The First AI Film Festival Paid Out $450,000 and Taught a Lesson About Story

The Astana AI Film Festival handed out its first awards on October 3, and the top prize went to a director who had been making films for four years and using AI for two. Nikolay Shestak, from Latvia, won $450,000 for *The First Honest Day*, a short about people whose faces change shape according to how they behave. Bad acts make them ugly. Good acts make them beautiful.

That premise could have been written by a human without any help from a model. It probably was. The film won because of how it handled the idea, not because of what generated it.

What the numbers looked like

The organizers said they received 8,067 submissions from 125 countries. Twenty-five made the final cut. The entries together used more than 80 different generative models, which is a useful reminder of how fragmented the tooling still is. Two Chinese films reached the finals, and one of them took home an award: *The Last Span*, from a team called 2xLabs, won Best Story. The film follows a homeless story along railway tracks through a father's monologue.

India finished second and Spain third. The prize pool was large by any festival standard, with $450,000 going to the winner alone. Finalists began screening publicly on October 2, and the jury included Audrey Azoulay, the former UNESCO director-general and French culture minister, along with Ma Ping of the China Film AI Research Institute.

The case the festival is making

Almas Zhali, a co-founder and council representative, said the most important part of an AI film is still the idea, because the tools have removed the resource barrier that used to decide who got to make one. He then made a point that sponsors of AI film festivals rarely make: AI films will hold a place in the market, but they will not replace traditional cinema outright. As long as directors like Christopher Nolan keep making the kind of film he makes, the older form does not disappear.

That is a more honest framing than the usual. The interesting question at a festival like this is not whether AI can produce a watchable minute. It plainly can. The question is whether it can hold a story together for ten or twenty of them without the seams showing.

What the screening actually revealed

Watching roughly ten of the finalists, a pattern showed up. The strongest entries used the technology to serve a narrative idea that already existed. The weaker ones leaned on spectacle and then ran out of road. Several films reproduced traditional cinematic grammar well, with wide shots and effects work that would have cost a production real money a few years ago. The ones that stayed in memory did something smaller.

*The First Honest Day* drew applause in the room. It did not try to out-scale anyone. It took one rule, applied it consistently, and let the audience work out what it meant. That is a writing decision. A model cannot make it for you, and no amount of resolution or frame rate substitutes for it.

Several other films were described in coverage as fast-food storytelling: strong openings, rising action, and then an ending that arrives because the runtime is up rather than because the story earned it. That failure mode is not unique to AI film. It is the most common failure mode in film.

Why the tooling matters less than it sounds

The 80-plus models figure sounds impressive until you remember what it measures. It measures how many tools the field is currently spread across, not how good any single one is. A director today can pick between text-to-video models with different strengths in motion, lip sync, and length, and will often use more than one on a single shot. The craft has moved toward supervision: choosing the take, fixing the continuity, and deciding what to keep.

That shift has a cost. Every model has its own quirks, and consistency across cuts is still the hard part. A character who looks right in one shot can drift in the next. Festivals like this one are, among other things, a public record of how well the field is closing that gap.

The money question

A $450,000 first prize invites a comparison. That is more than many national film funds hand out for a feature. Putting it behind AI work signals that the organizers see a viable category forming rather than a novelty act. Whether the money produces better films or just more submissions is something the second edition will answer.

For now the first edition left a fairly plain lesson. The technology has reached the point where it can carry a coherent short. What separates the winners from the also-rans is the same thing that always separated them, which is whether someone had something to say.

The cost side of an AI short

The festival also indirectly makes an argument about budgets. A short film that uses several generated models can, in principle, be made by a small team without a crew, a location or a camera. That is what Zhali meant when he said the tools have removed the resource barrier. For a director in a country with a thin film funding system, that changes who gets to make something at all.

It does not make film free, and the entries show why. Generated footage still needs selection, assembly, sound design and colour work, and those steps take people who know what they are doing. The savings land on production rather than on post. A team that skips the post work produces something that looks generated, and audiences notice.

What separates the entries

Across the finals, a rough hierarchy showed up. At the bottom were films that treated the models as the point, arranging impressive shots with no through-line. Above those were films that borrowed the language of conventional cinema and executed it well, which is harder than it sounds. At the top were films where the technology was almost invisible, because the story carried the attention and the generation stayed out of the way.

That hierarchy is not specific to AI. It is how short film competitions have always sorted entries. What is new is that the bottom category is now cheap to produce, so the volume of it is higher, and the judges have to wade through more of it.

What to watch next

Two things are worth tracking. The first is whether the winning films find distribution beyond the festival circuit, since a prize is not an audience. The second is whether the next edition sees fewer one-rule concept films and more that use the tools to do something the older production pipeline could not have attempted at that budget.

Zhali's comparison was to filmmakers who keep working at the top of the traditional craft. The fair test for AI film is not whether it can beat them. It is whether it can build a body of work that stands on its own without the technology as the headline.