Introduction

The field of generative music is not new at all: it dates back to the late 50s, when the first experiments of automatic music composition have been done. Since then, different approaches to music generation have been explored, first applying grammars and fractals. These are particularly good at expressing an aesthetically pleasing music composition, since they can easily encode the patterns that humans like and compose them in meaningful ways. Their nature however lacks of variety as they stick to the expressions that the author of the grammar wants to put in them, so they are pretty much constrained. A completely different approach to music composition has recently defined the field of evolutionary music: musical patterns are generated using evolutionary algorithms, which lead to very expressive systems. Evolutionary algorithms have been applied to interactive music production and have been proved to be very successful in this field, even capable of passing the Turing test for AIs in a jam session together with a human. Differently from other applications of evolutionary algorithms, generative music is particularly well suited for interactivity since the definition of a fitness function for music patterns is hard to produce and mainly biased by some expert's knowledge - while the appreciation of a particular pattern is a very subjective measure - so the use of the humans’ sense of aesthetic to give a fitness value to each pattern of a population is well suited for the goal.

While some good work has been done to interactively produce music pieces in jam sessions, its application is limited to work with expert musicians that have deep knowledge in the field, while applications in video games have yet to be explored.

Music games are not a new field either: the emerging popularity of rhythm games like Guitar Hero have proved that the interest for music and the desire for creativity can make the genre very interesting for the game industry and pleasing to the public. Most of these titles, though, are more oriented to music execution rather than interactive music creation.

This work describes “A Rock GAME” (A Rock Genetic Algorithm for Music Evolution), a music game that employs generative music to produce a creative and interactive experience.