Ghibli, glibly

What does Hayao Miyazaki really think about AI?

The New Yorker’s daily cartoon, by Ali Solomon.

First, a quick heads up. March marked two years since I launched this newsletter. (How far we’ve come since the Astro Boy fashion craze of 2023.) In that time, I’ve learned that being a newsletter writer is a privilege. It’s also a tough business. I deeply appreciate each and every one of you who has subscribed, or even simply takes time to read. But I’m able to continue doing this thanks to the generous support of readers who value it enough to become paid subscribers. If you enjoy what I’m doing, please consider clicking the button below. You’ll not only help me continue making the content you enjoy, but gain full access to a the full archive of stories, perfect for binge-reading whether you’re in Astro Boy boots or not.

Which brings us to the topic for today. The word “glib” means “fluent but insincere and shallow.” It seems an appropriate description of the most recent phenomenon to sweep the online world: the Chat GPT-4o “Ghibli filter.” This is shorthand for a prompt that, when input into the paid version of the AI chatbot, transforms any photograph into a simulacrum of a Studio Ghibli animated frame. It proved so popular after its stealth-release last week that OpenAI CEO Sam Altman bragged his servers were “melting” from demand.

As faux-Ghibli frames flooded the internet, countless fans jumped on the bandwagon while equal numbers of furious critics condemned the trend. The sheer emotional energy expended over the last week on this highlights, yet again, the integral role that Japanese dreams have come to play in the global fantasyscape. But “Ghibli on tap” also raises many other questions. What are the implications for anime as an artform if AI can seemingly copy it? How did this one anime studio come to have such a hold on the global imagination? And then, hanging over all of it, is what does Ghibli think of all of this?

Studio Ghibli is, almost without question, the planet’s favorite producer of animated fantasies. Thanks to their long association with Disney, co-founder and director Hayao Miyazaki’s twin Oscar wins, and an Honorary Palme d’Or for the studio as a whole, Ghibli possesses an air of respectability that elevates them above the scrum producing cartoons for mass consumption. The studio operates almost entirely outside the usual channels of anime production, resulting in fare that is quite different from the usual anime. Its virtuosity and wholesomeness make it unimpeachable, in ways that the average mainstream anime, with its bubbling stew of adolescent angst and sexuality, is often not. In the late Eighties and early Nineties, Ghibli was what young otaku showed folks to prove something existed in between Tom and Jerry and tentacle porn. We knew even then that Miyazaki was a genre of one, and Ghibli the anime to watch when you think other people might be watching you.

Even still, back then, Ghibli was a subculture within a subculture. Now it is a household word. You can see this in how quickly the fad for the filter seized the planet. (You know you’ve really made it when you get a New Yorker cartoon.) The craze kicked off what feels like many years ago on March 25th, when an OpenAI engineer tweeted (or Xed, whatever you call it these days) a family photo “converted to studio ghibli anime” (sic). This led, predictably, to practically every iconic photo of the last century being fed into the system in rapid succession, giving us the jarring experience of seeing the Twin Towers falling, JFK being assassinated, and Luigi Mangione in the soothing, soft linework of Ghibi house style.

In yet another example of how little of a difference exists between brain rot and the brain trusts running the world, authority figures quickly jumped into the fray. The White House posted a Ghibli-fied image of a woman being detained by ICE for fentanyl trafficking, followed by the Israeli Defense Forces posting a medley of super kawaii IDF soldiers on land, air, and sea. These associations would undoubtedly outrage Miyazaki, who flies a progressive flag in a very literal sense – he famously hung a banner protesting nuclear power outside the Ghibli offices after the Fukushima meltdown, and refused to travel to the US to accept his 2001 Oscar as a protest against the Gulf War. But outrage, is, of course, the currency of politics in 2025.

Nevertheless, and despite what you may have seen online, both Miyazaki and Ghibli have so far remained silent on the topic of the filter, and on AI in general. A supposed cease and desist letter making the rounds is a hoax, and Miyazaki’s comments about AI being “an insult to life itself” are being misquoted. This hugely-circulated quote comes from a 2016 documentary called “Never-Ending Man: Hayao Miyazaki.” Miyazaki did say those words, but they’re not exactly about AI.

