Building bridges with AI (Part 2) May 2019

Using AI to create images from text input became news early in 2018 but, its purpose still remains unclear. Far better would be the ability for AI to recognise images with the potential to create alt attributes on the fly. Just imagine how much more accessible the web would be? Even in this enlightened age 33.6% of all images have missing alternative text.

As a visual artist, the ability to input text phrases to create surprising and strange little images has its own lure. In my previous post I wrote about generating images from text inputs using AI. Based on the theme of bridges and using text phrases and idioms for the input resulted in 6 very surprising and captivating results.

The goal is to reproduce the images as intaglio prints but, as a diversion, I took one of the images and ran it through an arbitrary stylisation model in Runway. Using a variety of sources from Peter Doig to Jack B. Yeats, to my own paintings and drawings, it was images of Bazooka Joe comics which produced the most satisfying results.

I wanted the least amount of intervention between generating the images and printing them, but some manual processing was always going to be inevitable.

Proofing the experiments on an inkjet printer after producing a black separation from a CMYK conversion, then marginally adjusting the brightness and contrast proved to be very satisfying.

Next: Creating plates for intaglio printing using a 3d printer