Four years ago I made a series of AI generated portraits using an opensource GAN framework. These portraits have an artistic flair that I think is largely missing in the current crop of stable diffusion generators. The new generators specialize in photorealism and style transfer more than painterly images. One could say that the GAN portraits' painterly effect is just an artifact of the technologies limitations in crystalizing a particular image with definition. That is largely true. The fact that I had to find these images from the trained model by randomly generating images from the latent space also lent itself to a 'needle in the haystack' experience of finding usable painterly images.

Now that I work in AI for my day job (twisty.ai - An AI video platform) I have been exploring the new generation of tools to see what is being creating by artists and where I can lean in myself.

https://fabulistmagazine.com/mirrors-of-an-artificial-mind-gallery/