I’ve been trying to create Mexican courtyards ever since I first started playing around with AI image generation.1 Back in July, during my wabi sabi phase, I created the image below. I liked it but thought it was somewhat incomplete.
Midjourney had just implemented the ability to add canvas to existing images, so I asked for an addition to the right side. If I’d been paying attention, this would’ve been my first chance to snap to Midjourney’s propensity for hanging pumpkins in the air. They looked like gourds at the time though, so I didn’t even notice.
The image still didn’t quite work in my opinion, so I asked for an addition to the left side as well. I was especially pleased with the skulls that showed up in the version below.
A month or so later, I returned to the wabi-sabi courtyard images and asked Midjourney to add some human skeletons into the mix.
Up until that point, I’d had no success whatsoever in convincing Midjourney to create full-length human skeletons. Then all of a sudden, wham! I had skeletons. Drinking wine, no less.
I loved their interactions with one another. They seemed so … human.
Alas, it’s a feat I’ve been unable to replicate. The image below falls into the genre of what I call Epic AI Fails. The two skeletons in the center are pretty good (except for the backwards foot on the left one), but what in the world is on the right? Is it a skeleton or a chair? Both? But where are its legs? (Regardless of whether it’s a skeleton or a chair?)
More and more, it seems I’m creating AI images just so I can guffaw at some of the results.
On another note, I can’t believe it’s less than 10 days to Halloween, and a bit more than that to Día de los Muertos. I have many images saved up, including a plethora of black cats. More emails coming your way soon.
Actually, I’ve been trying to create New Mexican courtyards. But Midjourney couldn’t get a hold of that concept at all. It understands plain old Mexican courtyards just fine, so I’ve been running with that phrase for the past year.
Hilarious!