Over many sleepless nights in the past few months, the engineer and sometime-filmmaker Ash Koosha had one thought: this is not as easy as people think.
A British-Iranian Millennial with an earnest streak, Koosha had set about to make a movie about the Iranian regime’s January 2026 crackdown on protesters, which claimed the lives of at least 7,000 people and likely many more. With no way of safely making a film in Iran, let alone a financier or producer to back those efforts, Koosha decided to make it entirely using a mix of AI tools (mainly from Anthropic and Google).
But he was finding the process much easier said than done: crafting a meaningful narrative with just LLMs was a Herculean task, requiring a director to be in charge of departments a physical filmmaker never would.
“I slept three hours a night, I was having major headaches, it was nearly impossible,” he told The Hollywood Reporter Tuesday evening from outside an AMC screening room near Manhattan’s Union Square.
Still, some things proved a little simpler with this method. “I finished the first 30 minutes and realized I didn’t like it. So I just hit delete and started again,” he said.
The strange fruits of Koosha’s process premiered Wednesday night at the Tribeca Festival in New York — a work that was just another indie film on one hand but also a glimpse at the distorted mirror of film’s future on the other. Dreams of Violets, as Koosha named his piece, is a full 75-minute feature offering a look at modern Iran. Or, perhaps more accurately, the dream-diary of a person who knows the country but didn’t step foot anywhere near it to make his movie.
The industry has been keeping a close eye on what Koosha and his producer Tom Rogers have been doing — namely, conjuring a movie out of thin air without a single shred of an actor, location, crew member or camera. If Koosha was successful he’d have changed the entire aesthetic proposition, not to mention cost structure, of Hollywood itself. If he failed — and it was impossible to ignore that a number of folks in town were hoping he would — it would mean the AI revolution was overhyped and we could all go back to making movies the old-fashioned way.
“Tonight’s film has sparked a bit of conversation, and a bit of controversy,” Tribeca co-founder Jane Rosenthal deadpanned in introducing the film.
In an interview before the screening, Rogers, a longtime tech and media executive (he founded CNBC while running NBC Cable), offered to THR his own explanation for why this shouldn’t offend: a movie about a hotbed like Iran could never have been made conventionally; an unknown filmmaker could never have gotten the budget; and the whole process is additive more than substitutive.
“I understand the sensitivity because there will be disruptions and some jobs will go away,” said Rogers, who came to know Koosha when he employed him at his cloud-computing company. “But this is a movie that simply wouldn’t have happened if not for AI.”
Whatever you think of AI films as a general movement, the ire could feel a little misplaced when applied to this small film and its maker, who in Wednesday’s light seemed less like a job-stealing disruptor and more just a regular person feeling anger about injustice in his home country (he was raised in Iran) who deployed his skills to do something about it. “This was not a technical exercise. This is an artist finding a way to bear witness,” Rosenthal said.
Dreams of Violets can feel disjointed in spots, but perhaps not more than any impressionistic movie about a crisis. Koosha made the savvy choice of leaning into the form’s dream-like qualities, telling the story almost as though it’s being recalled by someone through the stylized haze of trauma rather than bothering with straight-ahead narrative. There are hints of emotion, and violence, and intimidation, and occasionally hope, but the whole thing can feel fractured, like postcards from a movie instead of the actual thing. A surgeon who watches as IRGC soldiers storm into the hospital and demand he step away from the protester patient he is operating on. An elderly woman recalling dancing with her husband, now gone. A sunny, wheelchair-bound boy who watches horrific acts of regime-led violence from his window. All of them collide in the streets of Tehran, but not always in our brain.
Not exactly anthology but hardly contiguous narrative, Koosha’s film works best as a kind of dispatch from a troubled moment, giving us human snippets to understand what can be an abstract violence.
A score that he generated plays over much of the movie, shifting the burden away from dialogue and longform action scenes (neither of which AI does especially well), if also grating with its insistence.
