A Pizza Hut Ad That Took Eight Months and $1.4 Million to Shoot Was Remade With AI in Two Weeks for $22,000

A Pizza Hut Ad That Took Eight Months and $1.4 Million to Shoot Was Remade With AI in Two Weeks for $22,000

A $1.4M Pizza Hut ad was remade with AI in two weeks for $22,000. Learn how it was made, the hidden token costs, and what it means for marketing budgets.

In this article

Let's Discuss your tech Solution

book a consultation now
November 17, 2026
Author Image
Suman Dilip Kumar
Marketing Analyst
Suman Dilip Kumar is a Marketing Analyst at Centric, experienced in managing client relationships and ensuring seamless project delivery. She is known for her strong organizational and communication skills, translates client goals into actionable plans, and ensures every project runs smoothly from start to finish.

On November 3, at an advanced-media conference in the South Korean city of Incheon, a studio founder named Kim Kwang-jib put two versions of the same Pizza Hut commercial in front of the room. The first had aired on Korean television the year before. It took about eight months to make and cost roughly 2 billion won, which is a little over $1.4 million. The second was made by his company, Studio Meta-K, with generative AI. It took two weeks and cost about 30 million won, or around $22,000. That is roughly 1.5% of the original budget.

Four days later Kim repeated the comparison on stage at the Taiwan Creative Content Fest in Taipei, where Variety was in the audience. He also added a broader claim. In advertising, he said, production costs can now fall by more than 90%.

Figures like these tend to travel as headlines and then disappear. What makes Kim's case worth a closer look is everything he said around it. Across the two sessions he explained how the work was actually made, why AI can still cost more than film in the wrong hands, what the savings should really be spent on, and which jobs on a production are already changing. It is one of the most detailed public accounts so far of what AI-made advertising really costs and who it changes.

Eight months and 2 billion won became two weeks and 30 million won

Studio Meta-K is a small company with a serious pedigree. Kim founded it in 2022 alongside well-known Korean drama directors and writers, and he spent nine years as a visual effects producer in Hollywood before that. According to Variety, the studio has completed more than 100 commercial projects. It raised around 10 billion won this month, is preparing for a public listing next year, and leads a government-backed research consortium that includes the broadcasters KBS and MBN and a subsidiary of MBC.

The Pizza Hut comparison was first reported by Bloomberg, and Kim has now used it at two industry events in the same week. It is worth being precise about what it shows. The original spot was a conventional production that aired on television. The AI version is one Studio Meta-K made in-house, and the reports do not say whether it ever ran on air. What it demonstrates is how close an AI-built version can get to a finished, broadcast-standard brand commercial, and what that version costs to make.

Kim was also careful about where the savings hold. His claim of cost reductions above 90% applies to advertising. When he turned to feature films and drama, his tone changed completely, and that contrast is the most useful thing in his talk. A 30-second commercial has a fixed brand world, a short running time and a clear approval path. Those are exactly the conditions in which AI production is already mature enough to replace most of a traditional shoot.

AI builds the 500-extra crowd scenes and cameras shoot the close-ups

The way Studio Meta-K works is more interesting than the phrase "made with AI" suggests. Kim generates the shots that are expensive to film and films the shots that carry emotion. A crowd scene that would normally need more than 500 extras, each with matching costume, hair and makeup, is built with AI. The lead actors' close-ups are then filmed for real on an LED virtual production stage or against a green screen. The studio has used this method for battle scenes in the KBS documentary series "History Special" and for match footage in an MBC sports program. It even generates aerial footage of Japan and Taiwan, where drone permits are hard to obtain.

Kim's rule of thumb is simple. "If you use AI, it has to be for the expensive shots," he told the Incheon audience. A person drinking coffee in a café, in his view, should just be filmed. The same logic runs through the studio's tool choices. On one animated project it builds the easy shots with free open-source models and saves the paid commercial tools for the difficult ones.

The bigger change is the order of work. On a traditional production, the shoot and the edit consume most of the budget, and any AI comes in afterward as a fix. Kim wants to reverse that. His studio is testing a pipeline that builds the entire production at near-final quality during preproduction, then sends cameras out only for the parts that still feel wrong, such as emotional close-ups. For a 100-minute film, he estimates about half would still need live shooting today. He expects that share to fall to zero over time. For a 30-second commercial, the share of live shooting can already be very small, which is why the Pizza Hut numbers look the way they do.

Talk to Our Experts Now!

Ten changes sent together cost one render, and ten sent one a day cost ten

The most practical part of Kim's talk in Taipei was about a cost that is easy to overlook. Generative AI services charge by the token, the unit of computing a model uses, and every time a shot is regenerated the meter runs again. Kim compared it to the history of film. When every take burned through 35mm stock, preproduction was meticulous because mistakes were expensive. Digital cameras made shooting and cutting almost free, and productions relaxed. AI brings the old discipline back, because every re-render is paid for again.

His example was simple. If a client sends ten changes in a single round, the studio can make them all and re-render once. If the same ten changes arrive one a day over ten days, the studio has to re-render ten times and pay ten times over. "Unless the preproduction stage is made solid, I think AI could end up spending even more money than in the 35mm film era," he said. For a feature film made with the most expensive commercial tools, he put the token bill alone at more than 500 million won, about $366,000.

