Computer-Generated Stylistic Compositions with Long-Term Repetitive and Phrasal Structure

Abstract

This article describes and evaluates an algorithm called Racchmaninof-Jun2015, referred to hereafter as Racchmaninof, which generates passages of music in a specifiable style. For generating all four parts of a Bach hymn (one of two target styles evaluated as part of a listening study – the other being Chopin mazurkas), we found that only five out of 25 participants performed significantly better than chance at distinguishing Racchmaninof’s output from original human compositions. These participants had a mean of 8.56 years of formal musical training and mode “daily/weekly” regularity of playing an instrument or singing. In the context of relatively high levels of musical expertise, this difficulty of distinguishing Racchmaninof’s output from original human compositions underlines the promise of our approach. Current trends and issues in the area of automatic stylistic composition are introduced and discussed and we consider the potential for applying our algorithm to additional composers and/or genres of music.