Did you know the average Hollywood movie takes 106 days to shoot? That's three and a half months of a massive crew burning through six figures (or higher) a day, praying the weather holds and the lead doesn't get food poisoning (and then God forbid, has another production scheduled right on top of this one, pushing reshoots out another 9 months to a year). And that's just production. Post-production, where editors and VFX teams assemble the movie, take a whole year after that.
Now think about an entirely different production model: concerts. Taylor Swift sold out 149 shows across five continents. Each one was a perfectly executed, multi-act performance captured in real-time. Her concert grossed something like $1B over the course of its run, and the film alone grossed over $261 million worldwide. Hamilton filmed a single Broadway run and became one of Disney+'s biggest launches. Many popular plays are now doing the same.
So here's the question nobody in Hollywood is asking (well, except me): what if we stopped treating filmmaking like a years-long construction project behind closed doors, and retool it into a live event?
The Pitch
Imagine a production team builds a set on a sound stage, constructing something elaborate like the kind of immersive production design that went into producing Wicked, where Nathan Crowley's team planted nine million real tulips, built a 58-ton mechanical train, and constructed entire university campuses on a sound stage. Our new production hires actors, choreographs the entire story, and rehearses it like a play. Then, for a ten-day run (or longer, if popular), they perform the full two-hour film live in front of a ticketed audience.
Multiple cameras capture every angle. Some are operated remotely, others positioned like concert rigs. The cast runs through the entire story, start to finish, each night. The director and DP use different camera setups each performance. The production performs in front of different audience energy each night.
After ten nights of shooting, the director has massive coverage of every scene. That's more than enough time to give notes from night to night, let actors try new things, and capture the kind of raw, layered performance that traditional shoots rarely afford. The actors get the luxury of performing the whole film from beginning to end, really living the performance in the right order instead of choppily assembling bits and pieces scattered across three months of fitting everything together wherever they can.
Now what makes this possible today, when it wasn't before? Generative AI, to fill the gaps in the production that would otherwise extend a production over months instead of days.
Need a wide establishing shot of a city that doesn't exist? Generate it and use it against The Volume (the giant industrial screens Disney and other productions use to fill in backgrounds when shooting on set). A stunt sequence too dangerous for a live stage? Model the actors' performances and extend the scene digitally. A close-up that wasn't quite right on any of the ten nights? Composite the best elements across takes. And best of all, you're incentivized to do as much as you can practically, which audiences love and looks better anyway, because people will see it live. But AI handles the rest.
You're using AI to complete the human performance. To fill in the world the around the production. To enhance the work of the humans instead of replace them.
We've Actually Done This Before
Before you dismiss this as crazy, consider: there's a type of production that used to work exactly this way. It's called television.
Remember the old adage, "filmed before a live studio audience"? I Love Lucy pioneered this format in 1951, shooting every episode in front of 300 ticketed viewers using a multi-camera setup that Desi Arnaz helped invent. The approach married cinema and theater, using multiple cameras to capture different angles simultaneously while a live audience added energy to every take. It became the dominant format for American network sitcoms for the next five decades. Friends, Seinfeld, Cheers, The Big Bang Theory; all filmed this way.
The format worked because network sitcoms had fixed sets and tight episode structures. The only reason you couldn't do this with movies was scale: films took months to shoot, cost exponentially more to produce, and required dozens of locations. The math never made sense.
AI changes the math. With enough cameras on set, you can capture multiple angles of every scene in a single run-through. With generative tools, you can fill in backgrounds, extend sequences, and create entire environments that would've been impossible on a practical stage. You can film the whole movie, all the way through, in two hours, for a ten-day run. You can transform the movie making process into literal event viewing.
Three Windows, One Production
The business model is what makes this more than a thought experiment. And if you understand how Hollywood currently makes money, this structure will look very familiar.
Movies already generate revenue through a waterfall of sequential "windows". Theatrical release comes first (studios split revenue roughly 50/50 with exhibitors), then premium video-on-demand (PVOD) rentals at $20-30 a pop, then licensing to streaming platforms, then linear TV and physical media. Each window extracts value from a different audience segment. Studios love this structure because the post-theatrical windows are high-margin revenue streams with almost no additional cost.
