In The Loop Episode 10 | Does ChatGPT's Viral Image Generator & The Ghibli Craze Spell The End Of Art & Creativity?

The sudden release of what can only be described as a magical image generator by ChatGPT has left the world stunned. Art styles that were once unattainable unless you spent years practicing and mastering them are now available to everyone with just a couple of prompts.

This phenomenon—what some are calling the Ghibli-fication of everything—lets users transform their photos into Studio Ghibli-style images. It highlights the playfulness and delight of this new technology, but also its darker side. Beneath the novelty lie legal, economic, and cultural questions that demand serious attention.

In today’s episode, I’m going to unpack the Ghibli-fication of everything. I’ll explain how the model works and explore the ways it might affect creative industries and everyday life.

This is In The Loop with Jack Houghton. Thanks for listening—and enjoy the show.

‍

‍

‍

OpenAI’s new ChatGPT image generator

So, OpenAI recently released a major update to ChatGPT. They've integrated one of the most advanced image-generation models directly into GPT-4’s interface. And... wow. It’s good. For those who’ve tried it, this feels like the first real “whoa” moment in AI in a long time.

I’ve spent hours—probably too many—uploading photos of friends and pets, asking it to turn them into a World War II soldier, a 1970s hippie, you name it. Eventually, it even stopped me from making more images because I'd hit the limit. That’s how addictive it is.

One of the things that the new GPT-4 image generator really unlocked was this Ghibli-ification of everything. Social media has been flooded with Studio Ghibli’s iconic animation style. Some of it looks like it could’ve been pulled straight from Spirited Away.

What makes this different from previous AI image tools or internet trends is just how easy it is. You can upload an image into GPT and simply say, “Make this in the Studio Ghibli style.” And it works.

It’s gone so viral that even Sam Altman, OpenAI’s CEO, changed his Twitter profile picture to a Ghibli-style portrait.

But underneath all the fun—and yeah, I’ve had a lot of fun with it—there are serious questions. This trend exposes real tension between imitation and creativity, automation and authorship, and a whole lot of gray areas in the law.

So what’s changed under the hood to suddenly let people create Studio Ghibli-style images so easily?

Let’s dig into that.

‍

‍

What makes the new ChatGPT image generator so special?

Previously, generating an image was a two-step process: 1) You’d type a prompt into ChatGPT to help describe an image; 2) then switch to a separate tool like DALL·E or Midjourney to actually create it. It was a bit of a pain.

Now, GPT’s new model merges everything into one interface. It’s incredibly simple — you can refine and iterate on the images directly. This represents a shift from the older, diffusion-based methods of image generation.

‍

Moving away from “denoising”

Earlier platforms like DALL·E, Midjourney, and Stable Diffusion relied on a process called “denoising,” which means they’d start with random pixels—visual noise—and slowly clean that up to match your prompt. Imagine an old-school TV screen full of static, gradually morphing into a clear image.

GPT-4 works differently. Instead of starting with noise, it builds the image step by step—more like writing a sentence than sculpting from chaos. That makes it much better at following instructions, especially for tricky things like rendering text accurately in an image—something previous models were notoriously bad at.

‍

Conversational design

Another big upgrade is the ability to have a conversation with the model. You can say things like “add a hat,” “make the lighting softer,” or “change the text,” and it updates the image immediately—without losing the original style or forgetting what it made before.

People believe it’s using an autoregressive architecture, which is what allows it to build images sequentially, step by step.

Another important new feature is multi-turn refinement. You can ask it to redraw parts of an image. You can even highlight a section with your finger on your phone, and it will rework just that area while keeping the overall structure and style consistent.

This conversational approach to image generation feels more like you’re directing a person than feeding prompts into a machine.

‍

Image referencing

There's also the ability to reference actual images. So if you’re after that Studio Ghibli look, you can show it an example. Or you could upload a photo of your living room and say, “Rearrange this in a Scandinavian style with a big window—move the chairs to the corner.”

That opens up huge potential for interior design, mockups, storyboarding, and even prototyping products.

‍

Accurate text rendering

The improvements in text rendering are also game-changers. If you’ve used these tools before, you’ll know how bad they were at generating readable text. Because diffusion models start with noise, the output often looks garbled. GPT’s new model is much better — it can render full sentences and paragraphs clearly.

