str9blk0rb

The rapid advancement of multimodal AI, such as ChatGPT and Stable Diffusion, is fundamentally transforming the creative environment in visual art and design. This study explores how image-generative AI can be utilized as a creative tool in the field of illustration, how it redefines the boundaries between traditional and AI-based creative practices, and what its potentials and limitations are. Focusing on Low-Rank Adaptation (LoRA) techniques within Stable Diffusion, the study combines case surveys, practical experiments, and the production of a final artwork.

             The final outcome is a custom LoRA model, str9blk0rb, trained on the researcher’s existing style, used in conjunction with GPT-4o to generate consistent outputs in Stable Diffusion. These results were then composed into a moving image via the Image-to-Video AI tool Higgsfield. The study reflects on how past memories, AI imagination, and LoRA outputs interact.