I’ve been seeing some twitter threads going around recently showcasing the beautiful horrorshows of AI-generated images that have been created using Big Sleep, and I thought it would be a fun experiment to create my own, uh, “whimsical dreamscapes.”
From my luddite’s understanding, Big Sleep is a program that takes texts and runs them through some computery-AI stuff and spits out an image. Very technical, yes. So, here are the results of my experiments:
For the first run, I used the parameters “a cat that looks like a bird.” This returned the following two truly terrifying “animals.” The first one is definitely pretty bird-like, but I worry that the second creature is vaguely, uh, “NSFW.” But maybe that’s just me being a pervert.
The second trial was to see what the AI would come up with for “Big Rock Candy Mountain.” I received two color variations, so BRCM light mode / dark mode, if you will. They truly are fair and bright /Where the handouts grow on bushes /And you sleep out every night.
Those were fun, but let’s try something a little different. What about the quote “Forget it, Jake–It’s Chinatown.” Except I wasn’t sure if the punctuation would break the code, so it was more like “forget it jake its chinatown.” I thought these ones were absolutely lovely, but then again, I’m a fan of abstract expressionism:
And also for some reason this terrifying ‘dog’ showed up as the first image in all of my runs after the first one. It’s probably some sort of omen about my impending death. He looks like a ‘Frederick’ but I’m open to other name suggestions.

Is it art? Is it science? Does it matter?
Anyway, here are some pretty great instructions if you would like to make your own. I’d love to see them!