I teach an AI and filmmaking course at USC’s School of Cinematic Arts, and lately I’ve been assembling my classes the night before rather than planning each session in advance. We browse platforms like X, Substack, and YouTube and select the most provocative articles and video clips to publish the next morning.
This proves how quickly the relationship between artificial intelligence and filmmaking is evolving. Each week brings new (often surprising) developments.
In class the next morning, my students and I discuss the changes in ethics, aesthetics, and storytelling that occur when collaborating with AI.
And we are not alone. Across Hollywood, everyone seems to have an idea about what’s next, including aspiring actors, filmmakers, stars, screenwriters, and studio executives. But I think three trends in particular will be hot topics at this year’s Oscar parties.
There’s nothing creepy about this clip
In February 2026, a 15-second AI-generated video clip of Tom Cruise wrestling Brad Pitt on a burnt-out freeway overpass went viral.
Depending on the viewer, the video evoked either admiration, anger, or existential constriction.
This video, created by Irish filmmaker Ruairi Robinson using a generative AI tool called Seedance 2.0, marked another milestone in the aggressive growth of AI tools.
Seedance 2.0 – developed by ByteDance, the Chinese company that created TikTok – is one of the many AI tools currently available for creating short video clips. But unlike most AI-generated videos, Pitt and Cruise don’t seem creepy, eerie, or lifelike in the clip, which mimics live-action footage almost perfectly. The appearance of two A-list stars in a fairly realistic scene created by a relatively unknown director using stolen likenesses shocked the industry.
The backlash was swift. Disney sent the cease-and-desist letter, claiming the video was generated from a data set that likely contained Disney copyrighted characters. Actors union SAG-AFTRA called the video a “blatant violation” of the actors’ likeness and voice.
“SAG-AFTRA stands with the studio in condemning the blatant copyright infringement enabled by Bytedance’s new AI video model Seedance 2.0,” the guild said in a statement. The practice “undermines people’s ability to earn a living” while ignoring “fundamental principles of law, ethics, industry standards and consent,” the guild added.
After watching the video, the class explored the ethics of using someone’s likeness without permission, the challenges faced by actors who build their careers on their unique ability to embody a character, and what the future holds for our understanding of acting.
If filmmakers can make fake actors perform accurately, what will happen to human actors?
with the old one
Since 2023, the Las Vegas Strip skyline has been dominated by an illuminated sphere called the Sphere, an entertainment complex with 360-degree LED screens covering 160,000 square feet (14,864 square meters). Sphere recently surpassed 2 million ticket sales for its re-filming of the 1939 classic film “The Wizard of Oz.”
The film, which premiered in August 2024, was shortened, the colors enhanced, and stretched to fill the entire interior of the dome. AI was used to transfer the image from the film’s original modest aspect ratio to the giant dome. This required generating a new image around the edges of the original shot, a method known as “AI outpainting.” This technology was also introduced to increase the resolution of the original film and enhance certain scenes.
Some critics worried that this rather radical expansion of the original classic would offend viewers. In return, the sphere attracted people willing to shell out $100 to $200 per ticket.
Not bad for a 1939 movie about a girl in Kansas.
Following the huge success of “The Wizard of Oz,” experts expect producers to explore the film archives for other potential hits and enhance them with AI before screening them at various venues, including IMAX theaters and the Cosm, another 360-degree dome in Los Angeles, Dallas and Atlanta.
Or, AI could be used to create material that was never completed in historical films.
In fact, The New Yorker recently profiled AI media entrepreneur Edward Saatchi, who is working to recreate and reincorporate lost footage from Orson Welles’ 1942 feature film The Ambersons. While Mr. Welles was filming the documentary in Brazil, executives at RKO Radio Pictures re-edited the film without Mr. Welles’ approval after poor preview results. They cut about 45 minutes, replaced the original ending with a happier one, and destroyed most of the deleted footage.
Saatchi’s idea is to build a dataset that includes not only existing movies, but also scripts, notes, images, and even new performances by actors. We will then use our AI platform, Showrunner, to create new scenes from this data.
Director Saatchi wants to honor the director’s creative vision by making the film he originally intended, but some troubling issues arise as a result of his efforts.
Is it appropriate to modify existing artwork without the input of its creator? Isn’t there something sacred about movies, such as the director’s intentions or the original actors’ performances? To what extent should these issues be ignored if remakes of older films can introduce films to new audiences?
Have there been fewer opportunities?
There is an undercurrent of anxiety in my class as well. My students often wonder what will happen when they graduate.
They fear that within a year or two, entry-level jobs in the film industry, from concept artists to apprentice-level editors, will be replaced by AI before they even enter the workforce.
They have reason to be afraid.
In 2024, the Animation Guild released a sobering report claiming that by 2026, “creative workers will face an era of disruption defined by the consolidation of some jobs, the replacement of existing jobs with new ones, and the complete elimination of many jobs.”
Some of those predictions have come true, with 41,000 jobs disappearing in film and television in Los Angeles County alone over the past three years.
But I’ve tried to counter this grim statistic with some stories of thoughtful practice.
For example, Paul Trillo, a filmmaker at AI studio Asteria, talked about how to put artists at the center of the process. When detailing the company’s work on singer-songwriter Cuco’s music video, he was keen to highlight the number of artists working on the project. Yes, AI tools were used. However, they were integrated as an alternative to boring tasks rather than creative practices.
“Instead of removing [artists] That process actually allowed us to do more and allow our smaller team to dream bigger,” Trillo explains at the end of the video.
In January 2026, management consulting firm McKinsey released a report that largely echoed Trillo’s positive outlook. We predict further adoption of AI across industries. But it also shows how this technology can lead to different types of jobs and open up new possibilities. For example, as AI-generated scenes become more common, studios will need technicians who know how to blend real footage with digitally created worlds. And by reducing the cost of producing sophisticated films and shows, AI could enable more “micro-studio” and independent filmmakers to produce professional-quality content.
At the same time, the report quotes a studio executive who acknowledged that AI could represent “the most profound platform shift our industry has ever seen.”
So it’s no wonder my students, along with various critics, commentators, and industry experts, are nervous.
However, from my standpoint, I am confident that the industry can survive this fundamental disruption. It has adapted to major changes in the past, including the addition of audio in the 1920s, videotape in the 1980s, and the threat of streaming in the 2000s.
After all, people are always hungry for new, well-told stories. The filmmaking tools and job market may be in transition, but the core need for storytelling will never go away.
author Holly Willis, Professor of Film Arts; University of Southern California
This article is republished from The Conversation under a Creative Commons license. Read the original article.
![]()

