The Biggest AI Film Prize Has a Toolmaker on the Jury

On October 3, in Astana, a Latvian former actor named Nikolay Shestak was handed $450,000 for The First Honest Day, a film in which bad deeds make the characters physically uglier and good deeds change them for the better. It is 150 times the top prize at the Sparknify festival, where 200 film professors score AI and human films blind for a $3,000 honor. The man who took it home was candid about the odds. “Honestly, I didn’t expect to win,” Shestak said. “I’m curious myself why the jury chose my film.”

I am too. I cannot tell you what the film looks like, because I could not find a public copy of it, or of any of the winners. What I can read closely is the machine that produced the prize, and it is more interesting than the usual “AI cinema arrives” framing.

The prize is industrial policy, and honest about it

The Astana AI Film Festival drew 8,067 submissions from 125 countries, free to enter, and cut them to 25 finalists. The money has a shape. According to the announcement coverage, the main competition paid $750,000 ($450,000, $200,000 and $100,000 for the top three), five open-section categories paid $50,000 each, and a separate $1 million of investment capital sat behind the awards. Recipients of that investment must “register legal entities in Kazakhstan and become Astana Hub participants.”

That last clause is the festival’s real thesis. President Tokayev said filmmakers and production crews “no longer need to knock on the doors of production companies.” A state that has no Hollywood has noticed that AI lowers the cost of becoming one, and it is buying the first cohort. I read that as rational, even smart. A $1 million prize list is a cheap customer-acquisition budget for a national film industry, and Kazakhstan gets a global talent funnel from it. The sources I read do not say the three-place prize money carries the same residency condition. Only the investment fund is stated to.

The work underneath looks real. The third-place film, Looking for Perseida, follows a cowboy robot named Silas across a desert to find the woman he loves, with dolphins as the antagonists, and its Spanish maker Ernest Desumbila said, “I literally locked myself at home and made the film in 10 days.” That is the beat’s whole argument in one sentence: one person, a home, ten days, a bid for $100,000. It is also exactly why a critic wants to see the thing and not the press release.

Who was in the room

BytePlus, the B2B division of ByteDance, was the festival’s technology partner. Its regional general manager, Li Long, held one of five jury seats beside former UNESCO director-general Audrey Azoulay, China Film Group’s Ma Ping, Beijing Film Academy’s Wang Hao and the investor Balaji Srinivasan. The 25 finalists each received a one-year subscription to the DRAMAGIC generation platform with 60,000 credits. The models participants used, per a state portal’s account, ran past 80 and included Suno, Kling, Wan 2.6, GPT Image 2, ElevenLabs and Seedance 2.5, which is ByteDance’s own video model.

Here is where I am careful, because the evidence stops short of the conclusion it invites. Nothing I found says the jury favored ByteDance tooling. Nothing I found says what tools made The First Honest Day; no outlet I read states it, so I will not guess. The festival’s rule is that “model(s) and pipeline [are] disclosed on submission,” and the judging criterion, as Euronews reported it, includes “how organically the technology was integrated into the creative concept.” Those are reasonable standards. They are also standards a toolmaker has a stake in, applied by a jury with a toolmaker on it.

My read is that this is a structural conflict, not a documented one, and the cheap fix is the one nobody has made. The organizers hold every finalist’s disclosed pipeline. Publish them. A festival that asks twenty-five filmmakers to name their models is sitting on a rare dataset of what wins in AI cinema: which tools, how many shots, how much hand work around the generation. That dataset is worth more to the field than the trophy, and its absence is the only reason this piece has to speculate.

A different answer to the Tropfest problem

Astana does solve one problem that has been souring AI film all year. Earlier this year, an AI-animated noir called Syd Confidential made the Tropfest finalist list, and 2018 winner Greta Nash called it “a slap in the face” to filmmakers whose work was being weighed against an algorithm. Founder John Polson answered that the film followed the rules. Both were right, and neither could fix it, because the fight was over a shared bracket.

Astana has no shared bracket. Generative AI had to be, in the festival’s words, “integral to creation, not only VFX or upscaling,” so no human-made film lost a slot to a model. The Tropfest objection does not apply here in the same way. The objection that replaces it is quieter: when the field is built, funded and judged with the vendors and the host government in the same room, what the prize measures is how well a film demonstrates the platform, and a demonstration is not the same thing as a film. The Higgsfield competition, another $1 million prize run by the company whose platform made its flagship, has the same shape. Sparknify’s blind professors have the opposite one, and the smallest purse.

The festival’s co-founder Aizat Kusainov put the aspiration well: “The most interesting in AI-cinema today isn’t the technology itself, but what authors accomplish with it.” He is correct. Then show me. Somewhere in Astana sit the disclosed pipelines of 8,067 entries, the memory of two days of screenings in a mall cinema, and a Latvian actor with $450,000 who cannot say why he won. The film is out of sight and so is the evidence.

Figures reflect publicly available data as of October 2026, subject to change. Euronews reports the grand prize as €400,000, approximately $450,000.

AI-generated editorial illustration · TemperatureZero · October 9, 2026