Streamlined Guides: K-Content – Korea Rents The Pipeline

PART FOUR: Who owns the production pipeline when the pipeline is software; what happens to the middle of an industry when the tools no longer need one, and whether Korean speed is still an advantage when the tools can simulate it.

Editor's Note: This is the fourth installment in a series that looks at the Korean content industry – how it evolved, where it stands now and which business models it could be adopting in the future. The series is written by Jaden Jaebum Park, a director and producer at Korean broadcaster MBC, who between productions writes about the economics of content industries. See the links at the end of this report for the first three installments.

Korea Rents The Pipeline

By Jaden Jaebum Park

This summer an SBS drama, Agent Kim Reactivated (2026), ran a caption at the top of its first two episodes telling viewers that the scene to follow had been made with AI. For about three minutes the series showed its lead, played by So Ji-sub, on an infiltration mission into North Korea years before the story begins. Not one frame of it had been shot.
 
Morpheus Studio, which made the sequence, generated more than 2,000 shots and kept 289. So Ji-sub's face, the studio's AI director told the Korea Times, was produced from prompts, “without a single reference photo”. The sequence was built to stand on its own, with no cutting back and forth between generated and filmed images, so that the join would never show.
 
Three months earlier, Netflix had bought a tool that runs the other way. It starts from what a production has already shot and finishes the shots the production could not get: the missing insert, the background that has to be replaced, the lighting that came out wrong, the pickup that would otherwise mean calling the actors back. The company that built it, InterPositive, had 16 employees. Netflix booked a March acquisition at $587m, widely identified as InterPositive, and said it had no plans to sell the tool commercially. One tool begins with a prompt and ends with a face conjured by name. The other begins with the footage and stays inside the space, the lens and the camera moves the director already chose.
 
Part Four in our K-content series is about Korea standing between those two tools. Korean crews have taken up the new tools as fast as anyone in the business, rent them and have not yet taught one with their own footage.
 
Previous editions of this series left three questions at this door. Who owns the pipeline when the pipeline is software? What happens to the middle of an industry, the layer that is already thinning, when the tools that used to require a middle no longer do? And is there a version of that change in which Korean velocity is an advantage rather than a liability? This edition takes them in that order and stops at what the change means for Korea. What to do about it is the final edition of the series.

Sixteen Employees

The dates matter, so here they are in order:

*In December 2025, Netflix agreed to buy Warner Bros Discovery for about $82.7bn, for the franchises and the HBO library.

*On February 26, Warner's board judged Paramount's $31-a-share offer the better one, Netflix walked away, calling the deal "no longer financially attractive", and collected a $2.8bn termination fee that Paramount Skydance paid.

*On March 5, Netflix announced InterPositive. The same week it signed a multi-year first-look deal with Artists Equity, the company Ben Affleck runs with Matt Damon.

None of this is cause and effect; an acquisition does not come together in a week. Set side by side, the two months still read one way: a company that had failed to buy scale and a library bought capability instead.

What the tool is, in Netflix's own words: Affleck's team had been shooting its own datasets on controlled soundstages since 2022: smaller sets of data, built around filmmaking techniques and kept away from performances. The company's own post said the tool was there “to expand what's possible on set, in post, and across the whole production process”, which is a wider claim than post-production and worth keeping in mind. Variety reported that Netflix's creative partners would get access and that the tool would not be sold. Asked whether Korean and Indian VFX houses count as creative partners, Netflix did not answer.

The patent says what the tool was built to do and for whom. US patent 12,438,995, granted in October 2025 to the company Affleck founded, puts the saving on principal photography at “conservatively from 10-20% of total production costs” and lists what those savings replace: background actors and stand-ins, second and splinter units, inserts, reshoots, distant locations. The business model is in the same document. The vendor would charge “a fraction of the cost savings as their fee”. A tool designed to be licensed to studios in exchange for a share of what it saved them was bought by one studio and taken off the market. The numbers are the applicant's own, with no method shown. The structure is the point.

And Korea, in the same year? CJ ENM did not buy. It partnered, joining Google and some 18 companies and universities in an AI content alliance announced in February. Studio Dragon adopted tools in-house and says, without naming titles or tools, that work which once cost billions of won is now done for millions. A survey by the Korea Creative Content Agency (KOCCA) in October gives the proportions: 32.1% of content companies used generative AI at the end of 2025, up from 7.8% two years earlier, and 2.9% own a generative system of their own.

In 2020 to 2022, Korean studios bought LED volumes, and they still own them; in 2025 and 2026 they rent the generative layer. One side bought. The other rents. At Busan this month, on a panel hosted by Netflix and the Motion Picture Association, The Hollywood Reporter's note on the room was that Hollywood's creative community remains deeply wary of the technology and Asia's content industry mostly sees an opportunity. Hollywood used the tools first; Korea adopted them aggressively.

Innovation for Filmmaking, By Filmmakers: Why InterPositive Is Joining Netflix - About Netflix

Netflix announced the acquisition of InterPositive, the filmmaking technology company founded by Ben Affleck that develops AI-powered tools built by and for filmmakers.

About Netflix

Shoot Close To Home

Eyeline, Netflix's VFX and research company, described its working method at a conference in Turin this month. Mohsen Mousavi, who runs what Netflix calls "visual pioneering", calls the method hybrid production.

The actor performs on a partial set under flat lighting. The surroundings and the light are rebuilt afterwards. The performance stays as it was shot. “It's an extension of the traditional art department,” he said, and drew the line himself between a demonstration and a finished show.

Put the patent's list beside that and you have what a production no longer has to film: distant locations (the patent's phrase is “shoot close to home”), crowds, second units, inserts, reshoots, the surroundings and the light. What stays in front of the camera is the performance, the location sound, the space, the lens and the camera move the director chose. That last criterion is mine, from the set; Netflix has not put it in those words. Hold on to the first list. It comes back later with people's names on it.
 
The Korean set sells the opposite. Part One of this series described the arithmetic: 130 to 140 minutes of finished television every seven days; shot in four to six, several locations in a day; a night that runs into the morning. Where the tool came in was the budget. Kang Yun-sung, on the Busan panel, said budget limits “have held the Korean industry back from growing to its full potential”, and that AI put back on the table what had been buried for money.