How Ghiblification Became AI’s Big Style Meme

On March 25, 2025, OpenAI released native image generation for GPT-4o.

By the next day, the internet had found one of its favorite ways to test it.

Not text rendering.

Not product shots.

Not diagrams.

Studio Ghibli.

People uploaded wedding photos, pets, movie stills and old memes and asked ChatGPT to remake them with the soft colors, expressive faces and hand-drawn visual cues associated with the Japanese animation studio.

Within days, there was already a verb for it.

Ghiblify.

And once the internet gives a product behavior its own verb, the feature demo is probably over.

You have a culture event.

It worked because you already knew the picture

AI image generation had gone viral before.

But Ghiblification had an unusually simple before-and-after.

You did not have to decide whether the machine had created a “good image” from scratch.

You already knew what you were looking at.

There was the original family photo.

And there was the same family, still recognizable, translated into a visual world that felt like animation.

That made the result almost perfectly suited to social media.

The comparison told the story instantly.

OpenAI’s new system was designed to work from uploaded images while preserving important details about the source. OpenAI launched the feature on March 25; Axios was reporting a flood of Ghibli-style portraits the following day, and other outlets quickly documented the same wave.

Why Ghibli became the dominant style of the moment is harder to prove precisely.

But the appeal was visible in the results.

Your dog was still your dog.

Your vacation photo was still your vacation photo.

It had simply acquired better clouds.

Then the internet conjugated it

The clearest sign that the trend had escaped the prompt box was linguistic.

People stopped describing what they were doing and started giving it ordinary grammar.

Ghiblify this.

I Ghiblified it.

Here is the Ghiblified version.

Wiktionary documents Ghiblify in a New York Times quotation from March 27, only two days after the feature launched. It records Ghiblification in published coverage from March 28 and derives the noun from Ghiblify + -ication.

That speed is part of what makes the episode interesting.

Most product features keep product language.

You “use image generation.”

You “apply an effect.”

You “create an edit.”

Ghiblification acquired its own language almost immediately.

And the word itself helped.

Ghiblification sounds less like a technical workflow and more like something spreading through the population.

For a few weeks, that was not a terrible description.

First the selfies. Then everything else.

The obvious use was personal.

Upload a photo of yourself, your partner, your children or your dog and see the alternate-universe animated version.

Then the internet did what it usually does with a reproducible format.

It stopped respecting the original use case.

People Ghiblified famous memes.

Film scenes.

Politicians.

Historical photographs.

Images from serious news events.

At that point, Ghiblification was no longer only an aesthetic preference.

It had become a meme format.

The underlying idea was simple:

What would this look like if the machine Ghiblified it?

Once a format reaches that stage, almost any recognizable image can become input.

The source supplies the context.

The visual style supplies the contrast.

The AI supplies the speed.

It also showed what had changed about AI images

Earlier image generators could produce striking pictures, but they often felt like machines you visited to ask for something new.

Ghiblification pointed toward a slightly different relationship.

The generator became an editor of your existing visual world.

Your photo went in.

Your meme went in.

A familiar public image went in.

And something recognizably connected came back.

That difference matters culturally.

Generating “a beautiful anime village at sunset” can be impressive, but the result still feels mostly like the model’s picture.

Transforming the photograph already sitting in your camera roll is personal.

The technology touches your memories, your jokes and your identity.

That made the feature easy to show to people who did not care about benchmarks.

Nobody needed a chart explaining why it was interesting.

They could Ghiblify their cat.

Then came the uncomfortable part

Studio Ghibli is not simply a bundle of visual characteristics.

It is the work of human artists associated with decades of painstaking animation.

So almost as soon as the images went viral, another question arrived:

Should an AI system be able to imitate a recognizable artistic style this easily?

Artists and commentators raised questions about training data, consent, creative labor and whether existing copyright law adequately addresses style imitation.

Nature Electronics later used Ghiblification as a case study in the tension between generative AI, artistic practice and copyright frameworks.

This is also where one piece of internet context often gets mangled.

A clip of Studio Ghibli co-founder Hayao Miyazaki condemning an AI-generated animation experiment as an “insult to life itself” resurfaced during the trend.

But the clip was from 2016 and referred to a different AI/CGI demonstration.

It was not Miyazaki reacting to ChatGPT’s 2025 Ghiblification wave.

The older quote fit the moment so neatly that the internet supplied it with a new context anyway.

A meme, a demo and an argument at the same time

The strongest internet phenomena often do several jobs at once.

Ghiblification was playful enough to be a meme.

Personal enough to make people try the product.

Technically impressive enough to become an unofficial demonstration of GPT-4o image generation.

And controversial enough to become a shorthand example in debates about AI and art.

Its lasting cultural contribution may be smaller and stranger:

it gave AI image transformation an everyday verb.

For a brief period, you did not need to explain what the new image model could do.

You could just say:

Ghiblify it.

And people knew what you meant.