A formal approach for high-level automatic rhythm generation

Abstract

In this paper, we present a novel algorithm for automatically generating rhythms in real time in a given meter. The generated rhythms are "generic" in the sense that they are characteristic of each time signature without belonging to a specific musical style. A stochastic model in which various aspects and qualities of the generated rhythm can be controlled intuitively and in real time was developed. Such qualities are the density of the generated events per bar, the amount of variation in generation, the amount of syncopation, the metrical strength, and, of course, the meter. The kin.rhythmicator software application was developed to implement the algorithm.