BHenne confronted with a little downtime, many of my friends will turn to the same party game. It is based on the Surrealists’ Exquisite Corpse and involves rapidly translating short written descriptions into frantically drawn drawings and back again. One group calls it Telephone Pictionary, another refers to it as Writey-Drawey. From the Internet, I know that it is also popularly called Eat Poop You Cat, a phrase surely inspired by one of the game’s outcomes. While Exquisite Corpse can produce absurd moments, they rarely match the comedy of Eat Poop You Cat. Its outcomes, however, like inside jokes, are not always communicably funny. (I have often tried and failed to convey the heartbreaking silliness of a particularly memorable round in which the phrase Panic in the disco gradually transformed into a well-drawn horde of abnormal worms — some vomit, some smoke, some fuck, one sips a novelty beer hat.) But as a rule, bad translations tickle, and the shift between the verbal and the visual and the subsequent switch reinforces these humorous juxtapositions and inconsistencies.
Just three years ago, it was rare to encounter strange mistranslations of text to image or image to text in everyday life, which is why Eat Poop You Cat’s hair-raising results seemed particularly novel. But we have now entered a new era of image creation, powered by tools that explicitly rely on this kind of multimodal translation. Using AI image generators like DALL-E 3, Stable Diffusion and Midjourney, and the generative features built into Adobe’s Creative Cloud programs, you can now turn a sentence or phrase into a highly detailed image — or more — in just a few seconds. Images can also be translated into descriptive text almost instantly. Today, you can play “Eat Poop You Cat” alone in your room and play around with the algorithms.
In summer 2023 I tried a browser-based version of Stable Diffusion — an open source model for generative AI (or GenAI) for text-to-image conversion — and another AI browser application called CLIP Interrogator, which translates any image into a text prompt. It took about three minutes to play two rounds of the game: between twenty seconds and two minutes to generate an image from a prompt, and about another minute to convert that image back into a text prompt. I started by typing “Eat Poop You Cat” (why not?) into a box that asked me to “enter your prompt.” Then I clicked “Generate Image.”
The Stable Diffusion applet generates four images for each prompt. I cheated a little and just picked my favorite image to continue. Staring at me from the center of the image was a somewhat realistic tabby cat, with wide green eyes and an open mouth with a salmon-pink tongue sticking out. The background was a dingy gray without much detail; in the bottom third of the image was a bubbly white text in bold capitals: FOOD, FOOD, SHIT, SHIT, YOU’VE GOT A FEW LEFT OF YOU!
I dragged this image into CLIP Interrogator, which spat out the prompt: “A closeup of a cat with green eyes, blue text that says 3kliksphilip, epic urban background, poop, white border and background, licking, epic poster, office cubicle background, golden toilet, funny cartoon, Erin, classic gem, messy eater, exploitable image, abandoned, motivational, moving poetry, toilet.” Alongside the development of GenAI tools, a nuanced syntax for image generation prompts has emerged, and CLIP Interrogator’s “prompt” mimicked this accretionary layering of styles, details, and descriptors — although this list seemed exaggerated, like a psychedelic extrapolation of the image I was glad was already a “classic gem”. Apparently 3kliksphilip is a gaming YouTuber, but who is Erin? Why is background misspelled the first time, but correct the second time? Where does the golden toilet come from? Leave?
After some back and forth, I landed on a painting of a black and brown cat lounging on a dresser that could have been designed by Frank Lloyd Wright. A piece of toilet paper that had fallen from the roll above onto the cat’s head resembled a hat. The painting was flat and looked painted; dark contour lines conveyed the shapes. The style felt familiar — Expressionist? German Expressionist? Pseudo-naive? Certainly influenced by Modigliani, early Picasso and some of the later still lifes of the Polish Cubist Henri Hayden. Next to the cat and toilet, a copper cup balances on a matching copper saucer. A series of blocky words written in capital letters are scattered across the floor: RULE OF TWO FREE PROSDY.