This model adds a brand new window at the very top of the waterfall: live.
Window 1: The live event. Tickets sell for premium prices. Audiences get a front row seat to a movie being made, making them feel like a part of the production. Think of it like the Eras Tour, except instead of songs, you're watching a story unfold in real-time on a built set with professional actors. Imagine you could see Ryan Gosling or Zendaya or Michael B. Jordan or Michelle Yeoh act in person, on a massive set, performing a full narrative in front of your eyes. This initial audience becomes your first marketing engine. They leave buzzing, posting clips (where allowed), telling friends. Word of mouth around a film spreads fast; clips go viral, creating buzz and UGC-style marketing before the "finished film" is even done (though after the first couple nights, you'll have enough content to already produce a trailer). Built-in virality before the film even exists in its final form.
As they say in Hollywood, the movie is never really done; each draft of the script, each day of production, and each cut of the edit is another pass at rewriting it. The final cut is never really the "final" draft. You just "finally" decide to release it. How is this any different from that?
Window 2: Theatrical release. The edited, polished, AI-enhanced film hits theaters weeks (not months) after production wraps. The live event audiences become your opening-weekend evangelists. Everyone else wants to see the movie they've heard so much about.
Window 3: PVOD and streaming. Premium rentals, then wide release on your platform of choice.
Each of these windows funds the next. The live tickets offset production costs before the movie is even finished. The theatrical run benefits from word-of-mouth that's already been building for weeks (they say these days that marketing a film is often $50M-$60M, or half the cost of some films). PVOD captures the "I'll pay to see it now" audience. And streaming gets a title with a story behind it, not another anonymous release in the algorithm.
Without This Model, There's a Huge Storm Brewing.
Now, at this point you might be thinking: this idea is dumb; you can't film a movie in ten days, and even if you could, many hundreds of thousands to millions of production teams rely on long productions to make a living.
Here's the thing about that 106-day production timeline: nobody actually wants it.
No director wants to be away from their family for four months. No actor wants to spend half a year on location grinding through a single project. No grip or gaffer wants the physical toll of consecutive 14-hour days, week after week, for a hundred days straight. What people in this industry want is steady work, fair pay, and the opportunity to make something amazing when they show up.
Ask most working crew members what the best gig in Hollywood is, and they'll tell you: a long-running TV show. Reliable income. The same team, year after year. You know where you're working and when your next check is coming. You can plan a life around it. For the below-the-line talent (the grips, electricians, set decorators, camera assistants, stagehands) who are most vulnerable to being replaced by AI, that kind of stability is everything. Think of them like junior developers in tech, except they're protected by unions.
And right now, those jobs are disappearing. L.A. County lost roughly 42,000 entertainment jobs between 2022 and 2024. Television production in greater Los Angeles has dropped 58% since its peak in 2021. Productions keep leaving California for states with better tax incentives, or for countries where labor is cheaper. Between 2020 and 2024, an estimated 71% of projects that didn't get California tax credits moved out of state. The result is a hollowed-out workforce, where veteran prop masters are applying to Costco and camera operators are driving ferries.
So what could bring consistent, paying gigs back?
How about doing the whole production in one location. Working on sound stages. Building real sets that real audiences will see in person. Y'know, the production model that made Hollywood "Hollywood" to begin with; filming here, in Hollywood?!
This model creates jobs for craftspeople, stagehands, and crew, protecting them from NVIDIA GPUs replacing all their hard work because building things in person will have actual audiences (and dollars) attached to it. It'll create consistent, repeatable work at the same stages, booked weeks to months to years in advance because the tickets are already sold. The local production infrastructure gets steady revenue instead of feast-or-famine project cycles. And if GenAI takes off like everyone thinks it will, famine seems most likely (without this model).
The obvious objection: building elaborate sets is expensive and laborious. And yes, AI will certainly try to eliminate those jobs by generating entire movies; no humans required. But what if those productions could earn their keep? What if it was worth it to build a set on the scale of Wicked's $145 million production, because real audiences would pay to see real actors perform on it, live, in person? Sure, it wouldn't justify the entire cost, but it could generate its fair share and offset marketing costs and shrink production timelines. What's not to like?
