June 2026

Where Are the AI Films?
Some Observations from Inside a Festival Selection

By Enrico Vannucci

Within the festival ecosystem, cinema made with AI tools remains a rare presence. Among the submissions I watched this year — almost 1400 short films — four were made entirely with AI tools, and two contained AI-generated inserts.

Most of the cinema being made with AI tools today does not reach the festival circuit at all. It lives online, on Vimeo, YouTube, and social platforms. Some of it even circulates on LinkedIn, which says something about its purpose: less a film made for an audience, more a demonstration of technical skill made for a professional network.

This raises a question that is rarely addressed directly: how does one evaluate a film made with AI tools? From a curatorial perspective, the answer may be simpler than expected. The criteria do not change. What matters is whether the work functions as cinema, whether it has something to say and whether it finds a form adequate to saying it. The tool is irrelevant to that judgment. A film made with AI can be formally accomplished or formally incoherent, narratively compelling or empty, visually interesting or merely spectacular. The difficulty is that the sheer volume of AI-generated content circulating online, much of it designed for spectacle rather than meaning, makes it harder to approach these works with an open eye. The association between AI and superficiality is not entirely unfounded, but it is not the whole picture either.

The works that did arrive varied considerably in approach. The films made entirely with AI tools tended toward animation or attempted to replicate photographic cinema, with mixed results. One used AI to reconstruct documentary material, which raised its own set of questions about representation and authenticity. The more formally interesting uses of AI appeared as inserts within otherwise conventionally produced works, where generated imagery served a specific expressive function rather than constituting the film itself. The experimental tradition, filmmakers who deliberately foreground AI's imperfections as an aesthetic strategy, was largely absent from the submissions I watched. Those works exist, and some are genuinely compelling. Whether their absence reflects the specific nature of this selection, or a broader pattern in how they circulate, is difficult to say.

This is not the first time a new tool has lowered the barriers of entry to filmmaking. The arrival of MiniDV in the late 1990s made production accessible to a generation of filmmakers who could not have afforded film stock. The result was an explosion in output and, eventually, a handful of works that demonstrated what the format could genuinely do. AI tools may be following a similar trajectory. The cost of certain images, effects, environments, visual experiments that once required significant budgets and large crews, is dropping. What remains constant, as always, is the need for a human mind behind the work. Democratisation of the tool does not automatically produce democratisation of quality.

At the same time, the industry is moving in the opposite direction. AI-assisted productions are beginning to appear at major market events, not yet in competition, but present, visible, looking for buyers and for legitimacy. The festival ecosystem, with its economy of prestige, is something the industry wants access to. One recent case is instructive, though not in the way its producers intended. An AI-assisted animated feature presented at the Cannes market this year, built across a generative pipeline that included a major text-to-video model, did not make it into the festival programme, in part because the model it relied on was shut down mid-production by the company that owned it.¹ The film made it to the Croisette. It did not make it onto the screen. The case illustrates something that goes beyond any single project: the tools on which these productions depend are not stable infrastructure. They are commercial products, subject to the economics and strategic decisions of the companies that build them.

Film festivals have responded to this moment primarily through critical debate, conferences, academic programs, theoretical frameworks. This is not nothing. But it is not yet a structural response. When virtual reality emerged as a serious audiovisual form, festivals eventually created dedicated sections and exhibition spaces to accommodate it. VR did not replace cinema, it established itself as a parallel practice, with its own works, its own audiences, its own logic. AI cinema has not yet reached that point, in part because the films themselves are still few. The two worlds, the festival ecosystem and the emerging practice of filmmaking with AI tools, are not yet in contact in any meaningful way. How and when that changes remains to be seen.

¹ Critterz, an AI-assisted animated feature from Native Foreign, Vertigo Films, and AGC Studios, was presented at the 2025 Cannes market. The film had been built using OpenAI's Sora model, which was shut down in March 2026. See: thenextweb.com/news/critterz-misses-cannes-openai-sora-shutdown