CROSS-DISCIPLINARY ART THROUGH AI-GENERATED MUSIC AND VISUALS

Abstract [English]

Multimodal generative systems change the current creative practices which is the focus of this study, looking at how the field of cross-disciplinary art has grown rapidly due to AI-generated music and visual synthesis. The study combines the concept of deep learning including GANs, VAEs, and transformer-based networks to learn the interaction between sonic and visual representations that can produce coherent works of art that combine sound, visuals, and real-time engagement. The proposed system, based on different data sets, incorporating annotated music collections, visual art collections, and records of multimodal performances, provides cross-modal associations that match rhythm, timbre, texture, color, and movement. The theoretical base integrates the cognitive theories of perception, emotion and synesthetic experience making AI not just a technical means but a creative partner who can assist new types of computational creativity. It uses a complete workflow pipeline of generative music-visual production, which can be used to create offline productions as well as in real time performance settings. It has a special user interface allowing artists to modulate parameters, control style transfer, and dynamically interact with developing multimodal content. Experimental testing is an integration of quantitative analysis (perceptual coherence, structural similarity, and temporal alignment) and qualitative response by the composers, visual artists and the viewers involved.