While I was offline for a while (working, coding and studying) a bunch of AI guys and anti-AI folks on social media became overnight experts in Studio Ghibli after OpenAI’s image generator let ChatGPT users generate images in the Ghibli style, or rather what westerners think is the Ghibli style.
The AI guys insisted this was a “game changer” and soon anyone would be able to generate any animation film from the comfort of their bed, while the anti-AI side overreacted by mislabelling videos of Miyazaki talking about computer animation to defend traditional animation.
I need to raise some points.
1. Unlike these overnight experts, I support the industry with hundreds of dollars worth of disc and book purchases every year. In the last year I spent about $1500 with ‘The Boy and The Heron’ blu-ray, art book and storyboards costing $300. If you support a filmmaker but you only watch their movies on a streaming service once every two years, you’re not supporting them. It seemed clear from the arguments I was reading that the AI side and anti-AI side had barely ever spent any money supporting the animation industry. They were just vying for attention on social media.
Recent purchases. Art books and storyboards should be your main source for studying animation.
2. Sites like Civit.ai have hosted downloadable LoRAs for generating Miyazaki-like images on your computer for almost two years. Where was the mass outcry then? What you saw in the last month from OpenAI and the media was marketing. You were emotionally and psychologically manipulated to think something new and threatening happened.
3. Generating images in someone's style doesn't mean anyone is going to respect you for it or let you commercialise it. At the point of sale copyright law will always apply. If you try to sell t-shirts with generated Totoro images a lawyer from Japan will reach out to you. They issue takedowns on eBay and YouTube all the time.
4. Commenters don’t know the difference between Miyazaki and Studio Ghibli. The studio doesn’t have a set style and employs various directors. Miyazaki’s style has changed over the years. He has said his influences have come from Jean Giraud to Roald Dahl and from Sanpei Shirato to Osamu Tezuka (who was influenced by Disney). One of Miyazaki’s most successful films was 'The Castle of Cagliostro', a departure from the style he is famous for.
5. A video of Miyazaki attacking AI was mislabelled. Even though he would feel the same way about gen AI, the video was actually about scripted animation algorithms used in 3D CGI (often used in gaming). The video predates the deep learning models we see today. A different kind of AI.
6. Gen AI is no threat to people like Miyazaki for a reason anyone intelligent can understand. Audiences go to the cinema or buy discs to support them because of the sheer amount of muscular and intellectual effort that is needed to make a movie like ‘The Boy and The Heron’. They won't spend money on a LinkedIn bro who generated a semi-compelling moving slideshow from his sofa. Talent matters to audiences just like it matters to sports spectators. Don't let anyone tell you it doesn't matter.
7. Don't be tricked by clips on social media. Even if Gen AI did meet the very high requirement of filmmaking and animation, it will struggle to find an audience beyond a niche. For most there’s no reason to watch automated videos. It doesn't inspire. AI videos won’t generate the kind of revenue that cinemas need to stay open. That's the key. You have to invest a lot of money to make the kind of revenue that all the connected sectors require.
So ignore the uneducated opinions that gen AI will replace all filmmaking and animation. It won’t because it can’t. Language models struggle badly enough as a coding assistant in their own domain of computer science. Ask ChatGPT to generate a Photoshop clone or refactor a 100,000 line codebase and watch it fall over itself. The language models will tell you their limits themselves.
Before I end this post I want to highlight a thread I saw on the Stable Diffusion sub-Reddit. They were experimenting with the latest models to "generate animé" and even though the models appeared to output impressive images (with minimal movement), the users themselves were very critical of AI and understood that it will never compare to crafting true animation by hand. Here’s some quotes:
