At Tribeca, an All-AI Iran Movie Shows Film’s Future (Or at Least Its Messy Present)

For months, Ash Koosha found himself awake deep into the night, wrestling with a realization that many AI optimists tend to gloss over: making a movie this way is far harder than it sounds.

Koosha, a British-Iranian millennial known both as an engineer and an occasional filmmaker, had set out to create a film about the Iranian government’s brutal suppression of protesters in January 2026, a crackdown reported to have killed at least 7,000 people, with the true number likely higher. Because filming safely inside Iran was impossible, and because he had no traditional financing, producer, cast, crew, or access to locations, Koosha turned to artificial intelligence. Using a range of AI systems, especially tools from Anthropic and Google, he attempted to build the entire film digitally.

The experience was punishing. Koosha said that shaping a serious story through large language models and generative tools required him to take on responsibilities that, in a conventional production, would be spread across many departments.

Speaking outside an AMC theater near Union Square in Manhattan, he described the process as exhausting. He was sleeping only a few hours a night, dealing with severe headaches, and constantly running into the limits of the technology.

There were advantages, though. At one point, after completing the first half-hour of the movie, he decided he was unhappy with it. In a normal production, that would have meant lost money, time, and labor. In his case, he simply erased it and began again.

The finished result, titled Dreams of Violets, premiered Wednesday night at the Tribeca Festival in New York. On one level, it was another small independent film trying to say something about political violence. On another, it felt like an early look at a possible future for cinema — strange, unsettling, and not entirely human in its texture.

The 75-minute feature is set around contemporary Iran, though calling it a direct portrait of the country may be misleading. It feels more like a memory, or a dream, created by someone who knows Iran intimately but never physically returned there to make the film.

Hollywood has been watching Koosha and producer Tom Rogers closely because their project tests a provocative idea: can a feature film be made without actors, sets, cameras, locations, or a traditional crew? If it works, the implications for the movie business could be enormous, both aesthetically and financially. If it fails, skeptics can point to it as proof that AI filmmaking is nowhere near replacing conventional production.

Tribeca co-founder Jane Rosenthal acknowledged the tension before the screening, noting dryly that the film had already stirred conversation and controversy.

Rogers, a longtime media and technology executive who helped launch CNBC during his time at NBC Cable, argued that the criticism was misplaced. In his view, a film like this could not have been made through ordinary means. A story about Iran would have been too dangerous to shoot there, too expensive for an unknown filmmaker to mount elsewhere, and too difficult to finance. AI, he said, did not replace an existing production. It allowed a nonexistent one to happen.

He admitted that AI will disrupt jobs and that some roles may disappear, but insisted that this particular movie was an addition to the cultural landscape rather than a substitute for a traditional film.

Whatever larger concerns exist about AI-generated cinema, they seemed somewhat complicated in Koosha’s case. He did not come across as a tech-world provocateur trying to eliminate workers. Instead, he appeared to be an artist responding to violence in the country where he grew up and using the tools available to him. Rosenthal framed the movie in similar terms, describing it not as a technological stunt but as an artist’s attempt to bear witness.

As a film, Dreams of Violets is uneven, though not necessarily more so than other impressionistic works about trauma and political catastrophe. Koosha wisely embraces the unnatural, dreamlike qualities of the medium rather than trying to hide them. The film does not unfold as a straightforward plot. Instead, it moves through fragments: a doctor forced aside by IRGC soldiers while operating on an injured protester; an older woman remembering a dance with her late husband; a cheerful boy in a wheelchair watching violence unfold from his window. These figures eventually converge in Tehran’s streets, but the connections are more emotional than narrative.

The movie is not quite an anthology, but it also does not operate like a conventional drama. It is strongest when it functions as a series of glimpses from a national crisis, offering small human moments within a larger atmosphere of fear, grief, and resistance.

Koosha’s AI-generated score runs through much of the film. That choice helps reduce the burden on dialogue and extended action scenes, two areas where generative AI still struggles. At times, however, the music becomes overly insistent.

The movie also contains the kinds of visual issues viewers now associate with AI: soft, unreal images, occasional continuity problems, and a slightly smeared quality in certain shots. Still, there are fewer of these problems than one might expect. Koosha relies heavily on close-ups, which appears to be a way of controlling the technology’s tendency to lose consistency in wider or more complex scenes.

Rosenthal urged viewers to focus on what the film was about rather than how it was made. But that proved difficult. The look and feel of the movie constantly remind the audience that this is not a traditionally filmed work.

The project also raises a deeper question about docudrama itself. When a film is based on real events, how much does it matter whether what appears on screen physically occurred in front of a camera? Dramatized historical films already rely on actors, constructed scenes, and recreated events. Does staging those recreations with performers in a real location give them more authority than generating them with a machine? Watching Dreams of Violets can make the viewer feel both less emotionally connected to this particular film and more suspicious of the truth claims made by all dramatized films. If fact-based cinema is always, in some sense, an imitation, then where exactly is the line between reenactment and deepfake?

That question may become less visible as AI tools improve. Eventually, audiences may not be able to tell whether a film was generated or shot. For now, though, we are in an awkward middle stage. AI is advanced enough to create a feature-length movie, but not advanced enough to disguise its own artificiality. In that sense, Dreams of Violets may be less a preview of cinema’s final destination than a record of this brief transitional moment.

Koosha himself is not a simple evangelist for AI. He does not believe every movie should be made this way. He argued that if a filmmaker can shoot in New York, for example, there is no obvious reason to use AI instead. He also said he dislikes much of what AI produces. But he does believe that filmmaking no longer has to depend on physically holding a camera.

For him, the method is especially powerful in cases where traditional production is dangerous or impossible. Iranian filmmakers, he said, often have to bury their meaning in metaphor to avoid punishment. But if a filmmaker does not need to enter the country, he argued, the story can be told more directly.

The screening itself was unusual, though not a disaster. Dreams of Violets was neither among the year’s best festival films nor its worst, but it was certainly one of the strangest. When the credits rolled, Koosha’s name appeared repeatedly, along with that of his brother, who helped with postproduction. The repetition almost became comic, like an exaggerated version of an indie film where one person does everything. Some audience members laughed, seemingly expecting a longer list of actors, crew members, consultants, and department heads that never appeared.

There was no traditional cast or crew to credit. Instead, the film offered thanks to advisers and to the technology companies whose tools made the project possible. Koosha said the entire film cost around $2,000, most of which went toward subscriptions to AI platforms.

The audience response was generally warm. The room, which held several hundred people and appeared mostly full, treated the film seriously and applauded strongly at the end. Some of that reaction may have been shaped by the subject matter; a film dedicated to victims of an IRGC crackdown is not something audiences are likely to jeer. But there was also genuine curiosity. After the screening, people approached Koosha to ask how he had made the film, what he thought AI meant for cinema, and to mention their own AI-driven projects. Those most opposed to the concept either did not attend or chose not to speak up.

Still, resistance from the industry remains. Rogers said one distributor responded enthusiastically to the movie until learning it had been made with AI. Koosha described a similar reaction from a festival representative. In both cases, they said, the people did not necessarily object personally but feared the controversy that might follow.

The film reached Tribeca through an unlikely route. Rogers happened to sit behind Rosenthal at a New York Knicks playoff game and struck up a conversation about the project. That chance encounter ultimately helped bring the movie to the festival.

Despite hesitation from distributors, Koosha and Rogers are still looking for a home for the film, either through streaming or theatrical release. They also intend to keep collaborating. Koosha said that next time he may bring in a few experienced department heads to offer feedback, but he expects to use a similar AI-centered process, possibly in other genres.