I say this with genuine love and worry. All of my friends work in production in some form or fashion, be it behind the scenes, on set, or in the writer's room. I WANT them to continue making a living, and to have plenty of work for years to come. If Hollywood doesn't change its production model, and do what it does best (create wonder, spectacle, and daring feats of honest humanity put on screen) then it could lose out to the slop driven algorithms that are already starting to fill our feeds.
People want to see people. People want to watch real people. And clearly, they want to do it in person, with other people. I repeat again: what's not to like?
Why This Could Actually Work
Now you might be wondering: surely genAI can't do ALL the work that takes a full year of post production. After all, when it comes to Marvel movies and big budget productions, most of the cost goes into that years-long post production cycle. And what about those jobs that could be lost?
Let's separate these into two arguments: the technology, and the impact of the technology. To start, the technology is closer than you think.
Generative video tools like SeeDance, Sora, Veo, and Kling can already produce photorealistic footage from text and image prompts. They're not good enough to make a whole film from scratch (not yet, and maybe not ever for anything audiences actually trust). But they're very good at extending existing footage, filling in backgrounds, generating B-roll, and creating VFX shots that would've cost millions two years ago (or currently).
The editing timeline collapses too. AI-powered editing tools can sync multi-camera footage, color-grade in real time, and rough-cut assemblies overnight. What used to take a year of post-production could realistically shrink to weeks.
And the live performance format solves one of Hollywood's most expensive problems: coverage. Traditional film shoots capture scenes in fragments. One angle, then reset, then another angle, then reset. It's incredibly time-consuming. A live multi-camera capture gets you everything at once. Ten nights of full run-throughs gives a director more raw material than most traditional shoots produce in three months. If you need something else, you have enough footage that you can stitch it together with the assets you've created. Actors' integrity is protected because real people saw them perform the movie, in real time.
Now the job question. This is where the argument actually flips.
The VFX industry has been in crisis for years. Studios squeeze VFX houses on pricing, demand impossible timelines, and then move to the cheapest bidder in another country. California lost roughly 42,000 entertainment jobs between 2022 and 2024 across ALL production categories. That includes VFX jobs.
The current system concentrates production into a handful of mega-budget tentpoles that employ massive crews for short bursts, then leave everyone scrambling for the next gig.
Meanwhile, a live-ticketed film production model inverts that dynamic. If the production timeline compresses from months to weeks, budgets shrink. If budgets shrink, the barrier to making a film drops dramatically. More films get made. More productions mean more consistent work for more crews in more locations.
And here's where it gets really interesting: this model doesn't stay in Hollywood. If you can film a full movie on a sound stage in ten days, edit it with AI in weeks, and release it through the same waterfall pay windows the studios use, smaller production teams all over the world can compete. You don't need a $200 million budget to make something that looks like a $200 million production.
We already have proof. Flow, the Latvian animated film directed by Gints Zilbalodis, won the Oscar for Best Animated Feature in 2025. It beat Inside Out 2 ($200M budget) and The Wild Robot ($78M). Flow was made by a team of about 40 people for $3.5 million, using the free open-source software Blender. As Zilbalodis said in his acceptance speech: "Any kid now has tools that are used to make Academy Award-winning films."
That was animation. Now apply the same logic to live-action. A regional theater company with a talented cast, a built set, a bank of cameras, and access to generative AI post-production tools could produce a feature film on a fraction of a studio budget. Small teams of VFX artists can band together and use local tools like LTX Studio and LTX Desktop plus Blender to make their own movies and compete with Hollywood. The human parts can stay human; you can shoot with real actors, use real performers to voice your characters, and do motion cap without wildly expensive rigs that previously only Hollywood could afford. The rest of it becomes easier, cheaper, and keeps getting better as these models progress. There is slowly becoming very little excuse to not just create things as a creator.
The tools are democratizing. As a small team outside Hollywood, you don't need them anymore. Inside Hollywood, you need to push the system to do something to keep up. The live-film model gives those tools a viable business structure to plug into: sell tickets to the live run, release the finished film, collect from every window downstream.
