AI-generated images have made significant progress since the days of abstract renderings and glitchy amalgamations. OpenAI’s newly released “Images for ChatGPT” has an uncanny ability to nail depth, shadows, and even text. It’s unleashed a frenzy of people recreating a familiar style: Hayao Miyazaki’s work at Studio Ghibli. The art style was already ubiquitous across the internet, thanks to its comforting, soft aesthetic (just look at Lofi girl) — and now, it’s a fully automated formula.
The trend kicked off pretty wholesomely. Couples transformed portraits, pet owners generated cartoonish cats, and many people are busily Ghibli-fying their families (I’ve stuck to selfies, not wanting to share with OpenAI my siblings’ likenesses). It’s an AI-generated version of the human-drawn art commissions people offer on Etsy — you and your loved ones, in the style of your favorite anime.
It didn’t take long for the trend to go full chaos mode. Nothing is sacred: the Twin Towers on 9/11, JFK’s assassination, Nvidia CEO Jensen Huang signing a woman’s chest, President Donald Trump’s infamous group photo with Jeffrey Epstein, and even OpenAI CEO Sam Altman’s congressional testimony have all been reimagined with that distinctive Ghibli whimsy (it’s not clear whether these users transformed uploaded images, or prompted the system to copy them). Altman has played into the trend too — he even changed his X profile picture into a Ghibli rendering of himself and encouraged his followers to make him a new one.
The guidelines for Images with ChatGPT are surprisingly flexible about producing this kind of content. “The model can generate images that resemble the aesthetics of some artists’ work when their name is used in the prompt,” according to the GPT-4o system card. “We added a refusal which triggers when a user attempts to generate an image in the style of a living artist.” (It’s worth noting that Miyazaki is, indeed, still alive.)
OpenAI spokesperson Taya Christianson told The Verge that the company isn’t “blocking the model’s capability to depict adult public figures, but we implement the same safeguards we employ for editing images of photorealistic uploads of people.” She added that “depicting violence in artistic, creative or fictional contexts is generally allowed to enable creative and artistic endeavors.”
“Our goal is to give users as much creative freedom as possible. We continue to prevent generations in the style of individual living artists. We do permit broader studio styles — which people have used to generate and share some truly delightful and inspired original fan creations,” Christianson wrote via email.
That doesn’t mean you get free reign over transforming photos, which a lot of the ones going viral seemingly do. “Users should only upload materials to our services that they own or have the rights to use,” Christianson said.
These guidelines, while flexible. haven’t stopped users from generating Ghibli-fied Hitler in Paris and Korean-American business owners with guns during the 1982 Los Angeles riots. There’s already a website dedicated to compiling all the AI-generated images, many of them featuring transformed photos of public figures.
ChatGPT does block direct requests to generate photos of public figures or cartoon characters — when I asked it to make a photo of Will Smith eating spaghetti, or Mark Zuckerberg wearing a Verge-branded hoodie, it declined those requests. But it’s not hard to bypass it by asking ChatGPT to generate photos with a Totoro-like character, or describing traits that point it to generating Zuckerberg. In principle, the guidelines state that you cannot use ChatGPT in a way that “infringes, misappropriates or violates anyone’s rights.” You can report copyright infringement to ChatGPT directly, per the guidelines, and OpenAI “will terminate accounts of repeat infringers where appropriate.”
OpenAI has faced numerous copyright disputes and conflicts over celebrity likenesses. Actor Scarlett Johansson threatened legal action for a voice tool that resembled her own, and there have been a litany of lawsuits against the startup from authors, major media publications, and artists who are trying to prove their content was used to train these models. For its part, OpenAI has been largely cagey about admitting what data is used and whether it was licensed, while it and other AI companies argue in court that AI is transforming the content into something completely new, not just regurgitating it.
Meanwhile, Miyazaki isn’t the only artist getting copied. People have been able to render pictures in the styles of Rick & Morty, Wallace and Gromit, The Simpsons, and South Park.
It isn’t clear how the animators of Rick & Morty and The Simpsons feel about their work being transformed without permission — or if they might question how OpenAI got enough data to generate such images. ChatGPT multimodal product lead Jackie Shannon previously told The Verge that users own the images they generated here “and are free to use them within the bounds of our usage policies as they would like.”
Miyazaki, the cofounder of Studio Ghibli, has actually made his feelings about AI-generated art quite clear. Before the rise of modern generative AI in 2016, the animator was shown a “deep learning”-created video clip from a production intern. His words have resonated with AI skeptics.
“I am utterly disgusted,” Miyazaki said. “If you really want to make creepy stuff, you can go ahead and do it, but I would never wish to incorporate this technology into my work at all. I strongly feel that this is an insult to life itself.”
But it didn’t take long for users on X to Ghibli-fy Miyazaki too:
Other Articles
No Comment! Be the first one.