How would you approach a procedural “Basquiatism” generator

p5.js Project Guidance
1

I’m trying to map out a project I’ve been calling Basquiatism.

The idea is to build a procedural UI component library + background generator that captures the raw visual essence of Jean-Michel Basquiat’s paintings in a respectful, rule-based way.

Im thinking of a UI component library that encapsulates a background generator and UI elements like buttons, input fields and so on.

My long-term dream is for it to become as reusable and recognizable as something like the Dracula color theme, except for a full visual/UI language.

A big part of what pushed me in this direction was seeing projects where people translate painterly logic into code instead of relying on prompts. In particular, Yusef28’s painterly shader work made me realize that if you want something that actually feels human, irregular, and intentional, you probably need to program the visual rules from the ground up rather than just ask an image model to imitate them. Here’s the youtube profile that helped shape how I’m thinking about this:

https://youtu.be/_VEUioLmyPI

Right now I’m still in the ideation / research phase. My rough idea is something like:

  • Research a set of Basquiat paintings I want to study.

  • Break them apart into recurring systems: motifs, icons, typography behavior, color fields, textures, collage structures, stroke irregularity, margin figures, etc.

  • Use graphics programming to rebuild those rules into a generator.

  • Build an interactive UI where users can dial elements up or down with sliders/buttons/toggles, like background density, text chaos, line roughness, symbol frequency, margin activity, and so on.

  • Eventually use this style system inside my own project, eccomuse.com.

I also reached out to James Dalzell Hodge (Jam2go), and his advice was basically: don’t lean on machine learning first, do a deep dive on the process yourself, break the paintings into layers (textures, palette, symbols, brushes), and build a component library of things that fit the style. That advice honestly made a lot of sense to me.

So I’m mostly trying to answer questions like:

  • Does this sound primarily like a shader problem, or more like a mixed graphics pipeline problem?

  • If you were building this, what tools would you start with first: GLSL/Shadertoy, p5.js, Processing, TouchDesigner, Three.js, OpenCV, custom SVG tooling, something else? Consider that i want to host this UI style generator on a site.

  • How would you approach the extraction side without drifting into “just train a model on Basquiat” territory?

  • How would you handle the typography side, since a lot of the feeling comes from irregular text, inconsistent spacing, changing stroke feel, lists, crossed-out words, etc.? I have some ideas, such as using multple Basquiat fonts available online edited to fit the project.

  • If I want reusable UI components out of this and not just static images, what would be the smartest architecture?

  • What problems do you think I’d run into early - technical, aesthetic, or conceptual?

I’m especially interested in hearing from people who have built painterly shaders, procedural design systems, or generators based on analyzing an artist’s visual process rather than just copying surface appearance.

I’m not looking for “just use Midjourney” type answers. I’m specifically interested in methods, tools, pipeline ideas, and warnings from people who’ve done adjacent work. I dont even think Midjourney could do something like this.

Thanks.