Automatic DJ Transitions with Differentiable Audio Effects and Generative Adversarial Networks

TL;DR

This paper presents a data-driven method using GANs and differentiable audio effects to automate seamless DJ transitions, learning from real mixes to produce realistic and competitive results.

Contribution

It introduces a novel approach combining differentiable digital signal processing with GANs for automatic DJ transitions, enabling learning from real-world mixes.

Findings

Model achieves competitive results in listening tests.

Differentiable EQ and fader effectively simulate real DJ transitions.

Approach outperforms several baseline methods.

Abstract

A central task of a Disc Jockey (DJ) is to create a mixset of mu-sic with seamless transitions between adjacent tracks. In this paper, we explore a data-driven approach that uses a generative adversarial network to create the song transition by learning from real-world DJ mixes. In particular, the generator of the model uses two differentiable digital signal processing components, an equalizer (EQ) and a fader, to mix two tracks selected by a data generation pipeline. The generator has to set the parameters of the EQs and fader in such away that the resulting mix resembles real mixes created by humanDJ, as judged by the discriminator counterpart. Result of a listening test shows that the model can achieve competitive results compared with a number of baselines.