So now you can create infographics, signage, t-shirts, speech bubbles in comics—anything that involves text. This opens up a ton of commercial possibilities, from brand mockups to flyers to social media content with minimal back-and-forth.

‍

All these new features have sparked an explosion of new use cases. Marketing agencies can create entire campaign mockups. Product sellers can generate realistic product shots in different environments. Families—myself included—are turning photos into Ghibli-style portraits or fun historical reimaginings.

Even educators might use it to bring historical scenes to life or turn a sketch of a scientific concept into a polished digital visual.

This update has brought a real sense of magic back to AI. It’s suddenly fun again to make things—to send in a photo and get back something great with hardly any effort. And because it’s so good, it’s going to have a huge impact.

‍

‍

ChatGPT image generator copyright, cultural, and ethical implications

Let’s take the Studio Ghibli trend as a kind of case study—a lens to explore the legal, ethical, and cultural implications of this new technology.

Studio Ghibli’s art is deeply personal to its co-founder, Hayao Miyazaki. He once called AI-generated animation “an insult to life itself.” His films are handcrafted, layered, philosophical—years in the making. He still draws every day. They are the antithesis of automation.

So when thousands of AI-generated images start circulating in the name of Studio Ghibli, it’s no surprise that many people are uncomfortable. It’s not just about aesthetics—it’s about authenticity.

These aren’t reinterpretations by human artists paying tribute. They’re algorithmically generated imitations, and it’s very likely they’re trained on thousands of actual Ghibli film frames. That’s where the discomfort really sets in.

Still, many have defended the trend. To them, the “Ghibli-ification” of photos is no different from remix culture or cosplay—playful, participatory, and creatively harmless. You could argue that reimagining your dog as a Ghibli character or transforming a wedding photo democratizes creativity. It gives professional-grade tools to everyone. And crucially, most of these images aren’t being monetized—they’re shared for fun, not profit.

But culture doesn’t exist in a legal vacuum. And this is where things enter a gray area.

The heart of the debate is this: style isn’t protected by copyright law. Copyright protects specific expressions—not ideas, not aesthetics. That means you can’t copyright the “Ghibli look.” You can only copyright particular characters, scenes, or images.

So an AI-generated image that mimics Ghibli’s style isn’t automatically illegal — unless it starts to replicate protected elements, like Totoro or the bathhouse from Spirited Away. Drawing that line, however, is incredibly difficult.

Generative models don’t copy in the traditional sense. They interpolate—they synthesize outputs based on learned patterns. But, studies have shown that these models can output near-identical images from their training data under certain conditions.

So if someone prompts, “Make a photo of my friend in the style of Ghibli,” and the result mirrors a frame from Princess Mononoke — is that fair use? Or is it infringement?

Some legal experts say that if an AI output includes “substantially similar” elements — like a recognizable character, composition, or scene—it could count as a derivative work and require permission.

But others argue that AI outputs are transformative by nature, and therefore protected under fair use. So far, no court has given a definitive ruling.

What makes the Ghibli case especially tricky is the likely use of its work in training datasets. AI models like GPT are trained on billions of images—many scraped from the web without consent. OpenAI and Google haven’t disclosed what media they’ve used, but it’s safe to assume that Ghibli frames—which are widely available online—were part of that training data.

And given how uncannily accurate the stylistic imitations are, it’s hard to imagine otherwise.

So, it seems like Studio Ghibli should have a strong case. But this is where copyright law meets AI—and it’s anything but straightforward.

Several high-profile lawsuits are testing whether this kind of training is even legal. Artists like Karla Ortiz, Sarah Andersen, and Kelly McKernan are suing Stability AI, Midjourney, and others in a class-action lawsuit. Their claim: that training on copyrighted work without permission is infringement — and that enabling others to generate new work based on their art adds a layer of liability.

So far, the court has allowed the case to proceed. It hasn’t dismissed the core claims around copyright infringement and false endorsement.

Meanwhile, Getty Images is suing Stability AI for allegedly scraping 12 million copyrighted images from its database—and is seeking around $2 billion in damages. Getty called the scraping “pure theft.”

These cases matter. They’ll set a legal precedent.

If courts rule that using copyrighted images for training does violate copyright law—or that prompting an AI “in the style of” an artist implies false endorsement—then current practices in AI art generation could face serious restrictions.

Right now, many AI companies rely on a fair use defense. They argue their models don’t store or regurgitate exact images—and that training is a form of “transformative learning.” But if judges disagree, AI developers may be forced to license training data or offer artists the ability to opt out.

And if that happens, it could fundamentally reshape how large AI models are built. It’s not just a challenge—it’s an existential threat to the current model of scaling AI.

So what’s going to change?

Pressure is mounting from the creative community to protect their work and legacy. Some AI labs are starting to self-regulate. OpenAI, for example, has introduced prompt filtering to prevent users from mimicking currently living artists. But when it comes to studio styles — like Ghibli’s — they still consider that fair game.

That leaves us in a legal and cultural limbo.

If the courts side with artists, we might see studios getting paid to license their work for training—and artists getting credit for their styles. But if not, then “style theft,” as some are calling it, could become fully normalized. And artists will be left to find new ways to stand out in a world where AI can imitate them with ease.

‍

‍

The business implications of ChatGPT’s image generator

So, what’s the real impact on businesses and individuals?

Well, I think studios, agencies, and production companies will see tools like this as a fast, powerful way to produce concept art, background layouts, or even basic animations. These models can generate dozens of variations in minutes—supercharging creativity, saving time, and cutting costs.

But the reality is that junior artists and production assistants will likely feel the brunt of this shift. Instead of hiring early-career creatives to sketch concepts or help with in-between animation frames, some companies will simply automate the process. Freelance illustrators and animators will also feel the pressure. Where once there were commissions for children's books, pitch decks, marketing visuals, or early-stage concept art—AI tools can now handle much of that heavy lifting in seconds.

So many creatives are adapting. They’re repositioning themselves with the mindset: I use AI—I’m twice as fast, twice as productive. Pay me for that.

They’re incorporating these tools into their workflows—using them to generate rough sketches, thumbnails, or to quickly explore different styles. In a way, they’ve become curators as much as creators.

And the same shift is happening with influencers. Some will be replaced by AI-generated personas—fake influencers generating hyper-stylized, perfectly aspirational content that feels just real enough. People follow these accounts not because they’re human, but because they project a lifestyle people want to live.

All of this is happening fast, and it’s forcing adaptation. In Hollywood, unions like the Animation Guild are pushing for contracts that require human consultation before studios use AI—trying to safeguard jobs and artistic integrity.

That broader debate—about ethics, consent, and fair compensation—is just getting started. Many artists are now advocating for strict rules around how their work is used in AI training. From a legal perspective, copyright is being tested in ways no one anticipated—and many are asking questions like how to protect intellectual property from ChatGPT.

Courts and lawmakers now have to wrestle with big, messy questions: Who owns an AI-generated image? What counts as fair use? And can training data be used without permission or payment?

Economically, entire sectors are being transformed. Studios will need to retool their pipelines to incorporate AI. Freelancers may benefit from faster turnarounds—but they’re also facing real threats of job displacement.

Culturally, there’s a risk of homogenization. The same styles, the same aesthetics, endlessly replicated. But on the flip side, this could push people to forge entirely new artistic directions—styles so original that even advanced models struggle to replicate them. And I think, over time, people will crave the emotional depth of human-made work. That could actually strengthen parts of the creative sector.

‍

‍

Closing thoughts

So, the real takeaway? There is no single narrative here. These changes are complex, multidimensional, and still unfolding. The magic of AI image generation is undeniable—it’s captured imaginations everywhere. But with that magic comes pain: copyright battles, job losses, forced adaptation, and cultural shifts happening almost overnight.

For those who create, live off, or champion original work, this is a moment of both anxiety and new possibility. Because the role AI will play in our creative culture isn’t set in stone. It’ll be shaped by legal rulings, business decisions, and the collective values we decide to stand for.

Alright, that’s it for this week. I hope you enjoyed the show. If you did, share it with a friend, a colleague—drop it in that Slack channel if you want people to start learning about AI in under 20 minutes.

Thanks again. See you next week.

‍