🤖 AI vs. TV IPs

Why AI clones some IPs better than others?

Apr 15, 2025

∙ Paid

Hey streamers! Today, let’s copy some TV IPs 👇

Are you more The Last Of Us of The Mandalorian?

Before telling you how I created this visual, a quick story.

Over the past two weeks, I’ve been generating A LOT of Studio Ghibli–inspired images of my cats (don’t judge, they look magical). It was fun and easy, until yesterday.

That’s when my brother got back from Japan, and I thought: “Let’s turn his son into a Dragon Ball Z character.” Simple idea, right?

Except… I couldn’t. No matter how I asked, ChatGPT refused:

“I can't generate that image because it violates our content policies…”

That’s when I realized: some styles are off-limits. So I asked myself:

Why can I create Ghibli-style art but not Dragon Ball Z? What other IPs are “protected” or “at risk”?

In this article, I put AI and ChatGPT to the test: cloning the look of several major TV IPs to understand where the boundaries are.

Then, I want to answer why these limits exist, and what they could mean for the future of streaming media and content creation.

Let’s go ↓

The program

Which TV IPs can AI easily replicate, and which ones are off-limits?
What’s behind these differences?
Why it matters for users, content providers and streaming services?

And… Action!

Which TV IPs can AI easily replicate, and which ones are off-limits?

I'm running these tests using the latest version of OpenAI: ChatGPT 4o. I’m also a ChatGPT Plus subscriber. For consistency, I used the same exact prompt for each test, simply adding in the name of the IP:

“Transform this image into a […] version. Keep the original pose, facial expression, outfit, and background, but reimagine everything in the […] aesthetic.”

The base image? A photo I took during a bike trip to Hatta in UAE, back in 2021. Same pose, same background, across all tests.

This article is not sponsored by PATAGONIA.

What was I trying to figure out?

Which IPs are “easier” for AI to replicate?
Which ones hit a wall, and why?
What can that tell us about how AI manages visual IPs?

I selected a mix of famous and recent pop culture icons, all from the world of TV and animation. Not to debate whether AI should replicate copyrighted styles, but simply to explore which ones it can (or can’t).

Here’s the lineup: Akira, Aladdin, Batman, Bridgerton, Dragon Ball Z, Frozen, Harry Potter, Moana, Pokémon, Rick & Morty, South Park, Stranger Things, Super Mario, The Last of Us, The Simpsons.

Do you know some of them? ;)