A Fully AI-Generated Feature Is Premiering at Tribeca. The Industry Is Out of Excuses.

On June 10, during its 25th edition, the Tribeca Festival will premiere "Dreams of Violets," a docudrama feature in which every image and every person on screen was generated by artificial intelligence. 

The film was created by brothers Ash and Pooya Koosha and made through their AI studio Fountain 0, with Ash credited as writer and director and Pooya as producer. It dramatizes reported killings during Iran's January crackdown, following five people who meet in a Tehran alley before their execution, watched from a window by a 10-year-old boy with cerebral palsy. The real death toll from those events remains difficult to verify because of a communications blackout, and the film is openly a fictional dramatization built from journalistic reports, photographs, and eyewitness accounts. The subject is gravely serious; the production method was anything but conventional. It was reportedly assembled in about three months for roughly $2,000. Its makers describe it as the first feature-length film generated entirely by AI to be accepted into the official lineup of a major film festival, where Tribeca has programmed it as a Special Events screening rather than in competition.

That last phrase is the whole story. Synthetic films have been drifting around the edges of the festival world for a while now. Cannes is the clearest case: an AI action feature surfaced only in the market orbit around the festival in May, through the Marché du Film and not the official program, and Cannes has signaled that films primarily driven by generative AI are not eligible to compete for its top prize. The Warsaw Film Festival hosted an AI documentary as an international premiere. The pattern, at least at Cannes, was to let such work be seen without granting it the status of the festival proper. Tribeca did not draw that line in the same place. It placed a fully synthetic feature in the official program. That is the line moving, and once a major festival moves it, the question of how to treat AI-generated work stops being hypothetical.

This Is No Longer a Novelty to Be Indulged

For the past few years, the visual-arts world has mostly treated generative AI as a curiosity to be handled case by case. A contest disqualifies an image here, a gallery quietly pulls a piece there, a festival tucks an AI film into a side program rather than the main slate. Every one of those decisions has been ad hoc, made under deadline pressure, with no shared rule behind it. That worked when synthetic work was a fringe presence. It does not work now that a fully AI-generated feature has been programmed at a festival of Tribeca's stature, because the absence of a rule has itself become a decision. Drift is a position. It just happens to be a position arrived at by no one in particular, which is the worst way to arrive at one.

A Festival Is Not Making One Decision. It Is Making Several.

Part of why this question keeps getting fudged is that "should a festival accept AI films" is really several different questions wearing one coat, and they have different answers. Screening a film is not the same as putting it in competition. Competition is not the same as making it eligible for craft awards. None of those is the same as deciding whether the work must be labeled for the audience, whether its tools and sources must be disclosed, or whether the real people whose likenesses and voices it uses gave consent. Cannes is useful precisely because it shows how these distinctions can be pulled apart: technical or limited AI assistance may be treated differently from films primarily driven by generative AI, and a fully synthetic feature can appear in the market orbit without being part of the official festival program. Those are three different doors, and a serious institution decides which ones a given work may walk through rather than leaving them all ajar. The questions that need answering are concrete. Does a synthetic feature compete against camera-made films or in its own category? Must it be disclosed as AI-generated so an audience knows what it is watching? Must its makers document what was generated, what tools and source materials were used, and whose likeness appears? Can it qualify for funding and distribution? Each of those is a separate lever, and each is currently being pulled, or left alone, by default.

Be Clear About Who Has to Decide, and Who Already Has

Here is where these arguments usually go wrong, so it is worth being precise. When people say the industry must take a stand, the burden tends to land, in practice, on the individual maker: the working photographer, the freelance cinematographer, the documentary filmmaker who is told to hold the line by refusing the tools or competing harder against them. That is a misallocation of responsibility, and it is unfair. A photographer in a small market has no power over what Tribeca programs, what a contest accepts, or what a streaming platform buys. Asking that person to personally stop the normalization of synthetic media is like asking a single driver to fix traffic. They can make a choice for themselves, and that choice has dignity, but it has no leverage.

