A Physically Based Sound Space for Procedural Agents

Physically based sound models have a “natural” setting in dimensional space: a physical model has a shape and an extent and can be given a position relative to other models. In our experimental system, we place procedurally animated agents in a world of spatially situated physical models. The agents move in the same space as the models and can interact with them, playing the models and changing their configuration. The result is an ever-varying audiovisual landscape. This can be seen as purely generative—as a method for creating algorithmic music—or as a way to create instruments that change autonomously as a human plays them. A third perspective is in between these two: agents and humans can cooperate or compete to produce a gamelike or interactive experience.