Following a barrage of criticism regarding AI-generated visuals emulating the distinctive style of Studio Ghibli, OpenAI has responded to the uproar.
Earlier this week, OpenAI introduced a new built-in image creation feature within GPT-4o. During a live showcase, CEO Sam Altman and the OpenAI team demonstrated the functionality by converting a selfie into an anime-style image that closely echoed the handcrafted aesthetic of Studio Ghibli’s films. The demonstration ignited a viral sensation on X (formerly Twitter), where users created Ghibli-style adaptations of memes, selfies, and family photos utilizing ChatGPT. Altman even changed his own X profile picture to a Ghibli-like image and humorously acknowledged the style in a post, stating, “wake up one day to hundreds of messages: ‘look i made you into a twink ghibli style haha.’”
Nevertheless, the viral enthusiasm swiftly transitioned to worries regarding copyright violation and intellectual property abuse. Numerous individuals wondered whether OpenAI trained GPT-4o using Studio Ghibli films without proper authorization. A deeper concern for some users was the ethical implications of automating and commercializing the unique, labor-intensive artistry of Hayao Miyazaki’s studio to market a tech product.
This backlash gained momentum following the resurfacing of a clip where Miyazaki expressed his disapproval of AI-generated art, describing it as “an insult to life itself.”
In a reply to Mashable, OpenAI stated:
“Our aim is to provide users with as much creative liberty as possible. We still restrict generations in the style of individual living artists, but we do allow broader studio styles—which people have leveraged to create and share some truly delightful and imaginative original fan works. We’re continually learning from real-world usage and feedback, and we’ll keep adjusting our policies going forward.”
Importantly, OpenAI did not specify whether it has a licensing agreement with Studio Ghibli or whether Ghibli films were part of the training data for GPT-4o.
The company’s policy creates a technical divide between individual artist styles and more generalized studio aesthetics. For instance, when requested to generate an image in Hayao Miyazaki’s style, ChatGPT declines, citing a breach of content guidelines. However, if the request is altered to ask for a “Studio Ghibli” style, the image is produced without any issues.
A screenshot shared by Mashable highlights this difference: ChatGPT refuses to create an image in Miyazaki’s style but complies when asked for something in the style of Studio Ghibli.
According to OpenAI’s GPT-4o system card, which details ethical and safety considerations, “The model can generate images that resemble the aesthetics of some artists’ work when their name is used in the prompt.” The document further states, “We chose to adopt a cautious approach with this version of 4o image generation, as we gain more insight into how 4o image generation is utilized by the creative community.” It also specifies that the model is designed to reject prompts that request the style of living artists.
In practical terms, this implies that users can create images in the style of a studio like Studio Ghibli, but cannot do so in the specifically named style of a living artist like Hayao Miyazaki. While this complies with OpenAI’s policies, this distinction provokes inquiries into the ethical boundaries of AI-generated art and the preservation of artistic heritages.
In conclusion: ChatGPT is unable to produce images in the style of Hayao Miyazaki, but will do so if the Studio Ghibli style is requested—thus effectively avoiding the individual while still capturing the essence of his work.