About

I had two motivations for this project. Firstly, an interest in modern art - more specifically geometric abstraction - and a hazy mind’s eye view of what I wanted to create. Where else to start? Secondly, my background in software development made me think about using technology to help create something aesthetically engaging.

I could have used vector graphics to create the polygonal geometry but decided on AI image generation instead. I have purposely used an early version of a diffusion model that I can run on my own computer. This avoided the more stylised and polished image generation of some later models. A great deal of time was spent refining a relatively simple structured prompt. This gave me geometry of infinite difference, along with a certain rawness and imperfection.

There are serious issues surrounding AI models being trained on works without consent. If they can be properly addressed, I think AI can have a positive role in art creation. I have tightly constrained its use in this project and situated it within a wider human–AI process rather than treating it as a means of producing finished works.

I wanted strong material texture in the compositions. The only realistic way to achieve it was by painting in acrylic, charcoal and ink - the analogue part of the project - followed by scanning. I experimented with different tools and methods, and it was surprisingly difficult to paint for surface texture rather than composition.

I developed a small suite of Python programs to pull it all together and make it into a repeatable working process. It is now quite streamlined, so I can concentrate on curation. I generate a large number of source images, select a few promising ones, and make decisions on orientation and arrangement to create potential compositions. From these, I select the final work.