And if this works in Hollywood, it could work elsewhere. Local talent can create local productions that punch above their weight. More productions, dispersed across more cities, employing more crews on steadier schedules. That's how you protect jobs in an industry being hollowed out by consolidation and runaway production. That's how you create jobs for creatives in cities and locations where previously they'd have to leave to do what they love. Just saying y'all. This could be a new renaissance for the entire global film industry.
Now Add Robotics, and It Gets Insane
Everything above assumes human actors on a practical set, or human artists behind a computer. But what about the creatures, the monsters, the characters that can't be played by a person?
Go see Stranger Things: The First Shadow on Broadway. A towering Mind Flayer descends from the ceiling. Characters levitate. Limbs snap through telekinesis. Pyrotechnics fire. Creature effects that would've required a full VFX pipeline on screen are happening live, in front of you, through stagecraft so seamless it plays like cinematic editing. The show received a special Tony Award for its technical effects. And the part that matters most for this argument: Netflix is filming the Broadway production right now for a future streaming release. Live performance → filmed version → streaming. Sound familiar?
Now layer in what's happening in robotics. Disney's Imagineering team has built free-roaming animatronic characters that use reinforcement learning to walk, balance, and react to their environment. Their Spider-Man "stuntronics" robot can launch into the air and perform flips autonomously. Their latest animatronic figures can blush, tear up, and produce micro-expressions that rival what you'd see on a close-up in a Pixar film. The animatronics market is projected to reach $18 billion by 2032.
Connect these dots. Instead of CGI bad guys, imagine actual robotic creatures on stage, performing live alongside human actors. Dinosaurs. Aliens. Fantasy creatures. Characters that used to exist only in rendering farms, built as physical animatronics that audiences can see with their own eyes. The practical effects create the magic in the room. AI fills in the environments and extends the sequences for the filmed version.
The live audience gasps. The cameras capture it. And when the final movie hits theaters, you get the best of both worlds: performances grounded in physical reality, enhanced by generative tools that would've cost ten times as much just five years ago.
The Hard Problems (And They're Real)
Let's not pretend this is simple.
Acting style. Stage acting and screen acting are fundamentally different disciplines. Stage actors project to the back row. Film actors communicate with their eyes. A performance that lands in Row 12 often reads as overacting on a close-up. You'd need performers who can split the difference, or you'd need to embrace the theatricality as a deliberate aesthetic. (Honestly, that could be a feature. Some of the most memorable screen performances have theatrical DNA.)
Alternatively, you could use generative editing to stitch together a "rough cut" of the film for audiences in the cheap seats to watch on giant jumbotrons. So you're in an arena like setting, where the production is happening in the center of the coliseum, and those who can't see up close can watch the film on the big screen. Eventually, with advanced-enough holographic technology, you could even stream a performance into an empty arena from an on-set location live, and it would feel as if you're watching it in person. But that's a topic for another decade.
Sound. A live audience means ambient noise. Either you're wiring every actor with individual mics and accepting some bleed, or you're re-recording dialogue in post (ADR), which is expensive and defeats part of the efficiency argument. Smart sound design could minimize this, but it's a genuine engineering challenge. And luckily, generative AI can help with this, too. Today's audio models from Adobe and others can reduce background noise and isolate those speaking. Running these models in real time could isolate the sounds you want, introduce sound design in real time, and even modulate how you distribute the sound to the live audience across multiple speakers without picking up an echo or feedback loop. Difficult? Yes. But not impossible.
Actor likeness and AI. This is the big one. Post-SAG strike, the rules around AI-generated performances are still being written. Any model where generative AI modifies, extends, or composites an actor's performance needs ironclad agreements about approval rights. Actors would need final say over any AI-altered version of their work. This is the deal that could make or break this whole process. But if you get enough coverage, you shouldn't need any generated takes of the actors themselves. And you'll have the live audiences there in person to verify the actor's performance. I'm not worried about this. We just need actors to control their own likeness and be paid for loaning it out to productions, and we're set.
