Next month, the Tribeca Festival will host the premiere of Dreams of Violets, a 75-minute film that was generated entirely with artificial intelligence—and produced for a reported cost of just $2,000. The project is drawing attention not only because it reaches a major cultural platform with an unusually low budget and a fully AI-made visual language, but also because of what it depicts: a fictional dramatization of the Iranian government’s mass killing of protesters in January, using AI-created people and images.
According to reporting earlier this year by The Hollywood Reporter, Dreams of Violets presents its story as a dramatization grounded in journalistic material. Fountain 0, the company behind the film, says it was created based on “journalistic reports, photographs, and eyewitness accounts.” In other words, the film is not presented as a documentary with real footage; it is a narrative reconstruction—one that uses generative tools to create the faces, bodies, and scenes that viewers would otherwise expect to see captured by cameras.
That distinction matters, because it sits at the center of the debate around AI-generated media: what does it mean to “depict” real-world events when the people on screen are synthetic? And what responsibilities come with using AI to visualize alleged atrocities, even when the work is framed as fiction?
The film’s creators are brothers Ash and Pooya Koosha, who left Iran in 2009. Pooya co-founded Fountain 0, while Ash serves as CEO. Their personal history is part of the context for why the project exists at all: the company’s stated approach suggests an attempt to translate reported accounts into a cinematic form that can be widely distributed—without the traditional costs and logistical barriers of production. But the same factors that make the film feasible—speed, low cost, and the ability to generate imagery without filming—also raise questions about verification, consent, and the risk of turning contested or traumatic events into content that can be consumed quickly and emotionally, without the friction of evidence.
Fountain 0’s press materials describe Dreams of Violets as the first feature-length live-action film accepted to a major film festival that is completely generated by AI. That claim, if taken at face value, signals a shift in how festivals and audiences may begin to treat generative media—not as a novelty or a technical demo, but as a legitimate filmmaking format. Tribeca’s programming choice suggests the festival sees enough artistic and cultural relevance to bring the work into the mainstream conversation.
Still, the film’s subject matter ensures that the premiere won’t be discussed only in terms of technology. The Iranian government’s alleged mass killing of protesters in January has been the subject of extensive reporting and documentation by human rights organizations. The Verge’s coverage points readers to Human Rights Watch reporting on “countrywide massacres,” underscoring that the film’s narrative is tied to claims that have been investigated and described by credible sources. Dreams of Violets, however, is not simply “based on” those claims in a general sense; it is a dramatization that uses AI to create the people and images viewers will see. That means the film is not merely referencing events—it is visualizing them.
This is where the project becomes especially revealing about the current state of AI media. Generative systems can produce convincing faces, environments, and motion, but they do so without the evidentiary chain that traditional filmmaking relies on. A camera records what is in front of it; an AI model generates what it has learned from training data and prompts. Even when creators use careful sourcing—journalism, photographs, eyewitness accounts—the resulting images are still synthetic reconstructions. They can feel immediate and intimate, but they are not direct evidence.
For some viewers, that may be precisely the point. Fictional dramatizations have long been used to communicate the emotional truth of events that are difficult to capture on film. Historical reenactments, docudramas, and investigative storytelling often blend documented facts with interpretive choices. The difference now is that AI can generate a level of visual specificity at extremely low cost, potentially compressing the time between reporting and cinematic representation. When the barrier to “making images” drops dramatically, the speed at which narratives can spread also changes.
A reported $2,000 production budget makes that dynamic hard to ignore. Traditional feature production involves crews, equipment, locations, casting, post-production pipelines, and legal processes that can take months or years. With AI generation, many of those steps can be replaced by computational workflows. That doesn’t automatically mean the work is shallow—creative decisions still matter—but it does change the economics of storytelling. It also changes who can participate in filmmaking. If a small team can produce a feature-length live-action experience, then the gatekeeping power of large studios and expensive production infrastructure may weaken.
