In his latest comic “Cyberpunk: Peach John,” manga author Rootport imagines the Japanese folklore hero Momotaro — who is said to have been born from a giant peach — living in a dystopian future. But while the writer created the storyline and dialogue, his sci-fi-inspired imagery was produced entirely by artificial intelligence.
In fact, the 37-year-old has never drawn a comic by hand.
The publishing house behind the work, Shinchosha, believes that “Cyberpunk: Peach John” is the world’s first complete AI manga work. On sale in Japan from Thursday, it was illustrated using Midjourney, an online image generator that can produce detailed pictures based on users’ prompts.
To create the panels, Tokyo-based Rootport entered a string of text descriptions, which he then refined using trial and error, to create images that matched his storyline.
Rootport gave his characters distinctive features that would help readers recognize characters as the story progresses. Credit: Shinchosha Publishing
Speaking to CNN via email, the anonymous author, who uses the pen name Rootport due to privacy concerns, said he completed the work in just six weeks. Spanning more than 100 pages and — unlike many manga publications — rendered in full color, a work of this scale would take over a year to complete by hand, he estimated.
Online AI imaging tools like Midjourney, DALL-E 2, Stable Diffusion and Google’s Imagen have exploded in popularity since they became publicly available last year. Yet they remain in their infancy, meaning that the author sometimes struggled to produce what he called “the perfect image for a specific scene.”
For one thing, Midjourney was not able to directly replicate existing characters in new poses or with different facial expressions. To get around this, Rootport gave his characters distinctive features (such as pink hair, dog ears or a red kimono) that would help readers recognize characters as the story progresses.
“(But) even in legendary manga works, it is commonplace for the character drawings to differ between the beginning and the end of the series,” he explained.
AI imaging tools also infamously struggle to accurately render human hands, which often appear with too many (or too few) fingers. For this reason, Rootport said he made a “significant compromise” by limiting scenes that pictured characters’ hands.
“Hands were difficult to draw, and details tended to appear as if they were melting,” he said.
Rootport, wearing gloves to protect his identity, demonstrates how he generates characters using text prompts. Credit: Richard A. Brooks/AFP/Getty Images
Redefining creativity
AI imaging tools are raising new questions about creativity and artistic integrity. In August, Colorado game designer Jason M. Allen sparked outrage when he won a $300 art competition with a futuristic image created using AI. Social media users questioned the artistic merit of Allen’s work, though he insisted that a huge amount of work had gone into his submission. “It’s not like you’re just smashing words together and winning competitions,” he told CNN at the time.
Similar controversies have touched the comic book world. Shortly after celebrated South Korean artist Kim Jung Gi died last October, a game developer published a tool allowing users to generate images resembling his comics using text prompts. The developer said he had intended it as an homage, but he soon faced furious backlash — and, as he told the online publication Rest of World, even death threats — from Kim’s fans.
But Rootport insisted that his comic book, which includes a 10-page how-to guide telling readers how to produce their own AI-generated manga, should be considered a work of art. He likened arguments in favor of AI art to those used to defend Marcel Duchamp’s “Fountain” — a porcelain sculpture of a urinal — or Andy Warhol’s “Campbell’s Soup Cans.”
“If you consider their works, which utilize existing industrial products and label designs, to be art, there is no rational reason to treat AI differently.”
The author reported that his work was positively received when he posted a preview of the comic online, although some people on social media have expressed their disapproval. One Twitter user described the project as “an absolute insult to manga and mangaka (manga artists) everywhere.” Another wrote: “Weird to publish a manga by AI when your country (has) some of the most talented artist in existence.”
But the author said he does not envisage AI putting artists out of work anytime soon. He compared his process to making music using MIDI instruments, in that the technology “excels at quickly turning mental images into reality.”
“Just as few composers would claim that MIDI makes human orchestras unnecessary, I do not believe that human manga artists will become unnecessary,” he added. “Both humans and AI create based on learned data from the past. However, humans can create not only from data, but also from emotion, experience and as a means of communication.
“Currently, AI still does not possess emotions or experiences, nor does it have a desire to communicate. In this respect, AI cannot yet create a perfect work on its own. Human assistance is essential.”
A new dawn
Beyond ethical issues, legislators and creators around the world are also grappling with copyright concerns raised by tools trained using large datasets of existing images. In January, stock photo giant Getty images announced that it was suing Stability AI, the company behind Stable Diffusion, for allegedly copying and processing its images without obtaining proper licenses. (In a statement to CNN, Stability AI said it “take(s) these matters seriously” and is “reviewing the documents and will respond accordingly.”)
Japanese manga artist and politician Ken Akamatsu has been among the most prominent voices calling for new guidelines on AI-generated art. Posting a video to his personal YouTube channel, Akamatsu, who serves in the Japanese parliament’s upper house, suggested that creators should be able to exclude their work from datasets used to train AI programs — or be compensated should they opt in.
Copies of “Cyberpunk: Peach John” at the office of publisher Shinchosha in Tokyo. Credit: Philip Fong/AFP/Getty Images
Rootport, however, believes that AI technology will ultimately liberate artists from the “grueling process” of creating manga, which he said often entails onerous deadlines that see artists suffering ill-health due to overworking. Tools like Midjourney could, he argued, improve the industry’s “inhumane working conditions.”
“It would not only make things easier for manga creators, but also has the potential to improve the quality of the stories themselves,” he said.
“By reducing the amount of time spent on labor-intensive tasks, creators can devote more time and energy to the creative aspects of manga, leading to more interesting and engaging stories.”
