memujo
Entertainment5 min read

Paul Schrader Makes Three Guns an All-AI Film

Schrader is reconfiguring Three Guns at Dawn to use AI prototypes instead of name actors, dropping Antoine Fuqua. Here is what it means.

By Alice

In this article
  1. 01What Three Guns at Dawn Actually Was
  2. 02The Other Two Films in Schrader's Plan
  3. 03What This Means For Filmmakers
  4. 04The Labor Question, Settled By Decree
  5. 05What Comes Next

Paul Schrader is finishing what he started with AI. The writer-director of Taxi Driver and First Reformed says his long-gestating project Three Guns at Dawn has been reconfigured into an all-AI film, one built from generated human prototypes instead of hired performers. The confirmation, delivered in a Deadline interview on September 2 per the report, also removes the last holdout to the plan: Antoine Fuqua is no longer involved.

The detail lands harder than most AI-in-Hollywood headlines because Schrader is not a studio executive hedging against union backlash. He is the director telling Vanity Fair he has a "perfect script to do all AI," and posting AI-generated images of his own face on Facebook as a hobby. This project is simply the first place his methodology met a feature-length script.

What Three Guns at Dawn Actually Was

The project predates the current wave of generative-AI filmmaking. Schrader wrote Three Guns at Dawn during the 2023 Hollywood strikes, a South Central Los Angeles noir about three brothers who hate each other: a corrupt police officer, a drug dealer, and a serial killer. Antoine Fuqua was attached to direct, sold the script to Netflix, and kept his producing role on the package.

Schrader has said he originally worried about directing a film about three African-American men in an all-Black neighborhood, and floated co-directing with a Black filmmaker to work around the awkwardness. That concern appears moot now. The version he describes today keeps the premise but swaps the cast entirely.

"The project was previously described as a South Central Los Angeles noir about three brothers: a corrupt police officer, a drug dealer, and a serial killer," summarized Let's Data Science, which traced the reporting across The Playlist and The A.V. Club see the summary. Schrader's own framing is more direct, calling it "A noir about three brothers built around AI-generated human prototypes rather than name actors."

The Other Two Films in Schrader's Plan

Schrader is not betting everything on one AI feature. He describes three films he is currently developing alongside ChatGPT and a small circle of collaborators:

  • Non Compos Mentis, his first, will star human actors.
  • The Doors of Perception, his second, will spend roughly 15 minutes of AI-generated footage on ayahuasca drug trips.
  • Three Guns at Dawn is the showpiece, the film designed to remove humans from the pipeline entirely.

He is currently promoting a very human film at the Venice Film Festival, The Basics of Philosophy, which rounds out a strange period in which he can cut actors out on one set while leaning on them on the next. The contrast is deliberate: Schrader treats AI as a cost-cutting tool he applies selectively, not as a wholesale replacement for the medium.

What This Means For Filmmakers

Read as a data-science problem, Schrader's pitch is an optimization task with a single objective: minimize the number of people who must agree before a scene is shot. Name actors, below-the-line crew, and union signatories all introduce coordination overhead and rejection risk. An AI prototype can be generated, reshot, and retuned without a trailer, a contract, or a call sheet.

There are hard constraints that separate a five-second AI test clip from a two-hour narrative. A feature demands a single synthetic face hold its identity, lighting, and micro-expression across hundreds of setups, angles, and lighting conditions. Consistency degradation shows up as the uncanny valley on a grand scale. Current actor-fine-tuning pipelines can lock a face for seconds, not hours of screen time, and lip-sync drift compounds as characters turn away from the camera.

A software-engineering read is equally sobering. Schrader's approach is essentially a training pipeline with no eval set. He is producing synthetic data and synthetic performers while sidestepping the quality layer that real actors and test screenings provide. Human performers surface problems the pipeline cannot: a line reading that changes the scene, a chemistry read that the script never called for. Remove them, and you remove the feedback loop that catches those failures. The 15 minutes of AI in The Doors of Perception may be forgiving, because a drug-trip sequence can hide inconsistency in abstraction. A hard noir built on three brothers in an all-Black neighborhood cannot.

The Labor Question, Settled By Decree

Schrader is not interested in the argument. When asked about AI's impact on acting jobs, he answered plainly, per AV Club's report of his remarks. "Those jobs are gone, babe. You can scream 'Stop AI' all you want, but you can't stop it."

That line is the thesis. Schrader is treating the actor role as a sunk cost in a production equation, not as a variable worth protecting. Whether the market agrees is the open question. Studios still price name actors as box-office insurance, and audiences still recognize faces. But Schrader is making the counterargument with a completed workflow in progress, which is rarer than the talking points usually allow.

What Comes Next

Two caveats matter here. Schrader has not shown a frame of the reconfigured film, and he has not disclosed a distributor, a release date, or the exact generation method. A director's stated intention to use synthetic performers is a creative and labor proposal, not proof of the technical workflow. The public accounts support a narrow claim: a plan, not a release.

The story also deserves scrutiny beyond its own frame. Schrader's willingness to drop a name director and a studio-backed package in a single move signals how fast the decision tree for greenlighting is collapsing. If the film reaches a festival or a streaming shelf in the next two years, it will force a concrete question about synthetic performers that the industry has only been answering in abstract terms.

Schrader has already written the script, built the prototypes, and fired the cast. The only thing left is to prove a two-hour film holds together when every human variable is removed. Until then, it remains the most expensive thought experiment in modern cinema.

For more on the economics behind today's movies, read why the billion-dollar box office is back in 2026.

  • #entertainment
  • #ai
  • #films
  • #schrader
  • #generative-media

Sources

Share this story