He also warned that today's prices may be temporary. AI companies are losing money, Kim said, and are keeping prices low as a marketing play. If token costs rose roughly tenfold, a 500 million won bill could become 5 billion won. Lee Sang-wook, who runs the AI Content Lab at MBC C&I and shared the Taipei stage, added a second kind of cost. His lab's 139-minute AI feature "Raphael" took about 16 months to make, and a single shot sometimes kept around ten artists busy for a full day. New tools arrive weekly, he said, and "if you apply technology from a year ago, the scenes look very old."

The lesson sits in an unexpected place. The biggest variable in an AI production budget may be the approval process. A campaign that passes through a brand team, a regional office, legal review and a CEO's comments in separate rounds will pay for each of those rounds in compute. The organizations that capture the full saving will be the ones that agree on what they want before anything is generated.

Studio Meta-K wants a 10 billion won budget to look like 200 billion

The sharpest idea in Kim's Incheon talk was about ambition. A Korean film with a budget of 10 billion won, he pointed out, competes for the same audience as American films that spend 200 billion. "The time has come to spend the full KRW 10 billion with AI and make it look like a KRW 200 billion film," he said. His frustration is that many companies are still using AI to shave 200 million or 300 million won off a budget and keep the quality exactly where it was.

He has watched what happens when a market treats AI purely as a discount. When Studio Meta-K started making AI video, he said, it was the most expensive option in Korea. Since then so many companies have cut their prices so far that he now describes the Korean market itself as "very difficult." Once AI production became a race to the cheapest bid, the savings went to buyers and the quality stayed flat. Few people in that market are using the technology to produce work that would have been impossible before.

Kim's own answer is to reinvest. The studio says it uses AI to cut production costs so that it can build and own intellectual property. Its first big bet is an animated feature based on Im Kwon-taek's 1993 classic "Seopyeonje," made with the director's involvement and due in cinemas next summer. Kim puts the cost of an animated feature at 5 billion to 10 billion won in Korea and believes his studio can make one for around 500 million. Koo Yong-nam of the short-drama producer Hello Works, who shared the Incheon panel, described the same strategy. Both companies said they are putting their AI savings into owning source material and co-production stakes.

The efficiency gain from AI production is now large enough to be a strategic decision in its own right. A brand can return it to the finance team as a smaller production line. It can also spend the same budget on far more ambitious work, many more market-specific versions, or creative assets it owns and reuses for years. The Korean producers who are most experienced with these tools have chosen the second path, and their reasoning applies to marketing budgets just as well as to film.

One Korean studio already has more than 50 AI artists working together

Both speakers in Taipei were direct about jobs. Kim said crews of hundreds on a single scene will become rare, and roles such as drone operators and set builders could largely disappear. Some production companies, he said, are already keeping their preproduction staff and letting production crews go so they can co-produce with studios like his. His conclusion was blunt. "I don't see AI replacing that workforce," he said. "I personally think workers who have learned AI will replace workers who haven't."

Lee described the jobs that are appearing. He expects demand for previsualization artists who break a script down into an AI-built draft of the film, and for continuity supervisors who keep characters and locations consistent from shot to shot. He said one studio in Korea already has more than 50 AI artists working together, and that technical directors will matter more as the tools improve. He also noted a change of heart among actors. On one daily drama, the cast was skeptical until the previsualization showed that AI would handle a dangerous accident scene that could have injured them.

Kim's most useful observation was about skills. He said the AI tools can be learned in a few months, while film grammar takes years. Rules about camera angles, lenses and editing are what separate professional work from amateur output. Most AI video he sees defaults to the same eye-level, front-facing shot of a person from the chest up, a framing that film and television rarely use. Studio Meta-K hires accordingly. Most of its staff studied film first and learned AI afterward. "Every VFX company in the world uses the same tools," Kim said. "In the end, I think that human 5% decides the quality."

Advertising is where AI production pays off first

The gap between commercials and feature films in these talks is striking. Korean AI features have struggled to find paying audiences so far. Two of them, "I'm PoPo" and "Man in Hanbok," drew 781 and 640 cinema viewers and left screens within two weeks, according to Korea JoongAng Daily. Kim himself does not think audiences will pay to watch fully AI-made films or series yet. MBC C&I's "Raphael" opens in Korean cinemas on November 25 and will be the next public test.

Advertising has none of those barriers. A commercial is short, its visual world is already defined by the brand, and its audience is used to stylized images. That is why the same studio that warns about runaway costs on a feature film can show a broadcast commercial remade for 1.5% of its original budget. The hardest technical work has largely been done. The real decisions now concern how a marketing organization briefs, approves and staffs its creative work around these tools.

Those decisions are where the advantage will come from. The Pizza Hut comparison shows that the cost of making a polished commercial can fall by an order of magnitude. Kim's talks show what separates the teams who benefit from that and the teams who only cut a line item. The winners lock their creative decisions before the first render, put the savings into more ambitious work, and keep people with real craft in charge of the final result. In a market where everyone has access to the same tools, as Kim put it, the human 5% is what decides the quality. 

Sources

Contact_Us_Op_01
Contact us
-

Spanning 8 cities worldwide and with partners in 100 more, we're your local yet global agency.

Fancy a coffee, virtual or physical? It's on us – let's connect!

Contact us
-
smoke effect
smoke effect
smoke effect
smoke effect
smoke effect

Spanning 8 cities worldwide and with partners in 100 more, we're your local yet global agency.

Fancy a coffee, virtual or physical? It's on us – let's connect!