Apr 12
Because of the semantic overlap, referencing any artist named Jean in a stable diffusion prompt tends to add gratuitous denim to the image. Every once in a while the results are funny enough to be worth keeping and improving on.
seen from United States
seen from China
seen from China
seen from Georgia
seen from Singapore
seen from United Kingdom
seen from Malaysia
seen from Malaysia
seen from South Korea
seen from Singapore
seen from Türkiye
seen from Malaysia
seen from Türkiye
seen from Malaysia
seen from Vietnam
seen from Japan
seen from Poland
seen from Netherlands
seen from Romania
seen from Japan
Because of the semantic overlap, referencing any artist named Jean in a stable diffusion prompt tends to add gratuitous denim to the image. Every once in a while the results are funny enough to be worth keeping and improving on.
Decided to take a brief break from Blender projects and finish up a few of my in-progress 2D AI images. This one was just a semi-randomized recombination of prompt fragments that turned out particularly interestingly, upscaled and with all details meticulously refined as usual.
Another old one I started in like 2022 finished. I like how you can definitely tell from the compositions that these older images were made with much more primitive tech, even if I've polished up the details with a newer model.
Found this interesting character while wandering through latent image space and decided to polish up the image a bit.
Just spent the last several hours meticulously refining details on this one. I'm tempted to keep going, but I should probably avoid following that impulse on these randomly generated images that have no greater purpose.
Here's a little preview of a project I'm working on in IntraPaint, this is maybe 10% of the full image. For once I decided to actually pick up a pen and use an ink on paper sketch as the starting point. The rest of it was my usual process of controlled AI inpainting mixed with manual edits.