Ghibli-style AI art ‘melting’ GPUs saga continues. Just days ago, the excitement surrounding ChatGPT’s enhanced and user-friendly image generation features prompted OpenAI to impose a temporary cap on requests. In a Twitter update, OpenAI CEO Sam Altman wrote just a say after GPT-4o was rolled out, “It’s awesome to see people enjoying images in ChatGPT, but our GPUs are overheating. We’re rolling out some temporary rate limits while we optimize things—shouldn’t take too long! Soon, the free tier of ChatGPT will get 3 generations daily.” While Altman didn’t detail the specifics of the limit, he expressed hope that it would be a short-term measure as OpenAI works to manage the demand deluge more efficiently.
Days later, the GPU “meltdown” narrative continues with a fresh update. This time from Rohan Sahai, product lead for Sora—OpenAI’s video generation tool built by the ChatGPT team. OpenAI Sora, which crafts videos from text, images, and video inputs, too seems to be feeling the ‘heat’. Sahai tweeted, “The demand for 4o image generation has been unreal. It’s been a blast watching the Sora feed overflow with amazing creations. Sadly, our GPUs are frying in Sora territory too, so expect longer wait times or capacity hiccups in the days ahead. We’re hustling to get things back on track.”
Sahai quoted CEO Altman’s earlier post about Ghibli-style AI art straining OpenAI’s systems. The company is reportedly tapping its GPT-4o model to power this upgraded image generation tool, yielding visuals that are strikingly lifelike—though users can steer them in countless creative directions. The model has also tackled previous challenges, like rendering text, with notable success.
When ChatGPT turned everything into Studio Ghibli art
The Ghibli AI art phenomenon kicked off earlier this week when ChatGPT began transforming everything into Studio Ghibli-inspired art. OpenAI’s “Images for ChatGPT” feature nails depth, shadows, and even text with eerie precision, sparking a wave of users mimicking Hayao Miyazaki’s iconic Studio Ghibli style. Already a widespread favorite online for its soothing, gentle vibe (think Lofi Girl), this aesthetic has now become an effortless, automated trend.
Predictably, the craze quickly spiraled into wild territory. Internet users have reimagined everything from the Twin Towers on 9/11 and JFK’s assassination to Nvidia CEO Jensen Huang signing a woman’s chest, Sam Altman’s congressional testimony, and even the Babri Masjid demolition — all infused with Ghibli’s signature charm. Altman has leaned into the hype, swapping his Twitter profile picture for a Ghibli-fied version of himself and inviting followers to craft him a new one.
Interestingly, the rules for Images with ChatGPT allow this kind of experimentation. Per the GPT-4o system card, “The model can mimic the look of certain artists’ styles when their names are included in prompts.” It includes a safeguard: Requests to replicate the style of living artists are blocked. However, as ‘The Verge’ points out that it is worth noting that Hayao Miyazaki of Studio Ghibli is still alive.