Nanobanana2Nanobanana2

ChatGPT Ghibli Style: Essential Copyright Guide

Nanobanana2 TeamMarch 28, 2026

When OpenAI launched its GPT-4o image generator in March 2025, "Ghibli-style" images flooded every social platform within 48 hours. Even Sam Altman changed his X profile picture to a "Ghiblified" selfie. But behind the viral fun, a serious copyright storm was brewing — and it hasn't stopped yet.

Japan's Content Overseas Distribution Association (CODA), representing Studio Ghibli and major publishers, formally demanded OpenAI stop training on their copyrighted content (TechCrunch, 2025). Japanese lawmakers called it "a violation of the law." This article unpacks what happened, what it means legally, and why every AI creator should pay attention.

Key Takeaways

  • ChatGPT's Ghibli-style trend went viral in March 2025, triggering copyright backlash from Studio Ghibli and Japanese publishers
  • Copyright law generally doesn't protect a "visual style" — only specific characters, scenes, and compositions (TechRadar, 2025)
  • Artists Karla Ortiz, Sarah Andersen, and Kelly McKernan won the right to pursue copyright claims against Stability AI, Midjourney, and others (Artnet, 2024)
  • The legal landscape is still evolving — no definitive ruling exists yet on AI training and copyright

What Happened With ChatGPT's Ghibli-Style Image Explosion?

Within days of GPT-4o's image generation launch, millions of users discovered they could prompt ChatGPT to recreate photos in Studio Ghibli's signature style — dreamy watercolor backgrounds, oversized expressive eyes, and whimsical natural settings (Fortune, 2025). The trend spread so fast that "Ghibli AI" became a top search term globally.

Hayao Miyazaki, Studio Ghibli's legendary co-founder, has been vocal about his stance for years. He once described AI-generated animation as "an insult to life itself." The viral trend effectively turned his life's work into a free filter, and neither he nor the studio consented.

OpenAI stayed quiet initially. But the backlash built fast: Japanese lawmakers began drafting responses, CODA sent formal letters, and copyright lawyers started publicly analyzing the legal exposure.

Here's the uncomfortable truth for artists: copyright law, as it currently stands in most jurisdictions, does not protect a visual style (TechRadar, 2025). The dreamy backgrounds, oversized eyes, and whimsical creatures that define Ghibli's aesthetic aren't individually copyrightable.

What IS protected:

  • Specific characters, Totoro, No-Face, Ponyo are copyrighted
  • Specific scenes and compositions, A recognizable frame from Spirited Away is protected
  • Specific artworks, Individual illustrations and film stills

What ISN'T protected:

  • General aesthetic, "Looks like Ghibli" isn't infringement on its own
  • Color palettes, Warm sunset tones with soft greens aren't ownable
  • Artistic techniques, Watercolor-style rendering isn't copyrightable

This means if ChatGPT generates an image that "feels like Ghibli" without reproducing specific copyrighted elements, it may be legally defensible. But "legally defensible" and "ethically acceptable" are very different standards.

What's Happening in the Courts Right Now?

The most significant case is Andersen v. Stability AI, filed in 2023 by artists Sarah Andersen, Kelly McKernan, and Karla Ortiz against Stability AI, Midjourney, DeviantArt, and Runway AI (Artnet, 2024).

In August 2024, U.S. District Judge William Orrick ruled that the artists could proceed with their copyright claims. He found they had "reasonably argued" that the AI companies violated their rights by illegally storing copyrighted work, and that Stable Diffusion was built "to a significant extent on copyrighted works."

Karla Ortiz testified before the U.S. Senate in 2023: "I have never been asked, never been credited, never been compensated" for her work being used in AI training data.

The case is still ongoing. Legal experts anticipate its outcome will set precedents that shape AI copyright law for years to come.

Japan's Regulatory Response

Japan has historically had more permissive AI training laws, allowing copyrighted material to be used for machine learning under certain conditions. But the Ghibli backlash is changing the political calculus.

CODA's formal request to OpenAI represents a shift in Japan's approach. Lawmakers in the Japanese Diet have called for tighter restrictions. If Japan, one of the most AI-friendly regulatory environments, starts tightening rules, it signals a global trend.

How Should AI Creators Navigate This Gray Zone?

Until courts deliver definitive rulings, here's a practical framework:

Lower risk:

  • Prompt for a general aesthetic ("watercolor illustration style, warm colors, soft lighting")
  • Use style descriptions rather than artist or studio names
  • Generate original compositions that don't reference specific copyrighted scenes

Higher risk:

  • Prompt for a named artist's style ("in the style of Hayao Miyazaki")
  • Generate images that closely resemble specific copyrighted characters
  • Use AI-generated images commercially without checking for copyrighted elements

Best practice:

  • Credit your tools transparently
  • Don't claim AI-generated art mimicking a specific artist as your original work
  • If you're using AI images commercially, consult a copyright lawyer, seriously

What Does This Mean for the Future of AI Art?

The Ghibli controversy isn't an isolated incident, it's a preview of a recurring pattern. Every time an AI model gets good enough to convincingly mimic a recognizable style, the same tensions will surface.

Three possible outcomes are taking shape:

  1. Licensing models, AI companies pay artists or estates for training data rights (some deals are already happening with stock photo companies)
  2. Opt-out systems, Artists can formally exclude their work from AI training datasets (technically complex but legally demanded)
  3. Style attribution, Generated images include metadata about which training data influenced the output

The AI image generation market is projected to reach $30.02 billion by 2033 (SkyQuest, 2025). With that much money at stake, the legal frameworks will eventually catch up. The question isn't whether regulation is coming, it's how harsh it will be.

Related Resources on Nano Banana 2:

Frequently Asked Questions

Currently, generating images in a general "Ghibli aesthetic" likely doesn't violate copyright because visual styles aren't protected under most copyright law (TechRadar, 2025). However, generating specific copyrighted characters like Totoro or scenes from Spirited Away could constitute infringement. The legal landscape is still evolving.

Has Studio Ghibli sued OpenAI?

As of April 2026, Studio Ghibli has not filed a lawsuit against OpenAI. Japan's CODA sent formal demands to OpenAI to stop training on their content, and Japanese lawmakers have pushed for legislative action, but no court case has been initiated by Ghibli directly (TechCrunch, 2025).

Can AI companies legally train on copyrighted images?

This is the central question in Andersen v. Stability AI. Judge Orrick ruled in 2024 that artists can proceed with claims that AI companies infringed their copyrights through training data usage. No final ruling has been issued yet, making this one of the most closely watched tech cases in copyright law (Artnet, 2024).

Use generic style descriptions instead of naming specific artists or studios. Avoid generating recognizable copyrighted characters. For commercial use, audit your outputs for similarity to known copyrighted works. Tools like Nano Banana 2 let you upload your own reference images as style guides, reducing dependency on any single artist's copyrighted aesthetic.

What did Hayao Miyazaki say about AI art?

Miyazaki has called AI-generated animation "an insult to life itself" and expressed deep opposition to AI involvement in creative work. His stance reflects a broader philosophical objection: that art requires human intention, suffering, and lived experience, qualities that AI fundamentally lacks (Fortune, 2025).

ChatGPT Ghibli Style: Essential Copyright Guide