As generative AI races to produce ever more sleek and sophisticated images, a new online trend has users asking for the opposite: deliberately simplistic and terrible-looking ones. The worse, the better.
The resulting images flooding social media look like crude doodles out of Microsoft Paint in the ‘90s. Derpy faces have cartoonish rectangular teeth, half triangles for noses and chaotic squiggles for hair. Reimagined company logos look like chief marketing officers handed over their jobs to their 3-year-olds.
It all started when Korean creative director and graphic designer Wonjae Gi shared on Threads a prompt for basic line drawings, labeling it “the most trivial prompt in the world.”
“I kept seeing more and more polished, cinematic and high-quality AI images online,” Gi said in an email interview. “They were impressive, but at the same time I started feeling a kind of fatigue from how perfect everything was becoming. So I became curious about going in the completely opposite direction making something intentionally awkward, low-quality and trivial, just for fun.”
After OpenAI CEO Sam Altman reshared an X post about the prompt in late April, it quickly took off on the company’s ChatGPT conversational chatbot. So much so that OpenAI added the prompt directly into Images 2.0, a new upgrade to ChatGPT’s image generation tool that promises a higher-quality output.
You can forget higher quality here. Those taking part in the viral trend want lower quality. Much lower. ChatGPT competitor Grok Imagine, the AI chatbot launched by Elon Musk, has also added a “Scribbli” template that outputs images in a similar rough-hewn style.
What Is The Exact Prompt?
“Redraw the attached image in the most clumsy, scribbly and utterly pathetic way possible,” the widely used prompt reads. “Use a white background, and make it look like it was drawn in MS Paint with a mouse. It should be vaguely similar but also not really, kind of matching but also off in a confusing, awkward way, with that low-quality pixel-by-pixel feel that really emphasizes how ridiculously bad it is. Actually, you know what, whatever, just draw it however you want.”
OpenAI says it saw the trend gaining traction about a week after Images 2.0 launched on April 21.
Gi, the Korean designer, said it’s been more than a little surreal watching how quickly, and widely, the prompt has caught on.
“At some point, it stopped feeling like something that belonged to me personally,” he said. “It started feeling more like an internet phenomenon that people were collectively playing with.”
The silly images evoke nostalgia and simpler times, with Adele Li, OpenAI’s product management lead for ChatGPT Images, calling the craze “joyful, social and instantly understandable.” It comes at a time when public discussions around AI, and the companies that produce it, tend to be serious, technical and sometimes highly fraught, as reflected in the trial between Elon Musk and OpenAI, which has become increasingly messy and personal.
FORBES | By Leslie Katz
ChatGPT Trend Turns People Into Caricatures — And Shows How Well AI Knows Us
This isn’t the first viral ChatGPT image trend to inspire a wave of user-generated content. Earlier this year, another fad had people uploading photos of themselves to the OpenAI platform and asking it to produce a workplace caricature based on their chat history. And last year, Studio Ghibli-inspired images overtook social media after OpenAI released a new image generator.
But while that trend often led to rich, polished visuals, the latest obsession unapologetically embraces lo-fi and sloppy.
This article was originally published on Forbes.com