Scale limitations. This works beautifully for dialogue-driven films, contained thrillers, character studies, and stage-friendly genres. It's harder to imagine for massive action blockbusters or films that require dozens of real-world locations. I can imagine it though, and it would be sick. I imagine it would not be unlike Universal Studios' studio backlot tour, where you could have audiences on some sort of drivable rig, where the audience could follow along with a huge actions sequence that takes place over miles of footage. Though you could argue that probably wouldn't be necessary, and for that sort of thing, that's exactly where AI scene generation fills the gap most effectively.
In my opinion, this model opens up the landscape so that some films can still be filmed on location, the old fashioned way. For films that want to film on location, that option will oftentimes make sense. Does it make sense to film on location for a Marvel movie (which often take place on other planets), or a period piece (where you have to recreate old architecture anyway), or a contained thriller (where the entire film takes place in a few locations)? Probably not. Does it make sense to film on location for a small drama, or a Fast and the Furious film, or an international spy thriller set in an iconic European city set entirely in daytime? Yeah, probably.
The Deeper Question
Here's what this concept is really about, underneath the logistics.
The film industry is caught between two futures. In one, AI gradually replaces human creativity and human labor, costs plummet, and movies become content generated at scale with diminishing artistic returns. In the other, human performance becomes the scarce resource that audiences pay premium prices for, and AI handles the commodity work around it. We build a system that benefits working in entertainment as a career, that supports creative talent, and that reduces production costs through diminishing production timelines, not how many films get produced or how many people get to work on them or how high-caliber the quality is.
This live-film model bets on the second future. It says the thing people actually value is watching talented humans perform, whether that's set-builders or actors or cinematographers or editors stitching a rough cut together live. The product is the performance of the whole production, working together, making something for people to enjoy. If audiences aren't going to the movie theaters because they can stream the content at home, what will bring them back? Event viewing, and live in person productions. This model combines both.
And performers want to perform in front of people. This is why so many screen actors eventually try their hand at theater, or grew up in it. This model combines theater and movie making under a business model that actually makes sense. You sell out the live production for the full 10-15 day run. If it's a hit, or you need more time, you sell more tickets. The local crews get consistent, reliable work at sound stages booked well in advance. You can re-use dormant office buildings and turn them into more live production venues. Then you get the full waterfall of pay windows Hollywood already loves (theatrical, PVOD, Pay-1, streaming). You just have a new one at the top: live.
It's the same bet that live music made when streaming killed album sales. Recorded music became free. Live performance became priceless.
And the timing matters. Audiences are already trained to pay for experiential entertainment. Immersive theater (Sleep No More, Punchdrunk's productions) has proven people will pay $200+ to watch a story performed live in front of them. Concert residencies have proven the economic model of a performer in one city for a set run. What's missing is combining these patterns with film production and AI post-production into a single pipeline.
What Would Need to Be True
For this to move from thought experiment to reality, a few things have to land:
- Generative video tools need to reliably match the visual quality of live-captured footage. We're close but not there yet. Give it 12-18 months (if not sooner).
- A director willing to work in this format would need a theater background and a film eye, or just love one-ers. Someone like Alejandro González Iñárritu who directed The Revenant, or Greta Gerwig, who directed Barbie's massive practical sets, or Denis Villeneuve, who obsesses over in-camera practical effects.
- SAG-AFTRA and the major guilds would need to negotiate a framework for AI-assisted performance modification that actors actually trust. This could potentially be the slowest variable, but they need to iron this out ASAP or Hollywood will use more trash digital actors like Tilly Norwood. Boooo.
- A studio or streaming platform willing to bet on a proof-of-concept. Probably a mid-budget film ($15-30M) that's dialogue-heavy and built for a contained set to start. Then work your way up to the most high production budget films of all.
The first person to pull this off might invent an entirely new art form. One where the audience is part of how the finished product gets made.
AI can already generate entire short films from a text prompt. So the most valuable thing a filmmaker can offer might be the one thing AI can't fake: a room full of humans, performing live, creating something unrepeatable.
Could the death of cinema actually be the beginning of its next act?
