A.I. Alternate Art

I’ve been wasting time experimenting with Stable Diffusion, a text-to-image generator. Input a string of words and the application’s algorithms will output 2-4 images. Artists are not amused. Its legality is still controversial. Results vary widely, particularly with the setting of the “guidance” slider. Here are some visual experiments as imagined by text prompts (in italics) to the program.

Giant Bart Simpson sitting on a building:

Homer Simpson sitting on a huge donut:

With less concrete suggestions (and a lower guidance number)‚ the results can be surprisingly artistic: Madonna, crucifixion, snakes, beetles:

While more specific instruction can yield uncanny results: Witch, pig, Durer, engraving:

The possibilities are endless: Oil refinery, garden of Eden, Van Gogh:

You do have to know when to quit: