More experiments with CLIP+VQGAN image generation. Turns out I can give it the names of completely fictional painters and it’ll generate distinct styles for each of them.
“Cathedral of Giants by Carmine Nottyors”
“Cathedral of Giants by Picov Andropov”
“Cathedral of Giants by Janelle Shane”
Interesting to think about the factors it’s using in coming up with the painting styles. Similarity to the names of other artists? Similarity to other words? I wonder how (if at all) this space could be explored.