Automatic Playlist Sequencing and Transitions

Abstract

Professional music curators and DJs artfully arrange and mix recordings together to create engaging, seamless, and cohesive listening experiences, a craft enjoyed by audiences around the world. The average listener, however, lacks both the time and the skill necessary to create comparable experiences, despite access to same source material. As a result, user-generated listening sessions often lack the sophistication popularized by modern artists, e.g. tracks are played in their entirety with little or no thought given to their ordering. To these ends, this paper presents methods for automatically sequencing existing playlists and adding DJ-style crossfade transitions between tracks: the former is modeled as a graph traversal problem, and the latter as an optimization problem. Our approach is motivated by an analysis of listener data on a large music catalog, and subjectively evaluated by professional curators.

Related

July 2020 | IJCAI - International Joint Conference on Artificial Intelligence

Seq-U-Net: A One-Dimensional Causal U-Net for Efficient Sequence Modelling

Daniel Stoller, Mi Tian, Sebastian Ewert, and Simon Dixon

July 2020 | WCCI/IJCNN - IEEE World Congress on Computational Intelligence / International Joint Conference on Neural Networks

Using a Neural Network Codec Approximation Loss to Improve Source Separation Performance in Limited Capacity Networks

Ishwarya Ananthabhotla, Sebastian Ewert, Joseph A. Paradiso