Hollywood's Digital Ghost Problem: When AI Can Recreate Anyone, Who Owns the Performance?
Picture a scene: a major studio is in post-production on a blockbuster. A supporting character needs three additional scenes that weren't in the original shoot. The actor who played that character is unavailable — maybe they're filming something else, maybe the budget doesn't support a recall, maybe, in the most unsettling version of this story, they passed away years ago.
In 2024, the studio doesn't necessarily need to recast. They might not need the actor at all.
This is not a hypothetical pulled from a Black Mirror episode. It is a real, active conversation happening in studios, legal offices, union negotiations, and tech development labs across Los Angeles and Silicon Valley right now. And it's moving faster than most people in the entertainment industry are comfortable admitting.
What the Technology Can Actually Do
Let's be honest about where deepfake and AI generation technology actually stands, because there's a lot of hype on both sides. The sensational version — fully AI-generated feature films indistinguishable from live-action productions — is still a ways off for most practical applications. But the incremental version? That's already here and being used.
Background performers in crowd scenes can now be digitally multiplied, repositioned, or entirely replaced with AI-generated figures that move, react, and behave convincingly. Facial de-aging, which started as a painstaking VFX process requiring extensive reference footage, has gotten dramatically cheaper and faster. Voice cloning has reached a point where a few hours of recorded speech can produce a synthetic voice that passes casual listening tests.
Companies like Metaphysic (which made waves at AGT with its hyper-real deepfake performances) and Respeecher (used on The Mandalorian to recreate a young Luke Skywalker's voice) have moved from novelty to legitimate production tools. These aren't garage projects — they're vendor relationships with major studios.
The more controversial frontier is full digital recreation of real performers. James Dean, Marilyn Monroe, and other deceased celebrities have already been licensed for AI-generated appearances in commercials and projects, sparking significant public debate about consent, legacy, and the commodification of likeness.
The Union Fight Is Already Underway
If you followed the 2023 SAG-AFTRA strike, you know this isn't abstract territory. One of the central sticking points — the one that actually extended the strike for months — was AI. Specifically, the question of whether studios could scan a background actor's likeness once, pay them for a single day of work, and then use that digital replica indefinitely without additional compensation.
The deal that eventually ended the strike included some protections, but many performers and advocates consider them a starting point rather than a solution. The language around consent, compensation for AI use, and the definition of "likeness" is still being litigated in real time — in courtrooms, in contract negotiations, and in the court of public opinion.
The Writers Guild fought similar battles around AI-generated scripts. The consensus that emerged — AI can be a tool, but human writers must be credited and compensated for work that uses their material — is likely to become the template for how the industry approaches AI performance as well. The question is whether that template can keep pace with the technology.
What Creators Are Actually Worried About
Talk to working actors, voice performers, stunt coordinators, and background artists, and the anxiety isn't primarily about the sci-fi scenario where AI replaces movie stars. It's more granular than that.
Voice actors are watching the commercial and audiobook industries experiment with AI-generated narration at a fraction of the cost of hiring a human. Background performers — the people who populate the diners, office buildings, and city streets in every scene you've ever watched — are seeing their work systematically replaced by digital crowd generation. These aren't glamorous Hollywood careers, but they're real jobs that support real people, and they're quietly disappearing from production budgets.
For independent creators operating in the digital space, the calculus is even stranger. On one hand, AI tools offer solo creators capabilities that previously required entire production teams — voice synthesis, digital avatars, realistic visual effects. On the other hand, those same tools are being used by well-funded competitors to flood platforms with content at a volume no human creator can match.
The Philosophical Knot Nobody Wants to Untangle
Here's the question that keeps surfacing in every serious conversation about AI in entertainment: what is a performance, actually?
When an AI system trained on thousands of hours of an actor's work generates a new scene in that actor's style, is that a performance? Is it art? Is it theft? The legal frameworks we have — copyright, right of publicity, contract law — were built for a world where human beings made creative decisions. They're being stretched in ways that sometimes make them snap.
Deceased performers present the sharpest version of this problem. Studios argue that licensed digital recreation honors a legacy and gives audiences more of what they love. Critics argue it's exploitation dressed up as tribute, denying the finality that gives any artist's body of work its meaning. Neither side is entirely wrong, which is what makes it genuinely hard.
For living performers, the consent question should theoretically be simpler — you either agree to license your likeness for AI use or you don't. But the power dynamics of the entertainment industry mean that "consent" can be complicated by contract pressure, career necessity, and terms buried in agreements most performers don't fully understand.
The Signal From Here
AI-generated entertainment isn't going to be stopped. That's probably the most honest thing to say about it. The technology is advancing, the cost efficiencies are real, and studios facing brutal economic pressure are going to use every tool available to them.
What the industry — and frankly, the audiences who consume it — gets to decide is the terms. How much human creative labor should be required in a production that calls itself entertainment? What does informed consent look like when your face can become a perpetual digital asset? And at some point, does AI-generated content need its own label, the way organic food or fair trade coffee carries a certification that lets consumers make informed choices?
Those aren't just Hollywood problems. They're signal breaks — moments where the old broadcast of how entertainment works gets interrupted, and something new has to be negotiated in its place. What comes through on the other side of that static is still being written.