Abstract

Synesthesia is a neurological phenomenon in which certain stimuli (e.g., music, poems) automatically trigger a secondary experience (e.g., color perception, shapes, flavors). In this context, this work aims to develop a system capable of transforming a song into colored digital paintings. The proposal utilizes the lyrics recognized on the track to generate a target image using Stable Diffusion. The generated image is then progressively reconstructed in grayscale on a canvas using an existing Evolutionary Algorithm, based on brushstrokes, named the Genetic Painting Algorithm. This algorithm evolves multiple sets of brushstrokes, iteratively selecting the most similar ones to the target image. Finally, the grayscale painting is colored using an emotion-based rule system that considers the spectral characteristics of the song. Experimental results show that the grayscale approximation converges effectively within a limited …