AI is coming to the film industry—but what will it mean?

A 16-year-old movie that underperformed at the box office might not seem like the ideal choice for a project to repurpose, but Baz Luhrmann reworked material from his 2008 World War II romantic epic Australia into a six-part series called Faraway Downs for Hulu late last year. He saw it as a way to expand on themes like institutionalized racism toward First Nations peoples while using previously unused material and taking advantage of the longer-running, episodic medium.

As a filmmaking tool, AI is poised to do the kind of repurposing of footage that Luhrmann and Hulu did to an astonishing degree and with relative ease. Indeed, the technology may create completely new footage from scratch, without the need to bring big-name actors back to the set.

But fears about AI are rife in the media and entertainment industry. In 2023, it was a major component of stop-work actions by two of the most prominent artists’ industry unions. And as protesting actors and writers on one side and the studios on the other know all too well, AI is here to stay. The question is, how much of an impact will it make on the M&E industry?

Greig Fraser is the Academy Award–winning cinematographer behind Dune: Part One, Dune: Part Two, The Batman, and Rogue One: A Star Wars Story. “There are two sides of the fence,” Fraser says. “Back when the automobile was becoming a thing, carriage makers might have said, ‘Hey, we make carriages for a living, what’s this going to do to us?’. The other side of the fence was, ‘Hey, we make carriages for a living, and my God, we can now put an engine in one.’ Suddenly you can be a car maker.”

Fraser points out that there already is a market for repurposing old or existing films, it just hasn’t involved AI yet. Director Peter Jackson used groundbreaking image and sound enhancement techniques for the 2021 documentary series The Beatles: Get Back and his 2018 documentary They Shall Not Grow Old. For both projects, aged footage was digitally restored to the level of modern standards. In the latter project, the footage had to be retimed to appear more natural than the characteristic fast motion jerkiness of early 20th century hand-cranked film, and new high-definition sound was recorded to match up, written courtesy of lip-reading experts.

Couldn’t AI do all that, and more importantly, wouldn’t studios pay willingly for it if it could?

Two advances from academia show how AI tools might help make, remake, or remix films. RE:Color, a recent breakthrough from Graz University of Technology in Austria, used AI to colorize monochrome (black and white) footage. Automatic image colorization isn’t new, but previously it was a labor-intensive process and the results were dazzlingly bright but not very authentic.

The second advance is a collaboration between Cornell Tech and Google Research called DynIBaR, which takes existing footage and creates a clip from a new point of view, showing perspective from another point in the scene depicted. The main application of DynIBaR so far is image stabilization, but it’s not hard to imagine a subsequent algorithm that can remake a movie from another physical location inside a scene, maybe concentrating on the POV of a background character to tell a whole new story.

Tools like these are already having a real-world impact on the M&E industry. In November 2023, Warner Music Group announced it was producing an animated biopic of French vocalist Edith Piaf, with depictions of the late singer created by AI from film and audio recordings.

Building on the AI that’s already been developed, the possibilities for repurposing content are nearly endless.

Vlad Susanu, founder of Game Clubz, points out a parallel in the way gaming realizes ongoing revenue. “These tools make a strong case for squeezing more profits from studio archives, kind of like how downloadable content (DLC) and expansions keep extending games’ money-making lifespan,” he says. “Colorizing and remixing timeless films could attract streaming subscriptions or pay-per-view purchases, especially among younger crowds who might pass on old black-and-white versions.”

Imagine a film festival or streaming service showing the entire Marx Brothers oeuvre in realistic color, or popular franchises retold from a secondary character’s point of view—maybe Hagrid in the Harry Potter films or C-3PO in the Star Wars universe.

But even if this is technically doable, should it be done? A major concern is the preservation of artistic integrity. Art has been considered humanity’s sole domain for millennia, so is it still art if computers do the creating? And does allowing AI to rework a piece of human-created art negatively affect the artist and their original creation?

With streaming services still relatively new and a glaring hole in the 2024 release calendar following six months of strikes in 2023, artistic integrity might be of secondary concern to studios right now; they probably won’t give up the vast untapped markets promised by AI without a fight.

“For sure there’s a strong market for it,” Fraser says. “To be able to do it with old classic movies that maybe don’t hold up to IMAX or large screen viewing because the negatives have degraded so much. Corporate dollars are going to drive everything, that’s unfortunately the reality, and while I can’t see a Marx Brothers movie making a billion dollars or an old Elvis movie having much of a theatrical run, there might be some money to be made.”

As Jeremy Toeman, CEO of AI start-up Aug X Labs and former VP of product for WarnerMedia/Sling/CBS Interactive says, this is about companies that are sitting on vast stockpiles of assets—a century of them, to be precise.

“If there was something I learned at WarnerMedia, legacy content companies are sitting on massive libraries of visual content,” he says. “Most of it isn’t in use and isn’t being monetized. AI will empower rights-holders to bring their properties out of the dustbin and shed new light on them. Whether to train new models or reuse in a generative world, there’s an opportunity for anyone who owns a catalog of IP [intellectual property]. Media companies should see this as a huge opportunity to expand their reach.”

As sweeping an impact as AI has had across the M&E industry and in the news, there’s a lot more to it than loading up a movie file and clicking “make it color.” The computing power that enables these changes and versioning isn’t free. As Toeman says, “training new large language models from content archives is very costly.”

That said, Hollywood has deep pockets—especially if millions of dollars of box office revenue or the promise of skyrocketing streaming subscribers are on the table. Toeman adds that most studios and rights-holders should be able to leverage their content without incurring high costs.

In many ways, the risk might be more reputational and cultural. The last high-profile effort to colorize black and white films was by media mogul Ted Turner in the late 1980s, and the backlash by prominent critics and many filmmakers was swift. Though it was initially profitable, ratings for colorized films steadily declined, and when Turner launched the TCM network in 1994, black and white movies were shown the way they were originally made.

In some ways, repurposing old films into new forms of content seems like a no-brainer. The streaming revolution has given the industry an insatiable thirst for new content, and using AI to remix existing movies and shows might meet their demand to fuel long-promised revenue growth.

Kelly Indah, a software developer and security analyst who’s written about AI for Increditools, an online news source about the tech industry, agrees—with a caveat. “[Content demand] has increased to a rather voracious speed, and it’s quite worrying since the audiences don’t seem to know what’s enough, and quality material and originality are less and less appealing to the public.”

The Walt Disney Company CEO Bob Iger agrees with Indah about too much being a bad thing, explaining during an investor earnings call late last year that the company had “lost some focus.” Reprioritizing high-quality projects over a sheer volume of content, Iger said the company is moving forward with “creative excellence as our singular motivating priority with the content we create.”

Asking the question of whether to use AI to make “new” content out of existing content, the answer may be the same as to whether to make any art at all—as long as it’s good. If content is done right, it’s not just about selling a movie or a show, it’s about selling an experience, and that’s something people are willing to pay for.