Lately, every time I read about AI in entertainment, I see the same excitement.
AI can generate actors. AI can create dancers. AI can write scripts, compose music, and produce content faster than ever before.
And while all of that is impressive, I can't help but think about the people behind the performances we love.
Think about an actor for a moment.
Most people only see the finished movie. They don't see the years of auditions, rejections, acting classes, low-paying gigs, and sacrifices it took to get there. They don't see someone spending years learning how to become believable as another human being.
The same goes for dancers.
A dancer may spend ten or fifteen years training before ever getting a major opportunity. They practice until their bodies hurt. They miss birthdays, weekends, and holidays because they're rehearsing. They dedicate their lives to mastering movements that make difficult things look effortless.
Stand-up comedians go through the same struggle.
Nobody becomes a great comedian overnight. They spend years bombing on stage, testing material, learning timing, reading audiences, and finding their own voice. Every joke comes from experience. Every performance comes from lessons learned the hard way.
So when people say AI could replace performers, I think we're talking about more than technology. Read more here https://www.goldenwaymediafilms.com/ai-in-entertainment/
1 person likes this
Maria Johnsen Married with the latest in motion-capture, it does great animation (subject to the myriad other defects of generative AI video). However, it's not generating the movement in those cases, right? It's mathematical fiat that by itself, AI will only ever achieve an average of what it has been trained upon. It cannot, by definition, be original, and with rare statistical exceptions, it is incapable of anything original, especially where that is an interpretive function such as a dancer or actor undertakes.
Not yet, but soon, yes.
I completely agree that there are years of training, rejection, sacrifice, and real human experience behind a great performance. But I think we make a mistake when we use that as proof that AI will never be able to recreate what we see on screen. The audience only sees the end result.
Studios have decades of performances sitting in their libraries, and we are already seeing studios partner with AI companies. My personal theory is that AI is either already studying, or soon will be studying, facial tics, pauses, eye movements, posture, reactions, physical movement, comedic timing, and all the tiny choices that make up a performance.
We are already seeing the early version of this without anyone having to create a fully artificial actor. Adobe Premiere Pro’s Generative Extend can create new frames at the beginning or end of a shot, which means an editor can extend an actor’s movement, reaction, look, or pause beyond what was actually captured on set. It is only a couple of seconds right now, but it is still AI generating part of a performance the actor did not technically give.
That does not mean the AI trained for fifteen years, understood the character, or felt the emotion. It means it may eventually get good enough at recreating the visible result that most people cannot tell the difference. Whenever we say technology cannot do something, it usually turns out to mean not yet.
As an AI consultant focused on ethical integration in film and television, my approach is “Tools, Not Terminators.” The real question is not whether AI will eventually be able to imitate human performance. It is whether we use it to support performers or replace them, and whether the people whose work trained these systems are protected, credited, compensated, and asked for consent.
2 people like this
Hi Joshua, I get what you’re saying, and I agree with a lot of it. AI is definitely getting better at copying the small details that make performances feel real, and most of the time audiences only see the final result anyway. We’re already seeing tools that can extend or tweak real performances, so it’s not hard to imagine it going further.
Where I’m still not fully convinced is the idea that matching what something looks like on screen is the same as actually replacing a human performance. Acting, dancing, and comedy aren’t just movements and timing they come from real experience and intent, and I think that’s hard to fully recreate just from data. Even if AI gets really good at imitation, it still feels different from real interpretation.
But I do agree with your bigger point, which is that the real issue is how it gets used. Whether it ends up helping performers or replacing them is going to come down to ethics, consent, credit, and how the industry handles it.
2 people like this
No.
I think once the consolidation phase is over, all three forms, Theatre is going to be back big time as at some point people would want to see actors perform live, having said that the rest of the two categories, i.e films made where AI is being used an assistive tool and fully made AI films all of them will find their respective place in this eco-system.
and depending on the individual they can choose which one suits them the best. It's just that consolidation phase is always challenging and hence this insecurity.
Gen AI is just animation, working on a predict-the-next-pixel approach with a statistical base. It is not, is not intended to be, and never will be capable of simulating physical phenomena, or real performances.
You named the divide cleanly, Maria: imitation versus interpretation. Where I'd press is that audiences have never had access to a performer's intent. We only ever inferred it from what we could see and hear, so the question isn't whether AI can feel anything, it's whether the absence of feeling is something we can detect on screen. My worry is that for a lot of viewers, a lot of the time, it won't be. We as a species already assign emotion to inanimate objects, a car, a face in a wall socket, so a realistic human giving us every micro-gesture we read as feeling is hardly a leap. In just a few years we went from those horrifying Will Smith spaghetti videos to near-perfect realism. And that's why I don't think protecting performers can rest on "audiences will sense the difference," because often they won't, which pushes us right back to the ground you and I already share: consent, credit, and pay for the people whose work trained the system.
Joshua Young good points but - the audience has always had access to the artist's expression - intent is actually irrelevant and not the business of the artist whether anyone understands it or not. However with regard to your thoughts on whether an audience can detect the "absence of feeling" the jury came in a long time ago, and the answer is yes. Uncanny Valley, observed as robotics got into full swing in ~1970 is a clinically observed effect, essentially in our DNA, and more apparent the closer the object gets to human emulation. That Uncanny Valley applies to gen AI video has already been studied and confirmed. Here are two infographics on it: https://aerithai.com/papers/uncanny-valley-in-ai.html and https://aerithai.com/papers/uncanny-valley-interactive-page.html The takeaway is that you either ensure that the footage is clearly animation (which it is) and not supposed to be realistic (audiences have always forgiven that), or you risk alienating the majority of your audience entirely.