'Our GPUs are melting': OpenAI's Sam Altman after GPT-4o brings Ghibli-style creations to life

In a viral wave sweeping across social media, OpenAI’s latest AI model, GPT-4o, is capturing attention for its ability to generate Ghibli-style art from simple text prompts. This new feature, dubbed ‘4o Image Generation,’ has taken the internet by storm, with users enthusiastically uploading images to be transformed into enchanting anime-style portraits reminiscent of Studio Ghibli’s timeless animation.

The sudden surge in interest is largely due to GPT-4o’s ability to instantly replicate Ghibli’s distinct art style—characterised by dreamy landscapes, delicate pastel tones, and meticulous details. The nostalgia-driven trend has drawn interest from both casual users and brands looking to tap into the hype for promotional content.

Named after the renowned Japanese animation studio Studio Ghibli, founded by Hayao Miyazaki, Isao Takahata, and Toshio Suzuki in 1985, Ghibli art is known for its hand-drawn animation with rich storytelling and beautifully detailed visuals. Its signature style features soft, muted colour palettes and whimsical, slightly surreal aesthetics that evoke a dreamlike quality.

The AI-driven Ghibli art craze has seen a surge of popularity, especially in the United States, where an estimated 72% of the population watches anime daily.

OpenAI’s CEO Sam Altman joined the trend, changing his X (formerly Twitter) profile picture to a Ghibli-style portrait of himself. 

His participation in the trend further fuelled its popularity, with users now sharing their own AI-generated Ghibli-style images across various platforms.

Part of the appeal lies in how simple the process is. Users and AI enthusiasts are sharing various prompts to generate stunning Ghibli-style images, such as:

    •    “Transform photos into Studio Ghibli-inspired characters, with a focus on detail and whimsy.”
    •    “What would I look like as a Ghibli character?”
    •    “Can you Ghiblize my photo?”
    •    “Show me in Studio Ghibli style.”

The model’s ability to accept detailed prompts and customise the output with specific colours, styles, and even transparent backgrounds has enhanced the creative experience.

Unlike previous image-generating tools relying on external platforms like DALL-E, GPT-4o’s image creation abilities are fully integrated into its interface. This means users can request highly customised visuals directly within the model without relying on external software.

However, the AI-generated images from GPT-4o take longer to render due to their higher level of detail, with some images taking up to 2 minutes to process.