AI Could Cost Filmmakers More Than 35mm, Korean Pioneers Warn

A 100-minute AI-generated film can already run up a token bill above KRW500 million ($366,000) using the most expensive commercial tools, Kim Kwang-jib told Taiwan Creative Content Fest (TCCF) in Taipei. Kim, founder and CEO of AI content studio Studio Meta-K, warned that if AI providers raise prices tenfold, that figure becomes KRW5 billion ($3.66 million) — more than a typical Korean or Taiwanese feature costs today.

"If they raise token costs roughly tenfold, it could go above what the films we're shooting now cost," Kim said. He added that AI companies are all losing money and are keeping prices low as a marketing play.

What does this change for Asian producers?

Hollywood can absorb a tenfold jump on KRW200 billion ($146.6 million) budgets and still make very cheap AI films, Kim argued. Asian features, which typically spend KRW3 billion–KRW5 billion ($2.2 million–$3.66 million) on live action and VFX, would have to ask whether AI is efficient at all.

The mechanism, Kim explained, mirrors celluloid economics. Preproduction was meticulous when every take burned film stock; digital then let crews shoot and cut freely. With AI, every re-render costs money again. Ten client changes delivered at once mean one paid re-render; the same ten arriving one per day mean ten re-renders and ten payments.

"Unless the preproduction stage is made solid, I think AI could end up spending even more money than in the 35mm film era," Kim said.

A feature in cinemas, made with AI

Lee Sang-wook, head of the AI Content Lab at MBC C&I, a subsidiary of Korean broadcaster MBC, offered a working case. His lab's 139-minute feature "Raphael" took around a year and four months to make, and a single shot sometimes tied up about 10 artists for a full day. The project had backing from Korea's Ministry of Culture, Sports and Tourism and the Korea Creative Content Agency, and opens in Korean cinemas on Nov. 25.

Budgeting stays difficult because tools move fast. "If you apply technology from a year ago, the scenes look very old," Lee said. The market was initially hostile: "When we made AI short films in 2024, the market reaction was very cold, in Korea as well," he said. The lab began with generative AI shorts in 2024, then packaged AI content as "The Prompt: Next Drama" on Korean streamer Wavve. A cinema release can move to streaming later, Lee noted; the reverse order is hard.

Commissions vs. IP: where the money works

AI already pays off in advertising, Kim said. A Pizza Hut TV spot that aired last year took about eight months and roughly KRW2 billion ($1.4 million) to produce; Studio Meta-K's in-house version took about two weeks and around KRW30 million ($22,000) — about 2% of the cost. Advertising production costs can fall by more than 90%, he said.

Drama is harder. Kim doesn't think audiences will yet pay to watch fully AI films or series, so he favors a hybrid: a scene needing 500-plus extras could be generated with AI, leaving only close-ups of the leads to shoot live. Studio Meta-K is producing war scenes with AI for the KBS program "History Special."

The company's larger experiment reverses the pipeline: build an entire 100-minute film at film-like quality in preproduction, then reshoot only what feels off. For now, Kim said, perhaps 50 of those 100 minutes would still need live-action reshooting — but he expects that eventually nothing will.

Founded in 2022 with established drama directors and writers, Studio Meta-K has completed more than 100 commercial projects. This year it was selected for two government R&D projects worth about KRW15 billion ($11 million) combined to build SaaS-based platforms, leading a consortium including an MBC subsidiary, KBS and MBN. It raised around KRW3 billion ($2.2 million) in a Series A last year and around KRW10 billion ($7.3 million) this month, with an IPO planned next year. It is also converting Im Kwon-taek's 1993 classic "Seopyeonje" into an animated feature with AI, in collaboration with Dongseo University and with the director's involvement, for theatrical release next summer.

Open source, standards and jobs

The two companies split on open-source tools. Kim said free models have improved sharply, and Studio Meta-K builds easy shots with them while reserving paid tools for difficult ones. Lee said MBC C&I avoids open-source because it cannot verify training-data transparency, noting no AI model yet meets EU AI Act guidelines; the company has joined C2PA (a content-provenance standard). Lee also said streamer delivery is tougher than cinema DCP specs, since platforms analyze files mechanically, and AI output still struggles with bitrate and color grading.

On jobs, Kim said crews of hundreds on one scene will grow rarer, and roles like drone operators and set work could largely disappear. "I don't see AI replacing that workforce," Kim said. "I personally think workers who have learned AI will replace workers who haven't."

Lee compared the shift to sound's arrival, which cost theater musicians their jobs, and predicted new roles: previsualization artists (who break scripts into previs) and continuity supervisors. One Korean studio already has more than 50 AI artists working together, he said. Actors, he added, have warmed to AI after previs of a dangerous accident scene turned skepticism into approval.

Asked how directors can keep AI from flattening their vision — the session was moderated by Archin Yen, director and VFX supervisor at WWWIND Studio — Kim noted most AI videos default to eye-level, front-on bust shots that film and TV rarely use. "Every VFX company in the world uses the same tools, but I think it differs depending on who uses them," he said. "In the end, I think that human 5% decides the quality."