But there’s another side to the equation: when production costs fall, the temptation to treat complex events as “content” can rise. The ethical question isn’t whether AI can make films; it’s whether the incentives around speed and virality will outpace the care required for representing suffering. Dreams of Violets is already being framed as a milestone—“first feature-length live-action film…completely generated by AI”—and milestones tend to attract attention. Attention can be good for awareness, but it can also flatten nuance. A film that looks like live action can blur the line between dramatization and perceived authenticity, especially for audiences who don’t read the fine print.
Fountain 0’s stated methodology—grounding the film in journalistic reports, photographs, and eyewitness accounts—suggests the creators are trying to anchor their work in existing documentation. Yet the process of converting sources into scenes is inherently interpretive. Photographs show moments; eyewitness accounts provide perspectives; journalism provides synthesis. Turning those inputs into a coherent narrative requires choices about chronology, character behavior, setting details, and what is shown versus implied. With AI, those choices become visible as generated imagery. The film’s “people” are not just actors; they are synthetic constructs shaped by the system’s capabilities and the creators’ prompts and direction.
That raises a practical question for viewers: how should audiences evaluate the film? If it were a conventional docudrama, viewers might still ask what is dramatized and what is factual. With AI, the question becomes more complicated because the film’s visuals are not traceable to a camera. There is no original footage to inspect, no production record that can confirm what was filmed. Instead, evaluation depends on transparency from the creators: what sources were used, how they were interpreted, and what limitations exist.
At the moment, the public reporting summarized in coverage of the film emphasizes the inputs and the fact that the people and images are fully created by AI. It also notes the creators’ background and the film’s acceptance into Tribeca. What remains less clear in the available summaries is the level of detail about the production pipeline—how the AI was guided, what constraints were applied, and how the team handled sensitive content. For a project dealing with alleged mass killings, those details aren’t just technical; they’re ethical.
There is also the question of impact. A film like Dreams of Violets could influence how people understand the events it references, particularly if it circulates beyond the festival circuit. Even if the film is clearly labeled as fictional dramatization, the emotional force of realistic imagery can shape memory. Viewers may remember the faces and scenes as if they were representative, even though they are generated. That effect is not unique to AI—fiction has always shaped perception—but AI’s ability to generate photorealistic content at scale may intensify the phenomenon.
This is why the Tribeca premiere is likely to be discussed as much in terms of media ethics as in terms of innovation. Festivals have historically served as spaces where new forms are tested against cultural norms. Accepting an AI-generated feature-length film suggests that those norms are shifting. Yet the subject matter ensures that the conversation will be sharper than usual. It’s one thing to premiere an AI-generated fantasy or a stylized animation; it’s another to premiere a film that dramatizes alleged real-world violence.
The Verge’s reporting frames the film as a fictional dramatization of the Iranian government’s mass killing of protesters in January, with people and images fully created by AI. The Hollywood Reporter’s earlier coverage similarly emphasizes the AI-generated nature of the film and the grounding in journalistic and eyewitness material. Together, these reports position Dreams of Violets as both a technological milestone and a test case for how institutions handle AI representations of real events.
There’s also a broader industry implication. If a film like this can be made for $2,000 and accepted to a major festival, it signals that AI filmmaking is moving from experimental corners into mainstream validation. That could accelerate adoption across advertising, music videos, and independent cinema. It could also increase the volume of AI-generated content that claims to be “based on” real events. As that happens, audiences may need better literacy tools to interpret what they’re seeing.
In that sense, Dreams of Violets may function as a kind of stress test for the cultural ecosystem. Can festivals create frameworks for disclosure? Can press materials and screenings communicate clearly that the imagery is synthetic? Will audiences ask the right questions? And will creators develop standards for sourcing and representation when the output is visually indistinguishable from conventional live action?
The film’s trailer—shared in coverage—adds another layer to the discussion. Trailers are designed to hook viewers quickly, and they often compress context. When a trailer shows realistic scenes, it can create a sense of immediacy that discourages skepticism. That doesn’t mean the film is misleading; it means the marketing format is optimized for emotion and momentum. For AI-generated works, that optimization can be double-edged. It can help the film reach audiences, but it can also blur the boundary between dramatization and perceived documentation.
Fountain 0’s framing—based on journalistic reports, photographs, and eyewitness accounts—suggests the creators want the film to be understood as a narrative reconstruction rather than a fabricated