What is an AI film?
- Alex Iwanoff
- 12 minutes ago
- 5 min read
It’s been nearly four years since AI first landed on our computers with ChatGPT, closely followed by image, sound and video generators, effectively ‘killing cinema’ and everyone’s career with it, as now you can generate a script, actors, lighting, VFX and all. Or so the internet said.
But in real life... how is it effectively changing productions today? And we’re not talking here about resuscitating actors from their graves (Val Kilmer being the first one to be fully generated for the movie As Deep as the Grave). No, we’re looking at main ways that together paint a much bigger picture of where production is actually going.
Let's go through them.
BITCOIN: KILLING SATOSHI — THE GRAY BOX AI FILM
Bitcoin: Killing Satoshi tells the story of the real-world legal saga around Craig Wright, the Australian computer scientist who publicly claimed to be Satoshi Nakamoto, the anonymous inventor of Bitcoin. Casey Affleck plays Wright. Gal Gadot, Pete Davidson and Isla Fisher co-star. The script is by Nick Shenk (The Mule, Gran Torino) and the director is Doug Liman (Edge of Tomorrow, Mr. & Mrs. Smith, The Bourne Identity).
Needless to say that this is no indie production.
The movie called for 200 locations across four continents (Antarctica, Antigua, Las Vegas, etc.). Shooting that conventionally would have cost $300 million, the producers at Acme AI & FX told TheWrap in an exclusive set visit. Instead, they converted a former car showroom in West London into what they're calling a grey box: grey walls with with X everywhere, after “blue and green tests produced subpar results”. Proxy set pieces were built for anything an actor needs to physically touch: a staircase, a motorcycle, a desk. More like a theatre piece than a movie set.
“The entire focus on the set was on our performances. It was much more like acting in a Broadway play than in the giant event film that Doug's final product will actually be”, Casey Affleck
Though many roles were kept, there was one team that was missing: the lighting crew. All environments and lighting will be added in post. On set, only basic utility lighting was used. The sets themselves, however, were designed the traditional way. As production designer Oliver Scholl told TheWrap, his team used renderings, 3D models and hand drawings, then fed that reference material into the AI. The environments were generated virtually before principal photography even began.
In total, 55 AI artists are working in the post-production of the movie and the producers declined to reveal which model or AI tool they were using.
HOUSE OF DAVID — THE HYBRID
The Biblical epic premiered on Amazon Prime Video in February 2025. The second season was released on March 2026. House of David is now the clearest documented example of what a working AI-integrated production pipeline looks like inside a conventional TV show. The series has drawn over 40 million global viewers, according to Amazon MGM Studios’ press release.
And here are other numbers, because they are the story.
In season 1, 72 shots were generated using AI tools, concentrated mainly in Episode 6 "Giants Awakened", a sequence depicting the origin of Goliath, that sparked debate online about the use of AI in visual storytelling.
Midjourney, Magnific, Topaz, Runway, Kling and many more, layered onto After Effects and Unreal Engine. The production rule is to use the tools for what they’re good and never generate from scratch. Always augment something they had actually photographed or built. In other words, AI as a skin, not a skeleton.
“They don’t bear the wonky hallmarks of generative AI output from years past, but it’s not hard to believe they were AI-generated”, WIRED
In Season 2, the AI shot count jumped to 400 shots and the production had expanded to 30 different tools (both traditional and AI), covering image, sound and video generation, the LA Times reported. The key technical leap, detailed in a VP-Land breakdown of the production, was style transfer: training the system on the show's actual photographic look and applying that aesthetic over AI-generated content, so the join disappears in the final cut.
“It’s not a comparison of what would “Moses” have cost otherwise. It’s a comparison of “Moses” would have never been made otherwise, and that’s the way you have to think about it”, Jon Erwin, series creator and co-showrunner to LA Times, when asked about using AI to replace background extras in crowd scenes.
HELL GRIND — THE FULLY GENERATED
And then there's this.
Hell Grind is an action-fantasy made with no camera. No set. No actors.
The film was made entirely on Higgsfield, a San Francisco-based AI video platform, by director Aitore Zholdaskali and a team of 15 people, including directors, DPs and editors. The craft was not about the set, but more about choosing and selecting.
But here are the numbers, because they matter here too.
According to Alex Mashrabov, CEO and co-founder of Higgsfield, the 95-minute feature cost under $500,000, the majority in compute costs. The first 25-minute segment required 16,181 video generations to produce 253 final shots, a 64:1 curation ratio. Mashrabov told Screen Daily a comparable traditional action-fantasy feature would cost around $50 million, but he's candid about the limitations.
"There is definitely a feeling of a slot machine in complex scenes", he said, noting that some shots still require hundreds of iterations before spatial logic, character placement and action work correctly.
The script, however, remained human-written. A deliberate choice that points to something the numbers don't capture: the story is still the thing. What changed is the access. Epic-scale filmmaking was, until very recently, a privilege reserved for studios with deep pockets. Hell Grind suggests that's no longer true.
“Some people might use AI to even write the scripts. But I think we’ll see the difference between AI-generated scripts and human-crafted scripts. Then the audience will decide which one they prefer”, Dinara Mamleyeva, Higgsfield's communications lead
The movie premiered in May 21 at Cinema Olympia, a market and sidebar venue during the Cannes festival.
SO, WHAT ARE WE LOOKING AT?
Three productions. One with a $70M budget and a Hollywood director. One inside a full-scale TV production at Amazon. One made by a tiny team with no camera and big ambitions. The word "AI film" is being used to describe all three.
They don't have much in common technically. Killing Satoshi generated environments around real performances. House of David generated specific shots inside a traditional production pipeline (and backgrounds too). Hell Grind generated everything.
But they share one thing: cost was the problem AI was hired to solve. Making a film is expensive and that cost has always been the wall between a story and its audience. The reason countless ideas never leave the page. What these three productions show, each in completely different ways, is that the wall is getting lower. Now, you only need a good story.
What do you think about it? What's your take? Drop it in the comments.
Comments