[Review] "Done in Seconds" Family Photos Turned into Ghibli Masterpieces by AI Fingertips

From Ghibli to Tim Burton
Utilizing 'ChatGPT-4o' Image Generation
Korean Text Rendering Remains a Weakness

Family photo uploaded to ChatGPT transformed into animated styles. From the top left clockwise, the styles are Ghibli, Simpson, Conan, and Doraemon. (Photo by ChatGPT)

OpenAI's ChatGPT is captivating users worldwide with a ‘Ghibli vibe.’ At the heart of this craze is the recently introduced image generation feature in ‘ChatGPT-4o.’ People are fascinated by the AI’s ability to instantly create fairy-tale-like pictures and comics with just simple command inputs. This feature, which transforms user-uploaded photos into the style of famous animations, gained popularity immediately upon its release on the 25th. Sam Altman, CEO of OpenAI, joyfully exclaimed, "On the 31st, the number of ChatGPT users increased by one million in just one hour. The GPUs are melting down."

Uploaded a Photo... "This is the Ghibli Vibe"

"Can you turn an old family photo filled with childhood memories into a Ghibli style?" I handed ChatGPT a family photo taken over 20 years ago and typed, ‘Make it into a Ghibli-style picture.’ A simple user interface appeared on the screen. Without complicated settings or technical terms, all I had to do was attach the photo and type my request. I pressed Enter and waited a moment. In less than a minute, the screen displayed a result as if Hayao Miyazaki himself had visited. The picture perfectly reimagined the original photo’s subjects and composition in Studio Ghibli’s warm and nostalgic animation style.

The grass background, which looked like it was casually taken with a phone, transformed into vibrant pastel tones full of life, the family members’ eyes grew larger, and gentle smiles were added to their expressions. The blue sky and leaves even gave the illusion of a scene from ‘My Neighbor Totoro’ or ‘Spirited Away.’ In just a few minutes, an ordinary everyday record was transformed into a fairy-tale-like work.

It’s not just the Ghibli style. I decided to try converting the same photo into more animation styles. I requested, "Make it Disney style," "Change it to The Simpsons style," "Express it in Sailor Moon style," and "Draw it in Conan animation style."

ChatGPT-4o successfully fulfilled all these requests. It captured Disney’s vivid and bright colors, The Simpsons’ distinctive yellow skin and exaggerated features, Sailor Moon’s large, sparkling eyes and sleek lines, and Conan’s sharp contours and serious atmosphere. It accurately grasped the characteristics of each animation style and transformed the image accordingly. This demonstrates that ChatGPT’s image generation ability is not just a simple filter effect but has reached a level where it understands the essence of each style and can apply it anew.

"So, what am I really doing now?"

A 4-panel comic in the style of Studio Ghibli requested by the reporter.

After successfully converting images into the Ghibli style, I tried making a popular 4-panel webtoon. While wondering what to draw, a sudden idea came to mind. "I want to draw a 4-panel comic about people who draw 4-panel comics using ChatGPT these days. Write me a scenario."

With this single request, ChatGPT instantly completed the storyline. It started with the protagonist sitting in front of a laptop thinking, "Shall I try drawing a 4-panel comic?" and "I’ll leave the storyboard to ChatGPT," then showed a scene of the protagonist discussing the topic with ChatGPT. In the third panel, the AI suggests, "How about a 4-panel comic about people drawing 4-panel comics with GPT?" In the last panel, the protagonist wears a hollow expression, saying, "So, what am I really doing now?"

This ironic situation, where humans entrust creativity to AI, are amazed by the AI’s output, and ultimately ask themselves "What am I really doing?" is expressed by the AI itself.

Based on this storyline, I asked ChatGPT, "Draw a Ghibli-style 4-panel comic," and the AI immediately got to work. Remarkably, within just a few minutes, a comic was completed in Studio Ghibli’s characteristic warm and soft colors and drawing style. The beige background looked as if drawn on old paper, with delicate shading and perfect expression acting. However, the Korean text rendering was still a weak point, with some broken characters visible here and there.