The capabilities of AI-generated images have evolved dramatically from simple abstract designs and glitchy combinations. OpenAI’s latest offering, “Images for ChatGPT,” showcases remarkable proficiency in capturing depth, shadows, and even textual elements. This new tool has sparked a surge in users attempting to replicate the beloved artistic style popularized by Hayao Miyazaki and Studio Ghibli. The familiarity of this aesthetic, characterized by its soft and comforting visuals, has been prevalent online, particularly seen in various memes and artwork around the internet.
The trend began with more wholesome creations. Couples were seen transforming their portraits, pet owners were generating playful images of their cats, and many users took to Ghibli-fying their families (some, like myself, opted to stick to selfies to avoid sharing family likenesses with a third-party system). This movement is reminiscent of the custom art commissions available on platforms like Etsy, where users can request similar stylings of themselves and their loved ones.
However, the trend quickly took a chaotic turn. Users began to create artworks depicting notorious historical moments, including the Twin Towers on September 11 and JFK’s assassination. Notable figures like Nvidia CEO Jensen Huang and former President Donald Trump have also been reimagined in this whimsical Ghibli style, leading to mixed reactions. OpenAI’s CEO Sam Altman joined in the fun by changing his profile picture to a Ghibli rendition of himself and encouraging others to do the same.
According to OpenAI, the model possesses guidelines that allow it to create images inspired by various artists’ styles, as long as users specify them in their prompts. The company noted, however, that it has safeguards in place to prevent images from being generated in the style of living artists, Miyazaki among them.
OpenAI spokesperson Taya Christianson elaborated that while the model is not restricted from depicting adult public figures, they apply similar safeguards to those used when editing photorealistic images. She noted that artistic representation of violence within creative contexts is typically permitted to foster artistic expression.
“Our aim is to grant users extensive artistic freedom. While we prohibit generating in the specific style of individual living artists, we do allow for broader studio styles, which can lead to some charming fan creations,” Christianson asserted in an email.
These guidelines do not, however, permit unrestricted photo transformations. Christianson reminded users that they should only upload images that they own or have the rights to utilize.
Despite the flexibility of these guidelines, users have still generated controversial images such as Hitler in Paris and Korean-American business owners during the 1992 Los Angeles riots. A dedicated website has emerged, compiling numerous AI-generated images, many of which feature altered representations of public figures.
While ChatGPT does block requests for direct photo generation of public figures or copyrighted characters, users have found ways around these restrictions by describing characteristics or requesting similar but fictional representations. The guidelines emphasize the lack of tolerance for actions that infringe on others’ rights, with mechanisms in place for reporting infringement.
OpenAI has faced scrutiny over copyright issues, with several celebrities, including Scarlett Johansson, taking legal measures against the company for likeness-related disputes. Numerous lawsuits have also been filed by authors and artists alleging that their work was improperly utilized to train AI models. OpenAI has been reticent about disclosing the specific data sources used in its training, although in legal arguments, it maintains that its methods transform the original content into something entirely new rather than simply reproducing it.
In addition to Miyazaki, other prominent creators have seen their styles emulated, with users rendering artworks in the likenesses of characters from shows like Rick & Morty, Wallace and Gromit, and The Simpsons.
It remains uncertain how animators from shows such as Rick & Morty and The Simpsons feel about their creations being reinterpreted without consent, or whether they have concerns about how the AI obtained the data necessary for these transformations. ChatGPT’s product lead had previously indicated that users retain ownership of generated images and may use them in accordance with the platform’s guidelines.
Miyazaki has been vocal about his opposition to AI-generated art. Prior to the surge of generative AI, he expressed strong negative sentiments regarding a clip created through deep learning, asserting that it was an affront to the essence of creativity. He stated, “I am utterly disgusted,” and signaled his unwillingness to integrate such technology into his artistic practices.
Ironically, despite his reservations, users on social media have also attempted to Ghibli-fy Miyazaki himself in this wave of AI-generated art.
