Fountain 0’s Ash Koosha says “AI film” risked being rejected for $2,000 Tribeca debut

When Fountain 0 announced Dreams of Violets as an AI movie that cost $2,000 to make, the reaction was immediate: eyebrows raised, debate already raging. The key detail, though, is that the creators insist this is not a “press a button” film. They describe it as a 75-minute fictionalized docudrama set in Tehran during the January 2026 protests, and they say it is the first fully AI-generated, feature-length film accepted at a major film festival.

Ash Koosha, CEO of Fountain 0, told Rotten Tomatoes that he does not like calling it an “AI film,” because the decision-making is deliberate and the goal is to keep the emotional center on the people affected. He frames the risk as both artistic and historical: he wants voices of victims to be heard “again and again,” and he says the film might not go anywhere, but if it does, it could be studied decades from now as one of the first AI films in history. That is the thesis hidden inside the headline-grabbing production budget: the creators are trying to win the authorship fight by arguing they controlled intent, not output.

So what does “fully AI-generated” actually mean here, in practical terms? Koosha draws a hard boundary. He says the only AI element that is fully generated is the visuals: the visuals, video, and the world-building components. He also explains why. Locations and worlds like those are “impossible to recreate” in traditional production terms, and he points out the difficulty of shooting in Iran to document events without being arrested or prosecuted. In other words, the AI is filling a constraint that traditional production would struggle to satisfy.

Everything else follows a more traditional filmmaking pipeline. Koosha says the film began with a script, with a screenplay, and that he “blocked it like you would a traditional film.” He describes work done before any AI button is pressed, including deciding timeline, decoupage and mise-en-scene for each block and sequence. He also lists the post-production craft that comes after: music, soundtrack, sound mix, editing, pacing, correction, and coloring. In his telling, AI generation is inserted as a component, not a replacement for the end-to-end production discipline.

When Koosha gets specific about the AI phase, he describes it as preparation first, generation second. He says pre-video preparation is the most important and most artistic step, because someone still has to imagine the world, the people in it, how they look at each other, how objects are designed and placed, and how motion happens. The way he contrasts the process matters: he says the intent is collaboration with AI rather than letting AI decide scene physics or scene logic. He compares it to cinema roles and logic control, arguing that in traditional filmmaking, logic often sits with the director and cinematographer, shaped through physical production and the right take.

In this digital approach, he says the model becomes the camera, becomes the world, and becomes the performance. That raises a real-world management question for executives and boards: if the bottleneck shifts from casting and set logistics to modeling workflows and “digital preparation,” what new roles need to exist? Koosha explicitly points toward inventing new positions for gaffers, lighting, and sound people to “operate these models correctly,” especially because he claims you need on-set or edit-room experience to do it well. He also argues that political advertising and headlines have scared people away from AI, and that scared folks are not being given the models and training needed to operate them.

If that sounds like culture-war framing, it still ties back to a concrete industry debate: authorship, labor, and training data. Koosha says the criticisms and skepticism are “valid,” and he says the reason they made the project was to find answers through doing the traditional route of filmmaking, rather than just posting social media content. He frames the project as a way to locate “errors” and “opportunities” that might impact people positively. That is a careful move, because these debates are not abstract for studios or investors. They affect who gets hired, what gets protected, what counts as a fair creative contribution, and what festivals are willing to program.

There is also an implicit regulatory and risk-management backdrop, even though Koosha does not cite regulators by name in the provided text. He highlights the practical danger of shooting in Iran and the reality that some content and documentation is constrained by law enforcement and prosecution risk. That matters second-order for decision-makers in media and tech: when AI can approximate inaccessible visuals, it can reduce certain physical production risks, but it can also increase reputational and ethical risks tied to authenticity, attribution, and data sourcing. The project’s acceptance at Tribeca, the “first fully AI-generated, feature-length film” claim, and the creators’ insistence on intent are all signals aimed at those gatekeepers.

Ultimately, Dreams of Violets is being used as a stress test for whether AI can be folded into filmmaking without erasing the creative chain. Koosha’s argument is that the director and cinematographer logic does not vanish; it moves earlier into studio building, and the on-set roles must be reimagined for model operation. For executives evaluating AI strategies, this becomes a decision framework: do you treat AI as a shortcut that competes away labor, or as a new production system that demands new craft, new tooling expertise, and new validation pathways? Fountain 0 is betting on the second model, and they have one more movie planned, with Fountain 0's second film “set to be announced at the end of the month.”