How AI Is Quietly Changing the Economics of Hollywood

The actors are real, but the desert, the stone giants, and this fall, all generated with AI. It would take two months for one character, but with AI, this was done in three days. Is this an indicator? That's really the magic of hybrid filmmaking is the ability to move into worlds that would otherwise have cost millions to make. Let's move! Hollywood is already using AI tools like these. You may just not notice. You kneel before me. Pretty tight performance transfer, right? A lot of the traditional studios are reticent to talk about their use of AI publicly because there is so much hatred for AI. AI is still kind of stupid in that it doesn't understand the physical world. So how far can the technology go? And how much money is AI really saving Hollywood? I'm technically in LA and the year is 2026, but I've been transported to the New York Stock Exchange and it seems to be around circa 1950s. Let's move into the future. So I've been to three different eras in the span of about 10 seconds. But of course, I never left Luma Soundstage in LA. Luma has evolved from a 3D capture app into an AI video company reportedly valued at more than $4 billion, with backing from NVIDIA and Saudi state-owned Humane. Luma is using what they call hybrid filmmaking, combining real actors with AI-generated worlds. They say it makes production faster, more flexible, and cheaper. I can just ask for something super simple, only one sentence, but the agent has like full context of like... It's all the rules of this world and like what it's allowed to do, what it's not allowed to do. Design starts at a desk. That's where Brian Segundo creates characters for Dream Lab, Luma's in-house studio. His team is currently working on Salt Punk, a story about futuristic pirates stranded on an island. One of our writers, Fishkin, kind of laid out the lore of this world. When we take them to AI, we have to do everything in our power to make sure we maintain their likeness. So then once we capture their performance. Their performance translates into AI one-to-one, so it's basically them in real life. I told them, it's like, hey, stand by that wall. I'm going to take a few pictures of you. From there, we were able to build out their full costume. From there, we developed a model sheet. How long would that have taken you previously versus now? I mean, how many layers, how much time? Like, make it really real for us. It would take two months for one character. This was done in three days. Substantially less time. We can ask for like a hundred variations and we decided to make Dusty into a female character instead of a male character. In production, that would have been a nightmare. We're already like hundreds of millions of dollars deep. This is not going to work. But with AI, it's like, hang on, let's go back to the drawing board and then let's figure stuff out and then we'll bring it back out. We didn't end up in the middle of nowhere nine months ago. Hello, we're watching you perform over there. Hi there. Hi. So good to see you. Hi. How does it feel different from a regular scene you would shoot or even one on the green screen? Look man, it's like GTA. They're throwing me in an open sandbox, I get to do whatever I want, go wherever I want. It's amazing, it's free. Luma says it has three revenue streams, subscriptions to its creative tools, paid access to its AI models, and consultations with studios, brands, and agencies on how to use AI in their creative workflows. For studios, part of the appeal is cutting production costs. A single day of shooting a feature film or TV series can cost hundreds of thousands of dollars. Luma says adding sets and costumes digitally can reduce those costs. For filmmakers like Doug Lehman, one big draw is the promise of lower costs. Lehman shot his new movie on a bare stage, using AI to create much of the world around his actors. The film's producers say it cost $70 million, less than a quarter of what it would have cost to shoot traditionally. There's so many stories that are not getting told today at high quality for budget constraints. And now we're seeing so many more stories being green-lighted all across the world, and that's really, really exciting. But despite the billions flown into the AI video market, the tech still has some basic limitations. You can't make a full TV show inside of Luma today. If I'm saying, hey, I want to move this pen from here to here, AI doesn't really understand distance. So it's going to move from here towards here, but not exactly. And I think that's one of the key things that is often getting missed. Can AI be done with one person sitting alone making a whole full movie? Yes. Will it be at this level of quality and this level of storytelling? I have not yet seen that. Getting good results takes more than a few prompts. That's according to Holly Willis, who co-founded USC's AI for Media and Storytelling Initiative. Things like consistency of character or really nailing the performance that you're trying to achieve, all of these things are still very, very difficult. And it takes a lot of time, a lot of work, and a lot of patience and a lot of skill to actually create something. When do you draw the line between AI-enhanced content and AI-generated content, is there a difference for you? Yes and no. The industry is realizing that it needs to have much more nuance in its conversation about what AI is. And right now we have a very strong binary between AI yes or AI no. You think they'll believe us? How widespread is it in Hollywood? A lot of the traditional studios are reticent to talk about their use of AI publicly because there is so much hatred for AI. So I think there's a lot happening behind closed doors. Only a little bit of that, though, is made public. Yet there are signs of how quickly it's spreading. A year ago, Netflix had publicly confirmed using generative AI in at least one show. So far this year, it's up to roughly 300 titles. Google invested $75 million in A24 to develop AI tools with its filmmakers. And Lionsgate took a stake in AI video company Runway to make short-form AI series based on its films. There's kind of a race right now to figure out who will be the leader in this industry and which will be the central set of tools that everyone begins to use. The global AI video market is projected to grow from $5.5 billion to $42.3 billion by 2033. But few studios have publicly disclosed how much it's actually saving them. What parts of the pipeline do you think are actually dropping to near zero? What still seems like a fixed cost? Areas of the pipeline that are getting less expensive could be something like concept art or previs, things that you can do quickly and that aren't having a direct impact on the final pixel that you see in the movie theater. Other costs, though, I think are rising. So for studios, especially, thinking about what does it cost to train a model? What are the compute costs of doing all of this AI work? These are costs that are kind of hidden that people don't talk about. Do you believe that audience members are resistant to AI as a whole in cinema? Or is it that they don't want it to be so obvious? Among the generation of students that we have here, the pushback is 100% around environmental impact and IP theft. Those IP concerns are already playing out in court. In August, the estate of astronomer Carl Sagan sued Luma. Alleging the company used a clip of his voice in an ad for its AI video model without permission. And kept it running after being asked to stop. According to the suit, Luma said the use was too minimal to constitute copyright infringement. In an email to Fortune, Luma declined to comment, citing ongoing litigation. AI studios like Luma are also facing resistance within Hollywood. Why do you think that there is that hesitation? There's a lot of, like... Oh, you can't use AI? This is like ugly and bad and like how could you? You're betraying our entire guild for using AI. And I think all guilds are going through this as we're adopting AI. AI uses on the rise! AI uses on the rise! I understand it because it comes back to fear. And when change happens, and this is the biggest change any one of us, regardless of what we're going to go through, it's natural to be afraid. I mean, there have been some studies that show that up to 75% of viewers don't want to see a film or something that's been generated with AI. How much of that do you think is the consumers as well, the hesitancy? Consumer wants... Content that entertains them, that makes them feel. Technology has been part of filmmaking from the very, very early days. This is the same thing. I think the consumer is going to be excited by seeing more great stories being told. Especially if it's done well. Yeah, it needs to be done well. It has to be done well.