Let's take a look at the scene. The film’s producers bring Miyazaki face to face with a team from the telecom company Dwango. They demonstrate for Miyazaki how AI can be used to train characters to move on their own, without needing human animators. This is 2016, remember. Years before the advent of modern generative AI systems. Today, even a fumble-fingered fan can generate an animated sequence with a text prompt. Back then, it took teams of programmers to painstakingly train AI models, and Dwango’s character hasn’t quite gotten the hang of walking.

Instead, it writhes grotesquely across the ground, limbs horrifically askew, like a frog that’s been half-squished by a car. It’s horrible and disturbing and absolutely not ready for prime time, or any time. The team member making the presentation gamely remarks that the model needs more training, but might be useful for portraying zombie characters. 2016 might be early times for AI, but the attempt to spin suckitude into success remains, then as now, the industry’s guiding light.

The room falls into silence as the group, undoubtedly raised on Miyazaki’s work since childhood, waits for ther idol to speak. But things don’t go as expected. Miyazaki says that he has a disabled friend, and that the imagery “utterly disgusts” him. Apparently mistaking the demo for finished product, he continues, “if you want to make creepy stuff like this, go right ahead, but I would never use this in our work.” Then he utters the oft-quoted line. “I feel it’s an insult to life itself.” (My translations differ from the subtitles in the link above, because the subtitles suck — I mean, “utterly disgust” me.)

The Dwango team sits stone-faced as their leader stutters out that it isn’t intended to ridicule anyone, but simply a demonstration of a technology. Miyazaki is having none of it. Cut to Miyazaki’s office. “I feel like we are nearing end times.”

We hear you, Hayao. We hear you. But his oft-quoted comment isn’t what it a lot of people seem to think it is. He may truly feel that generative AI is an insult to life. In fact, I’d be surprised if he didn’t! But he isn’t actually talking about AI here. He’s talking about what he interprets — or misinterprets — as mocking the physically disabled. Earlier this week, I reached out to Ghibli for comment. I am not holding my breath that I’ll get a response, but if I do, you’ll be first to know.

Now on to the other questions. What is the impact of a Ghibli filter on Ghibli as a producer of anime? Or any producer of anime for that matter. My feeling is that the filter isn’t nearly as big of a deal as it may seem. Many of its critics cite Walter Benjamin, whose prescient 1933 essay “The Work of Art in the Age of Its Mechanical Reproduction” argues that the presence of an original is the prerequisite to the concept of authenticity. No copy can ever hope to hold a candle to the original, Benjamin argues, because an original possesses an “aura” that sets it apart from mere duplicates. This makes intuitive sense. We value a painting higher than a poster of it, or the experience of seeing a stunning sunset more than a photograph of it.

And we value Ghibli’s style more than we do mindless low-energy AI copies of it. But Japanese cultural products have always thrown a monkey-wrench into this idea of what’s authentic and what’s not. Shinto shrines are dismantled, rebuilt with new materials, and still treated as originals. Mass produced Ukiyo-e woodblock prints are viewed as fine art. And manga and anime are largely produced digitally these days, meaning neither have any “original” in the sense we usually think of it. In other words, the Japanese economy is being supported by mechanically reproduced art that is treated as absolutely authentic by its legions of fans. In fact the difficulty of applying Benjamin’s rule to anime and manga is part of what makes them so interesting as artforms.

None of this is to say the Ghibli filter is a good thing. It feels crass at best, and thievery at worst: again, where did that training data come from? It throws all of the usual arguments about AI into sharp focus. Should the owners of the training data be compensated? Is it putting creators out of jobs? Is it compromising the meaning of the original works? Etc., etc.

But the push-pull over what’s original or not shows that “faux Ghibli on demand” isn’t the pivotal moment either OpenAI would have us believe, or that the critics of AI think it is. Striking as they may be in the moment, Ghibli filter images aren’t any more profound than an old Snapchat filter. They’re content with a lowercase c, low-effort, low-hanging fruit of the kind that snatches mindshare for a second until the next shiny thing comes along. In other words, more slop.

Above all, the Ghibli filter is yet another example of Silicon Valley’s longtime fascination with Japan, and of OpenAI’s repeated efforts to soften its image by association with Japanese imagery. The association with Studio Ghibli is the company’s most successful such attempt to rehabilitate its reputation to date. Whether it works in the long run remains to be seen, but one thing is for sure: this will not be the last time we see AI and anime making headlines together.