What Netflix bought when it bought InterPositive

On 5 March 2026 Netflix announced that it had bought InterPositive, a film technology company that Ben Affleck founded in 2022 and ran out of public view. Its 16 engineers, researchers and creatives moved to Netflix, and Affleck became a senior adviser to the streamer. Netflix rarely buys companies, and this deal came a week after it walked away from its bid for the studios and streaming business of Warner Bros. Discovery.

The price is now public. Netflix’s filing for the second quarter describes an acquisition completed in March, accounted for as a business combination, for a total of about 587 million dollars in cash. The filing does not name the company, and it describes no other acquisition this year.

Tools built from the footage

InterPositive does not make films from a text prompt. Its system builds a model from a production’s dailies, the raw footage of each shooting day, so that the model learns the look of that one project. Filmmakers then use it in post production for colour and mix, for relighting shots and for visual effects.

Affleck said he started the company after watching AI arrive in production and seeing, as a filmmaker, how the models came up short. What makes storytelling human, he said, is judgement, the kind that takes decades to build. Netflix will give its creative partners access to the tools and does not plan to sell them. Its content chief, Bela Bajaria, said new tools should expand creative freedom, “not constrain it”. Its product and technology chief, Elizabeth Stone, said innovation should serve storytellers, “not replace them”.

300 titles in half a year

The purchase sits inside a larger shift. In its letter to shareholders of 16 July, Netflix wrote that generative AI workflows had been used in roughly 300 of its titles in 2026, most of them in post production. It named three: Glory from India, Brasil 70: A Saga do Tri from Brazil and The American Experiment from the US, which used the tools for enhanced crowds, historical battles and establishing shots that build a world.

On the earnings call, co chief executive Ted Sarandos said The American Experiment contains 17 minutes of AI enhanced footage, made twice as fast and at half the cost of other methods, Cartoon Brew reported. In many cases, he said, productions would otherwise have left out key shots because they could not afford them. In March, Engadget’s report on the deal recalled a single visual effects shot made with generative AI in the series The Eternaut. Four months later the count is roughly 300 titles.

The same letter says Netflix now uses AI tools across its advertising business, from planning and creative production to campaign management and reporting.

What the deal says about production

Three things stand out. The AI arrives after the shoot, not in place of it: the camera, the actors, the set and the director’s choices come first, and the model learns from them. The model is trained on the production’s own material, which keeps the look consistent and the question of whose work it learned from simple. And the economics change at the edges: shots that used to be cut for budget now make it into the film.

There is also a competitive point. Because Netflix keeps the tools for itself and its partners, the advantage stays with the films it commissions. Other studios will have to build, buy or license their own, and the scarce asset is not the model but years of footage, workflows and the people who know how to use them.

The rules for people are being written at the same time. In June the members of Hollywood’s main actors’ union ratified a new contract with the studios, with 91.42 percent in favour, running until June 2030. It extends the protections for digital replicas, which require the performer’s consent and pay. A synthetic performer may take a role a union actor could play only if it brings significant additional value, and the producer must notify the union and bargain first, TheWrap reported. The studios rejected the union’s proposal of a payment equal to a member’s wage.

Brands and advertising

Advertising is moving the same way, with less patience. In November 2025 Coca-Cola released holiday films made by two AI studios, Silverside and Secret Level. In a behind the scenes video, the team said five specialists had produced and refined more than 70'000 video clips in 30 days. Some viewers called the result soulless, others called it better than the year before, and the company said the previous film had performed well with customers.

Volume is no longer the hard part. Choosing is. Netflix itself now applies AI to creative production for its advertisers, so the ads around a series may soon be made with tools much like those used on the series. The lesson from InterPositive applies to any brand: start from real material, lock the look, and let AI extend it, instead of asking a model to invent a brand from nothing. That is the idea behind Reel: styles locked from real references, every image and clip made to follow them, and a person who picks every one that is used.

What stays human

In film, the story, the casting, the performance and the decision about what the audience sees stay with people. In brand production it is the idea, the taste in the pick, and the responsibility for every face and voice that appears. The tools take on more of the labour every quarter. The judgement Affleck talks about does not move.

For a company that makes pictures and films, three steps follow. Build your AI work on your own footage and photography, so that the look is yours. Get consent and rights in writing for every person whose likeness or voice is used. And keep a named person on the final pick, because the audience will hold the brand responsible, not the model.