I started experimenting with AI Art in 2020. Long before Generative AI.
My first coding was based on the Neural Style Transfer process that was pioneered by Leon Gatys. The process involves imposing an artistic style onto a photograph. The image below is an example: it uses the style of Van Gogh imposed on a photograph of the north west corner of Tenerife at sunset.
Once the computer has been setup to impose a style on a photo, it is possible to tweak the variables to get a wide range of options.
Here are “impressions” of the Zhaozhou and Haozhou Bridges in the style of Turner. I have included the source photographs for interest.
Here is a photo of Mary, my wife, from the 1980’s with two renditions based on a “loose modern style”. In the first image I have asked the software to produce something closer to the original photo than the second version.
Stable Diffusion
Producing impressive images by typing in a few instructions is now commonplace. But there was a period when less sophisticated but interesting images were possible using a prototype of a process called Stable Diffusion.
I wonderd if this could be a source of inspiration for an artist to find a subject to paint. In the example below I asked the computer to produce 16 images of woman looking wistful in an expressionist style. The black rectangle shows when the process failed.
In other examples I asked the computer to generate poster style images of angry and elegant woman.
I then chose some promising images and painted in acrylic. Interesting but I wasn’t adding anything.