The decision belongs to the institutions that actually control access, because access is the thing in question. Festival programmers and selection committees decide what gets a premiere and whether it competes. Juries decide what wins and, by extension, what the field is told to aspire to. Distributors and streaming platforms decide what reaches an audience. Grant-making bodies decide whose work gets made at all. Awards organizations decide what categories exist and what is eligible for them. The encouraging part, and the part that removes the usual excuses, is that some of these institutions have already shown they can act. During the 2023 strikes, the Writers Guild secured binding contract language establishing that AI is not a writer, that AI-generated text is not literary material, and that AI output handed to a writer cannot be used as source material or to reduce the writer's credit or pay. SAG-AFTRA won enforceable rules requiring consent and compensation for digital replicas of performers. In May 2026, the Academy went further into the awards space, moving beyond its earlier neutrality to rule that, beginning with the 99th ceremony, performances must be demonstrably carried out by humans and screenplays must be human-authored to be eligible for Oscars, while still permitting AI tools in technical categories like visual effects. Whatever one thinks of those specific rules, they prove the point: these decisions can be made deliberately and written down. The labor and awards bodies have started. The festival and programming layer, the one Tribeca just exposed, has not.

This matters for photographers specifically, even though Tribeca is a film story, because photography is already living through the same reckoning one rung down, and its institutions have begun to move too. World Press Photo has barred AI-generated images from its contest. Other competitions now permit limited AI use but only under disclosure rules, and there have been cases of real photographs losing or being pulled because judges wrongly suspected they were AI, which puts the burden on honest photographers to prove their work is genuine. There is a difference worth naming, though. Photojournalism rests on an evidentiary claim, that this happened in front of a lens, so AI threatens the very thing a news photograph is for. Cinema has always included fabrication, performance, animation, and visual effects, so for film the threat falls less on a claim of captured reality and more on labor, authorship, provenance, and credit. The institutions answering these questions need to keep that distinction in view, because the right rule for a press photo contest is not automatically the right rule for a film festival.

Taking a Position Does Not Mean Banning It

None of this is an argument that "Dreams of Violets" should not exist or should not be seen. The case for it is genuine and worth stating plainly. The Kooshas could not get access to actors, a crew, or the country the film is about; Iran is not a place a Western production walks into, and the events it depicts unfolded during a communications blackout. AI let them build a dramatization that could not have been filmed any other way. There is a real democratizing argument here too, that a tool letting someone make a feature for a few thousand dollars lowers a barrier that has kept filmmaking the province of the well-funded. That low price tag is partly an illusion, though, resting on computing power currently subsidized by tech companies racing for market share rather than on the true cost of generating the footage.

But there are hard questions the enthusiasm tends to skip, and they are the institutions' to weigh. One is provenance: synthesizing photorealistic footage of a real atrocity, however well-intentioned, blurs the line between dramatization and fabricated evidence, and a festival that frames such a film as documentary-adjacent is making a claim about truth that deserves scrutiny rather than applause. Another is ownership: the U.S. Copyright Office has held that purely AI-generated material lacks the human authorship copyright requires, so the generated footage itself may receive narrow protection or none, at least under U.S. law, where some other jurisdictions handle computer-generated works differently. Human-authored elements like the writing, the editing, and the selection and arrangement of shots can still be protectable, but a work whose entire visual surface is synthetic occupies uncertain legal ground, and distributors are wary of investing in material they cannot cleanly defend. These uncertainties are not abstract. They surface as concrete distribution problems, including errors-and-omissions insurance, chain of title, and the clearance of any real likenesses or closely mirrored real events, which can raise rights-of-publicity, defamation, or false-light exposure when a film reconstructs identifiable living people or recent events. These are not reasons to forbid the form. They are reasons the gatekeepers cannot keep waving it through without a framework.

And a framework need not be a prohibition. Cannes has already become an example of institutional line-drawing, separating official selection, competition eligibility, and market-adjacent exhibition. A festival could decide that synthetic features are welcome but must be labeled, or that they compete in their own category, or that they are eligible everywhere as long as their tools, sources, and any real likenesses are disclosed. The single most useful word here is disclosure: what was generated, what tools and reference materials were used, whether any real person's image or voice was used and with what consent, and whether human performers and writers are credited. Any of those is a defensible answer. What is not defensible is continuing to have no answer while the precedents pile up, because each unexamined decision quietly becomes the rule that governs the next one.

Tribeca has every right to program what it wants, and "Dreams of Violets" may well be a moving and important film. But the moment a fully synthetic feature stands in a major festival's official program, the festival world is out of room to stall, especially now that the guilds and the Academy have shown the rest of the industry that rules can be written. They can decide what AI-generated work is and where it belongs, openly and on the record, or they can keep letting the question answer itself one premiere, one contest, one disqualification at a time. The first path is hard and contentious. The second is easy, and it ends with a field whose most consequential rules were written by default, by people who never quite admitted they were writing them.