There are also, it should be noted, discontinuities and some of those gauzy semi-unreal slop-shots, though not as many as you’d think. And closeups, so many closeups. (It’s a way of limiting the discontinuity and hallucination issues.) One might expect that after a few minutes you’d settle into the movie and forget the AI production — Rosenthal asked the audience to think of what the film was about instead of how it was made — but the stylization of the product can’t but remind of the uniqueness of the process.
Also, I found it hard to escape a philosophical question: for a docudrama to have impact, how much does it matter that what we’re watching physically happened? The genre by definition requires re-enactments, and does the act of gathering a bunch actors and going to Tehran (or Morocco) give more credibility to the project than having a computer do it? As I watched Dreams of Violet I found myself toggling between two answers— the synthetic nature of the project made me feel less of what I might otherwise feel, but I also doubted every other docudrama more. Why does any fact-based film have meaning if what we’re watching is at heart a deepfake?
Of course, the models will soon refine what we see even further such that eventually we may not be able to tell something was even made by AI, and thus will stop caring about such angel/pin matters. We occupy a weird liminal space, in which AI is good enough to produce a film but not so good we can’t tell it from the human-shot kind — or not so good it doesn’t have its own uncanny-valley aesthetic. In this regard Dreams of Violets feels less like a forerunner than a snapshot of a very specific moment we may soon look back at with some curious and even amused detachment
Koosha is an unlikely conduit for an AI revolution. He believes many movies should be shot the old-fashioned way — “if you can shoot a movie in New York, why would you shoot it with an LLM?” — and doesn’t like “80 percent of what AI does.” But he firmly believes that shooting a film does not require picking up a camera, a serious shift from our existing reality. Also, he finds it artistically freeing to do it this way. “You have so many Iranian filmmakers who don’t want to go to jail. So everything in their movies becomes a metaphor,” he said. “But if you don’t need to be in the country you can just tell the story directly.”
Dreams of Violets was hardly the best or worst screening at a festival this year, but it certainly was the strangest. When the credits came up at the end and showed Koosha’s name again and again (and his brother, who did work on post), it almost played like a parody of an indie film where one person did all the work — so much so that a few audience members laughed. You kept waiting for a credit roll — of crew, of actors, of consultants — but nothing came, because no one else worked on this. Instead you got some special thanks for people who advised Koosha, Rodgers, and a shout-out to the brands that pitched in. (Koosha spent $2000 to make the movie, most of it on subscription services to the various AI models.)
But the overall reaction was positive, and the screening room of several hundred – it appeared 80 percent full — was curious and largely respectful, applauding strongly after it ended. Some of this was no doubt due to the subject matter — when you see a dedication to the victims of an IRGC crackdown, the last thing you want to do is boo — but it also seemed like people genuinely came to learn more. After the screening at the front of the theater and again outside, person after person came up to Koosha to talk about how he made the movie, how he felt about the future of cinema, how they, too, were working on an AI project. Anyone who felt he was ruining cinema, stayed home, or quiet.
That doesn’t mean there isn’t plenty of backlash waiting. Rodgers told THR he had shown the film to a distributor who loved and wanted to buy it until he found out it was made with AI; Koosha said a similar experience happened with a festival representative. Neither person had an issue themselves, they said, but just worried about the controversy. The movie ended up at Tribeca because Rodgers was sitting behind Rosenthal at a Knicks game earlier in the playoffs and got to talking about what he was working on. And you thought the Anunoby tip-in was the most eye-popping act at MSG this postseason.
Despite the cool reception from distributors, the pair haven’t given up, and still plan on shopping their movie to either a streamer or theatrical distributor.
They also plan on making more films together. Koosha said he might have a few veteran department heads in the room next time to give more feedback but otherwise plans on the same process, just in different genres.
“You could do something crazy like a movie set in outer space,” Koosha said. “They’ve never give an indie director the budget do that. But isn’t it great that with AI you don’t need one?”
Leave a Reply