Raphael brings polished AI visuals, but a familiar plot
“Raphael,” Korea’s first feature-length film made entirely with generative AI, opens Nov. 25 after 14 months. Its polished 138-minute sci-fi drama tests technology, not narrative surprise.

Why it matters
- Korea’s first feature-length film made entirely with generative AI, “Raphael,” opens Nov. 25 after a 14-month, 138-minute production.
- Generative AI created character appearances, landscapes, voice acting, music and sound effects; the crew shot no scenes with real actors.
- Three co-directors—Yang Ik-june, Moon Shin-woo and Jung Joo-won—made the film for Matteo AI Studio and MBC C&I AI Contents Lab.
- The film was invited to an official Kling AI conference at the Cannes Film Market in May.
- The story is set in Delmore in the 2050s, where unregistered citizens face immediate execution.
By Elena Vasquez
A combat android begins to experience human feelings in “Raphael,” Korea’s first feature-length film made entirely with generative AI after 14 months of careful, cut-by-cut work. It opens in local theaters on Nov. 25. The crew shot no scenes with live actors.
Yang Ik-june, Moon Shin-woo and Jung Joo-won co-directed the 138-minute sci-fi drama. Matteo AI Studio and MBC C&I AI Contents Lab produced the film together.
According to the review, generative AI supplied every finished layer after initial planning and screenwriting. It created character appearances, landscapes, voice acting, music and sound effects. That makes the production more than an AI-assisted shoot. The film presents a complete synthetic workflow for a feature intended for theaters.
The experiment carries economic stakes. The review sets the release against rising production costs and shrinking investment in the Korean film industry. It presents the film as evidence of generative AI’s potential to support large-scale science-fiction productions.
What makes “Raphael” a full AI production?
The production lasted 14 months, according to the review, and covered five categories of finished output: characters, landscapes, voices, music and sound. The crew shot no scene with a real actor. The review says the team refined the movie cut by cut over those 14 months. Its judgment is blunt: the refinement paid off visually.
That distinction matters because earlier AI videos often broke immersion through obvious errors. The review says “Raphael” avoids the awkward visual mistakes and sudden changes in character appearance common in those videos. It also gives dialogue delivery and lip movements that look smoother and more natural than in earlier AI productions.
For a 138-minute film, consistency across faces, voices and mouths becomes a production test of its own. A striking image can carry a brief demonstration. A feature must hold its synthetic world steady through repeated scenes, transitions and performances. The review credits “Raphael” with meeting that visual test.
The review offers no production budget, staffing total or savings figure. Its case therefore rests on demonstrated capability, not a measured cost comparison. The film-industry claim remains about potential rather than a proven saving.
Who is Alexei, and what does he become?
Alexei begins as a ruthless combat android modeled on the young appearance of a dictator, according to the review. His assigned identity sits uneasily beside his emerging emotion. The story places him in Delmore, a strictly controlled virtual city in the 2050s.
By then, climate disasters and devastating wars have collapsed human civilization, the review says. Delmore enforces membership through immediate execution for unregistered citizens. Sophia, an unregistered girl, and Michael, humanity’s last priest, draw Alexei into the film’s central moral conflict.
The plot follows Alexei as he encounters Sophia and Michael and begins to experience human feelings. The review says the film turns that change into a question about whether an android can possess a soul and enter heaven. Michael’s presence gives the question a religious frame, while the combat setting gives it physical danger.
The film does more than ask whether the machine looks human. Its premise tests whether an emotionally changed machine can claim an interior life that Delmore’s rules deny it. The review praises the philosophical ambition of that question. Its complaint is that the story reaches it through familiar science-fiction machinery.
How convincing are the AI visuals?
An official Kling AI conference at the Cannes Film Market invited “Raphael” in May, according to the review. The invitation placed the film inside the AI industry before its domestic release and the Nov. 25 opening in local theaters.
Visually, the review says 14 months of cut-by-cut work produced polished images. The film remains visibly computer-generated, yet it avoids the jarring errors and appearance changes that the review identifies as common flaws in AI video. Its human figures stay consistent long enough for the narrative to guide attention toward Alexei’s changing behavior.
The generated audio forms part of the demonstration. Generative AI created the voice acting, music and sound effects, according to the review. Along with smoother dialogue and lip movements, those elements reduce the distance between synthetic characters and a conventional cast, at least within the review’s assessment.
The film’s strongest achievement, in the review’s view, is not photorealism. It is the ability to sustain a feature-length world with generated performers and environments. That achievement matters because it moves generative AI from isolated demonstration toward a production method that studios could theoretically use.
Why does the familiar plot undermine the experiment?
The praise does not extend to the script. According to the review, the story of a combat android finding humanity and freedom through human connection feels overly familiar because countless science-fiction classics have used that pattern. The broad theme carries more weight than the specific journey.
Across the 138-minute runtime, the dark future and familiar battle scenes can make the story predictable, the review says. Its dialogue and lip movements are smoother than those in earlier AI productions, but smoother delivery cannot make a familiar turn of the plot feel new. The source identifies predictable plotting as the film’s central weakness.
The review also asks why early AI filmmakers choose such heavy, big themes to display technology. Grand settings can showcase scale, and the generated ruined landscapes of Delmore give that scale an obvious subject. Yet the same approach can expose a script’s borrowed architecture.
“Raphael” has the look of a technical breakthrough while following the emotional route of established science fiction. The technology earns attention through what appears on screen. The narrative has to earn the audience’s investment through choices that a combat-android conversion story has not already made familiar.
This contrast explains the divided verdict. The review calls the visuals polished and the philosophical questions ambitious, but it finds the plot short of fresh ideas. Visual novelty and narrative novelty operate on different levels. “Raphael” succeeds more clearly on the first.
What could AI mean for Korean production?
The review identifies two pressures on the Korean film industry: rising production costs and shrinking investment. Against that backdrop, it presents “Raphael” as a case study in AI’s potential as a cost-effective tool for realizing large-scale science-fiction productions. The project suggests generative tools can support visual production without filming scenes with live performers.
But the review does not quantify that advantage. It reports a 14-month production, not a 14-month saving, and it gives no budget, labor-cost comparison or revenue forecast. “Potential” is the review’s chosen term. The film demonstrates scope, while the business case remains unproven in the material provided.
That caveat does not erase the significance of the release. A feature made without filming scenes with real actors puts a different burden on writing, visual continuity, voice performance and sound. “Raphael” completed that workflow and received an invitation to an official Kling AI event at Cannes. The remaining test starts with audiences and the Nov. 25 opening.
The review’s final point is direct: technology can expand visual possibilities, but compelling storytelling and original ideas remain the forces behind memorable cinema. The Nov. 25 release will show whether polished generative images can carry a familiar premise through a full theatrical run. It will also show whether AI filmmaking can move beyond spectacle without giving up narrative surprise.
Source: feed.koreatimes.co.kr
More from Elena Vasquez
Show full bio
Market editor covering technology and product news in artificial intelligence and machine learning at AI In Context.
708 articles
Related articles
- Epic Delays Its Disney Game to Go "AI First", Bloomberg Reports
- Seoul opens free AI art festival with deepfake Sean Connery film
- AI Can Produce Stories. Humans Still Give Them Meaning
- Google DeepMind Partners with Korea to Bring Frontier AI to Science
- Epic Games goes "AI first" on Disney project, Bloomberg reports