I’m mapping out a project called Basquiatism: a procedural UI component library and background generator that captures the raw visual essence of Jean-Michel Basquiat’s paintings. My dream is for it to become as reusable as the Dracula color theme, but for a full visual/UI language (buttons, inputs, backgrounds).This was inspired by Yusef28’s painterly shader work https://youtu.be/_VEUioLmyPI. It made me realize that to capture an artist's style, you can't just prompt an AI; you need to program their visual rules from the ground up to capture intentional, human irregularity.My rough workflow idea:Break Basquiat's paintings into recurring systems: motifs, typography behavior, color fields, collage structures, stroke irregularity.Rebuild those rules into a generator using graphics programming.Build a web UI with sliders/toggles to dial elements up or down (text chaos, line roughness, background density).Eventually use it in my own project, eccomuse.com.I reached out to Jam2go, and his advice was: don’t lean on ML. Do a deep dive on the process, break the paintings into layers, and build a component library of elements fitting the style. This makes total sense to me.I’m trying to figure out:Is this primarily a shader problem or a mixed graphics pipeline problem?What tools would you use for a web-hosted UI style generator?How would you approach the analysis without just training an ML model?How would you handle typography (irregular text, spacing, changing stroke weight, crossed-out words)? I might use randomized existing fonts.What's the smartest architecture to output reusable web UI components rather than just static images?What early technical or aesthetic problems should I anticipate?I’m actively avoiding "just use Midjourney" solutions, as AI can't dynamically render this kind of raw UI deviance anyway. Any advice on methods, pipelines, or tools from people who have done adjacent work is hugely appreciated!
