What a product film actually costs now

A product film used to be priced by the calendar. You were not paying for an idea. You were paying for a studio day, a crew, a director, a colourist, and the three weeks it took to get all of them into the same room. That is the part that changed.

What the old number was actually buying

Break a traditional quote apart and very little of it is creative. There is a location or a studio hire, billed by the day. There is a crew, billed by the day, whether or not the light is right. There is talent, equipment, insurance, catering, a post house, a colourist, a sound mix. Then there is the calendar that holds all of it together, because every one of those people has to be booked, and booking is what makes a six-week timeline a six-week timeline.

None of that was waste. It was the cost of physically capturing a thing that did not exist until you captured it. But it did mean the work went to companies that could absorb a six-figure line item, and everyone else ran a template.

The first credible numbers came from a company that published them

Most claims about AI production costs come from people selling AI production. Klarna is the useful exception, because it put the figures in its own press release in May 2024, where its investors could read them.

What Klarna reported:

• A $6 million reduction in image production costs, $1.5 million of it in the first quarter of 2024 alone.
• The image development cycle cut from six weeks to seven days.
• More than 1,000 images produced in three months.
• Spending on external marketing suppliers down 25%.
• Around $10 million in annualised savings attributed to AI, which the company put at 37% of its total cost savings.

The number worth sitting with is not the $6 million. It is six weeks to seven days. Klarna did not simply buy the same output for less. It changed how often it could make something at all. A six-week cycle means roughly eight campaigns a year. A seven-day cycle means you can answer a competitor, a season, or a bad week while it is still happening.

Then the largest advertiser in the world did it at scale

In November 2025 Coca-Cola remade Holidays Are Coming, its most recognisable campaign, with generative AI. Reporting on the production put it at roughly 70,000 generated videos with about 100 people involved — down from the prior year. Pratik Thakar, Coca-Cola's global VP and head of generative AI, described the work as guided by "human storytellers".

It is worth being straight about how it landed: the reception was mixed, and in places openly hostile. Viewers picked at visual glitches and at what the method implied for the people who used to make these films. If you are deciding whether to work this way, that reaction is data too.

But notice what the argument was about. Nobody disputed that it could be done, or that it was cheaper. The complaint was that parts of it did not look good enough. That is a different conversation from the one the industry was having three years ago, and it is the conversation that matters now.

What did not get cheaper

Generation is close to free. Judgment is not.

A model will return forty versions of a shot in the time a crew takes to reset a light. Thirty-nine of them are subtly wrong — the product reads as plastic, the hand has the wrong weight, the room says "nowhere", the colour drifts a quarter-stop between cuts and the film stops feeling like one film. Knowing which one is the keeper, and being able to say why the rest fail, is the entire remaining job. It is also the part that was always the expensive part; it was just buried inside a day rate.

This is why "we used AI" is not a service and never was. The Spring 2026 CMO Survey — run by Duke University's Fuqua School of Business with Deloitte and the American Marketing Association — found that no marketing technology activity its 308 respondents rated scored above 5 on a 7-point performance scale, and that performance has not improved in two years. Adoption is running ahead of the ability to use it well. The tools got distributed faster than the taste did.

What this means if you are the one spending the money

Three things follow, and none of them require you to believe anything about the future.

The quote you were given last year is no longer the market price. If a film is quoted at six figures and six weeks, the fair question is which part of that is studio, crew and calendar, and whether your film needs them. Some do. A film that depends on a real face, a real place or a real product in real light still has a reason to be shot.

Volume stopped being the constraint. When a cycle runs in days, the sensible number of films is not one hero piece a year. It is a hero piece and twenty variants that let you say something specific to each audience, each region, each week.

Your risk moved. It used to be budget risk: spend a lot, get one shot at it. It is now taste risk: spend little, publish fast, and look cheap in public if nobody is exercising judgment between the model and the feed. Coca-Cola had the budget and still took that hit.

Where this ends up

The constraint on brand content was never ideas. It was the cost of producing them, and that cost has fallen far enough that producing content quickly is now the ordinary way to work rather than the expensive way. Brands that keep a six-week cycle are not buying higher quality with the extra five weeks. They are buying a slower answer to a market that has stopped waiting.

The studios worth paying are the ones that treat the cheap part as cheap and charge for the part that is still hard: deciding which frame is the one, and being right.