Comparative Visual Aesthetics in Synesthetic Structures

Abstract

We describe a means of co-evolving parameters for procedural animation generation. Several experiments were completed in order to find an aesthetically pleasing animation mechanism, which could be modulated by music and have its parameters evolved over time. The selected music visualization is a set of vector fields for the purpose of moving particles whose paths create visual interest. To cause emergence of pleasing patterns in particle behaviour, several metrics were incorporated into the fitness function. To reflect the spatial characteristics of the vector fields a crossover operator was developed. The most critical aspect in creating a vector field which animates particle paths over time is that the vector field not allow the particles to become static or too tightly clustered, as either indicates the particles will no longer be circulating in interesting ways. We are interested in encouraging certain behaviours of the particles without doing so explicitly so that a local interaction approach allow for many aesthetically pleasing solutions to be reached, and rendered. This complex systems approach continues to work as interesting patterns emerge. We provide several examples with later generations creating more aesthetically complex renderings. Comparative visual aesthetics as a measure of such visual refinement is introduced. If pleasing patterns can be made implicitly with the vector fields in this specific audio-visual composition, it stands to reason that parameters for other visualizations may be evolved as well and this work serves as an argument for computer assisted comparative aesthetic refinement.