Florian Thalmann, Queen Mary University of London

Summary
Fast DJ is a web-based automatic DJ system and module that can be embedded into any website. It generates transitions between any pair of successive songs and uses machine learning to adapt to the user’s taste via simple interactive decisions.

Description
A web-based minimal-UI DJ system that adapts to the user’s preference via simple interactive decisions and feedback on taste. Starting from a preset decision tree modeled on common DJ practice, the system can gradually learn a more customised and user-specific tree. At the core of the system are structural representations of the musical content based on semantic audio technologies (Dynamic Music Objects) and inferred from features extracted from the audio directly in the browser. These representations are gradually combined into a representation of the mix which could then be saved and shared with other users. Different types of transitions can be modeled using simple musical constraints. Potential applications of the system include crowd-sourced data collection, both on temporally aligned playlisting and musical preference. The functionality is also available as an npm module with a simple interface which can be embedded into any website (e.g. used in Moodplay and Grateful Dead Live websites).

 Links
https://dynamic-music.github.io/fast-dj
https://www.npmjs.com/package/auto-dj

Publications
Florian Thalmann, Lucas Thompson, and Mark B. Sandler. “A User-Adaptive Automated DJ Web App with Object-Based Audio and Crowd-Sourced Decision Trees”. (2018)