Skip to content
Deepak Gupta markFuture Tech

Future Tech/consumer

AI Generates Film to the Viewer by 2032

I watched ownership turn into access. Next the film itself stops being fixed. By 2032 AI generates and re-cuts movies to the viewer, and the idea of a single ownable cut quietly fades.

// By 2032 · medium confidence · disruption 6/10

Prediction

// 2032

By 2032 a major streaming platform ships AI-generated and re-cut films personalized to the individual viewer, ending the canonical fixed cut as the default unit of film.

Confidence██████████medium
Disruption██████████6/10

What dies

  • → the home film library

Who wins

  • → Netflix
  • → OpenAI Sora
  • → Google Veo

filed: 2026-06-14 · guptadeepak.com

The hook

The home film library died because you could not own the catalog. The next loss is stranger: soon there is no fixed film to own at all. If the movie re-cuts itself for you, with your preferred ending and maybe your face in a scene, what exactly would a copy even be a copy of?

Thesis. Beyond streaming, AI generates and re-cuts film to the individual viewer, so the catalog becomes infinite and bespoke and the canonical cut stops being the default unit of cinema.

The story

Setup: ownership already gave way to access

The paired obituary, the-home-film-library, traces how the owned shelf became a rented catalog. You stopped possessing films and started authenticating into a library that rotates on the platform's schedule. That was step one: the copy stopped being yours.

Step two removes the copy entirely. If the film is generated and re-cut per viewer, there is no single artifact to own, sell, or revoke, because no two viewings are the same object.

The hinge: personalization moves from the shelf to the frame

Streaming already personalizes everything around the film: the thumbnail, the trailer, the recommendation row. The hinge is when that personalization reaches inside the runtime. Alternate endings, a length tuned to your attention, scenes regenerated to your taste, and eventually your likeness inserted with consent.

Once a film is a generative system rather than a fixed render, the platform can serve a different cut to every household at near-zero marginal cost. The catalog stops being a list of titles and becomes a space of possible films.

Current state: the models can almost do it

By 2025 text-to-video models produced coherent clips approaching a minute, and editing models could change a scene's content, length, or character. [verify] Netflix's 2018 Bandersnatch proved viewers will accept a film that forks under their choices. Studios already quietly use generative tools for de-aging, dubbing into new languages, and patching reshoots.

The pieces exist as separate features. The prediction is that one platform assembles them into a shipping product where the film itself is generated to the viewer, not just selected for them.

Trajectory and the provenance problem

If this lands, the idea of a canonical cut fades for mainstream releases the way the owned shelf already faded. That breaks rights, credit, and provenance: who is the author of a film that exists in a million private versions, and how does anyone cite or preserve a cut that was only ever rendered once?

Expect provenance and watermarking standards to fight a losing rearguard action while the default mainstream experience drifts toward bespoke film. The fixed cut survives as a deliberate artifact, the way the 4K disc survives ownership.

Holdouts: the director's cut as a stance

Filmmakers and labels will defend the single authored cut as the point of the art, not a constraint on it. Criterion-style preservation, theatrical exclusivity, and director-locked editions become the holdout, chosen by people who want the film the maker intended.

That holdout is the same shape as physical media today: a minority ownership choice surrounded by an infinite rented, re-cut default. The canonical film does not vanish; it stops being what most people watch.

First signals (verify today)

Text-to-video models crossed minute-long coherent generation in 2024 to 2025: OpenAI's Sora, Google's Veo, and Runway all shipped publicly. [verify] Netflix has spoken openly about interactive and generative formats since its Black Mirror Bandersnatch branching experiment in 2018, and studios are already using AI for de-aging, dubbing, and reshoot patches. Streaming personalization already reshapes thumbnails and trailers per viewer; re-cutting the film itself is the same logic applied one layer deeper.

Key data points

  • OpenAI Sora, Google Veo, and Runway all shipped public text-to-video generation by 2024 to 2025. [verify]
  • 2024 to 2025 models reached roughly minute-long coherent video generation, up from a few seconds in 2023. [verify]
  • Netflix's Black Mirror Bandersnatch (2018) was a mainstream branching, viewer-directed film. [verify]
  • Studios already use generative AI for de-aging, dubbing, and reshoot patching in released films. [verify]
  • Streaming platforms personalize thumbnails and trailers per viewer today, a precursor to re-cutting the film itself.
  • Provenance standards such as C2PA content credentials exist but have weak adoption for entertainment media. [verify]
  • Confidence is medium: the technical trajectory is clear, but rights, talent contracts, and audience taste could delay a shipping product past 2032.

Contrarian angle

The owned shelf taught people that access replaced possession. Generative film goes further: it removes the object that could be possessed at all. You will no longer authenticate into a catalog of fixed films; you will authenticate into a system that renders a film for you on the spot. Ownership became access, and access is about to become generation, where the very thing you watched cannot be copied because it never existed as a stable copy to begin with.

The flip side

What this kills

The paired obituary in Tech Graveyard.

Read the obituary

FAQ

Won't people reject films with no fixed cut?

Some will, and that is the holdout: director-locked editions and physical media for viewers who want the intended film. The prediction is about the mainstream default, not the disappearance of the canonical cut entirely.

Is putting my face in a film realistic by 2032?

As an opt-in feature, plausibly yes. Likeness insertion already exists in research and consumer apps. The harder problems are consent, rights, and abuse prevention, which is why the medium confidence reflects policy risk more than technical risk.

How does this kill the home film library?

The owned shelf was already dying as ownership turned to access. Generative film removes the fixed artifact entirely, so there is no stable copy left to own, sell, or even revoke. It finishes the trajectory the obituary describes.

Who is positioned to ship this first?

Platforms with both distribution and model investment: Netflix, plus model makers OpenAI Sora, Google Veo, and Runway, and a content owner like Disney. The winner needs catalog rights and generative capacity together.

What about authorship and provenance?

It becomes genuinely hard. A film in a million private versions has no single citable cut, and preservation breaks. Expect provenance standards like content credentials to fight a rearguard action while bespoke film becomes the default.

More from guptadeepak.com

Want the technical deep-dive behind this prediction?

Read the companion article

Related predictions

More from the consumer desk.