX: Controllable Symbolic Music Generation using Simplex Diffusion with Vocabulary Priors

TL;DR

This paper introduces SYMPLEX, a novel diffusion-based model for fast, controllable symbolic music generation that leverages vocabulary priors to enable detailed manipulation of musical features without additional training.

Contribution

The paper presents a new simplex diffusion approach for symbolic music generation that allows for direct control over musical attributes through vocabulary priors, without task-specific adaptation.

Findings

Effective control over music features via vocabulary priors

Fast generation of 4-bar multi-instrument loops

No need for extrinsic control or model retraining

Abstract

We present a new approach for fast and controllable generation of symbolic music based on the simplex diffusion, which is essentially a diffusion process operating on probabilities rather than the signal space. This objective has been applied in domains such as natural language processing but here we apply it to generating 4-bar multi-instrument music loops using an orderless representation. We show that our model can be steered with vocabulary priors, which affords a considerable level control over the music generation process, for instance, infilling in time and pitch and choice of instrumentation -- all without task-specific model adaptation or applying extrinsic control.