break me, genAI

Abstract

break me, generative ai is an audiovisual artwork that explores the complex and dynamic nature of musical perception through a hybrid human–AI creative process. The piece combines hand-drawn MIDI parameter curves, representing subjective perceptual qualities such as “grittiness” and “rattle,” with procedurally generated visuals inspired by synesthetic art traditions. These initial visualizations are subsequently transformed through multiple generative AI models, including StyleGAN and diffusion-based systems, introducing machine-learned aesthetics while maintaining traces of the artist’s intent. By layering human-authored structures with AI-driven transformations, the work creates an evolving visual interpretation of music that foregrounds ambiguity and perceptual change. The project challenges conventional audio-reactive visualization approaches by emphasizing context, subjectivity, and interpretation. It also raises questions about authorship, influence, and the preservation of artistic perception within AI-mediated workflows, positioning generative AI as a collaborative rather than autonomous creative agent.