If there’s one thing Hollywood loves more than a reboot, it’s a revolution. So when generative AI came along promising to rewrite the rules of moviemaking, the industry reacted like a kid in a candy store, stuffing its face and worrying about nutrition later.

Now, with Lionsgate struggling to make an AI film that looks like more than an expensive storyboard, and OpenAI announcing plans to create a full-length AI-generated movie, the hangover has officially begun. It turns out that making movies with machines is a lot harder than pitching the idea of making movies with machines.

OpenAI Takes the Director’s Chair

OpenAI recently made waves by revealing it is developing a feature-length animated movie using generative tools, a first-of-its-kind experiment that would rely almost entirely on AI to generate imagery, characters, and scenes. The company says it’s partnering with production teams in Los Angeles and London, using the technology behind Sora, its text-to-video model, to bring the story to life.

In theory, that means no cameras, no sets, and maybe not even actors, just prompts, algorithms, and servers doing the heavy lifting. It’s the kind of idea that sounds like it came from a Silicon Valley pitch deck titled “Disrupting the Human Experience.”

But here’s the thing: if this works, it could truly change how visual content gets made (depending on how much and which elements of it are actually created by AI vs humans). And if it doesn’t, which seems likely, given the state of the technology, it’ll become the most expensive proof-of-concept ever rendered in 4K.

Meanwhile, at Lionsgate…

While OpenAI is chasing Oscar glory with code, Lionsgate is already licking its AI-inflicted wounds. Last year, the studio behind John Wick and The Hunger Games struck a partnership with Runway, one of the leading AI video companies, to create a proprietary model trained on its own movie library. The goal was simple: give the AI access to decades of Lionsgate’s footage and let it learn how to produce new content.

The studio imagined being able to whip up entire scenes on demand, or maybe generate alternate shots, backgrounds, or action sequences without a full crew. It was supposed to be a technological shortcut to movie magic.

According to industry reports, what they got instead were clumsy visuals, half-coherent edits, and characters that moved like malfunctioning marionettes. The AI could copy the look of a movie, but not its rhythm, emotion, or intent. In other words: it could paint the car, but it couldn’t drive it.

As one insider told Gizmodo, the test footage looked less like a film and more like a very expensive proof-of-concept reel, all surface, no soul. Turns out, as vast as the Lionsgate library is, it is apparently too small to create a viable model to replicate the quality of the source material.

Three Worlds, One Tool That Doesn’t Get the Difference

There’s a strange irony in the world of AI right now: the tools don’t seem to understand the people using them. Everyone, from casual creators to studio professionals to visionary filmmakers, is being handed the same set of controls, even though they’re trying to make wildly different things.

On one end, you’ve got the everyday creators. These are your YouTubers, TikTokers and people who make funny videos during their lunch break. They want speed. They want convenience. If the AI gives them a silly-looking shot or an odd animation glitch, they’ll probably laugh and upload it anyway. For them, AI is like a toy box that happens to generate content really quickly. A weird result isn’t a failure; it’s part of the charm.

Then you have the working artists. These are the folks inside ad agencies, production companies and post-production studios. They’re not looking for shortcuts to creativity. They already know how to make the good stuff. What they want is help with the tedious parts: cleaning up footage, adjusting backgrounds, smoothing a scene that’s 90 percent there. For most, Adobe’s tools are the gold standard because they slot neatly into a workflow without trying to replace it. For these pros, AI isn’t a muse, it’s a highly efficient coworker who quietly handles the grunt work.

And then there are the dreamers. The filmmakers who design shots like they’re building worlds. The people who’ll spend twelve hours chasing a certain kind of light or tape a camera to a bicycle just to see what happens. They aren’t interested in “quick and easy.” They’re chasing originality. They want images nobody has ever imagined. AI, for them, isn’t yet capable of meeting that ambition.

But here’s the problem: all three groups are getting treated as if they have the same goals. Every tool tosses them into the same interface, gives them the same prompt bar and says, “Have at it.” It’s the digital version of handing both a child and a master painter the exact same starter kit and expecting both to walk away with a masterpiece.

Different people want different things from AI. Until the tools start recognizing that, they’re going to keep missing the mark for everyone.

The Real Cost of “AI Cinema”

Of course, there’s also the money problem. AI’s big selling point has always been efficiency, but that pitch starts to collapse when you try to scale from 15-second clips to 90-minute stories.

Rendering AI video that looks realistic enough for a feature film is still wildly expensive, both financially and computationally. And then you have to fix everything it gets wrong: the weird shadows, the flickering faces, the mysterious extra fingers that pop up.

By the time a studio cleans all that up, it would have been cheaper to just shoot the scene for real. And here’s the kicker: it would look better, too. Because even with all the computing power in the world, AI video still lacks the one thing real filmmaking has… continuity.

An AI movie doesn’t unfold like a story; it unravels like a highlight reel. It can generate pretty moments, but it can’t stitch them together. The result feels like an endless trailer for a film that never quite starts.

The Legal Minefield

And even if you somehow solved the technical problems, you’d still be up to your neck in legal quicksand.

Lionsgate’s AI model was trained on its own movies, but those movies contain licensed music, third-party footage, actor likenesses, and product placements. Each of those elements comes with its own set of rights, contracts, and limitations. If the AI regurgitates even a fraction of them, someone’s lawyer gets a new boat.

Now imagine OpenAI doing this with data from multiple partners across continents. Who owns the finished product? The coders? The producers? The training dataset? Until copyright law catches up, everyone’s just hoping not to get sued.

The Big Picture: A Tool, Not a Director

What OpenAI’s movie experiment and Lionsgate’s failed AI model both prove is this: AI can help make movies, but it can’t make MOVIES. Not yet.

It’s a fantastic tool for generating ideas, filling in gaps, and speeding up production. But when it comes to sustained storytelling -- characters, emotion, human nuance -- the technology still falls apart under the weight of its own cleverness.

AI is great at patterns. But stories aren’t patterns; they’re choices. They’re the messy, human bits between cause and effect, the look, the pause, the moment of doubt before the punchline lands. Those are the things no machine can fake.

So yes, OpenAI’s full-length AI movie might get finished. It might even look good. But it won’t feel like cinema, not in the way a duct-taped camera in the hands of a mad auteur still does.

Because the real secret of filmmaking has never been what you can create. It’s why you create it. And AI, for all its power, still hasn’t figured out the “why.”