Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesZone Out is a real, roughly five-to-six-minute science-fiction experiment made for a 48-hour challenge in 2018. Its software system, Benjamin, generated dialogue, music and manipulated performances, but the film was not made without people: Oscar Sharp, Ross Goodwin, actors, editors and technical crews built, supervised and completed the production.
Watch the film through the creators’ Therefore Films project page, or read Ars Technica’s presentation and transcript. WIRED’s separate behind-the-scenes video shows Thomas Middleditch and the production team discussing the experiment.
What you are watching
Zone Out is intentionally hard to follow. It combines science-fiction imagery, abrupt cuts, synthetic or heavily processed voices, warped faces and dialogue full of non sequiturs. The result can feel bizarre, funny, unsettling and occasionally affecting, but it does not offer a conventional plot that can be summarized scene by scene.
That confusion is part of the point. Zone Out is as much a demonstration of an automated filmmaking pipeline as it is a narrative short. Its roughness exposes what happens when software is asked to perform jobs normally coordinated by a director, writer, editor and post-production team.
Free tools Windows power users keep installed
One-click scans. No signup required.
#1 Best Overall
Who was Benjamin?
Benjamin was the name Oscar Sharp and AI researcher and creative technologist Ross Goodwin gave to their filmmaking system. It was not a conscious or autonomous person. It was a software pipeline assembled from machine-learning and media-processing techniques, running with human-selected inputs and infrastructure.
WIRED reported that Goodwin monitored the system from a laptop while it ran in the cloud, using AWS, TensorFlow and several generative adversarial networks for face-swapping work. The report mentions 11 GANs in that process; this is a description of the 2018 production, not a general definition of how face replacement works today. Read the technical account at WIRED.
What the AI actually did
The contemporary headline called the film “directed entirely” by AI because the filmmakers tried to delegate an unusually broad set of creative decisions to Benjamin. The phrase is shorthand, not a claim that humans disappeared from the production.
| Filmmaking task | Benjamin’s contribution | Human role that remained |
|---|---|---|
| Screenwriting | Generated the screenplay and dialogue. | Sharp and Goodwin designed the system, supplied the challenge constraints and used its output. |
| Performances | Generated or processed voices and transformed recorded actor material. | Actors Thomas Middleditch, Elisabeth Gray and Humphrey Ker supplied human performances and facial data. |
| Visual material | Helped determine how source footage and character assignments would be used. | The filmmakers selected, prepared and provided the footage and green-screen recordings. |
| Face replacement | Mapped actors’ faces onto characters in existing footage. | People recorded the source performances and operated the technical workflow. |
| Editing | The project attempted to automate footage selection and scene assembly. | Editors intervened when the automated selection process failed, then interpreted the generated screenplay. |
| Music | Generated the score attributed to Benjamin. | Human technicians integrated the audio into the finished short. |
Scroll’s account presents the strongest version of the “written, performed, scored, directed and edited” claim. Ars Technica’s reporting supplies the important qualification: humans still had to make editorial decisions.
How the footage was constructed
Zone Out was not generated from nothing by a modern text-to-video model. Sharp and Goodwin combined pre-existing public-domain film material—including reported footage from The Brain That Wouldn’t Die and The Last Man on Earth—with green-screen recordings and facial data from the actors.
Face-swapping tools then placed the actors’ faces onto characters in the older footage. Synthetic or processed voices, generated dialogue and an AI-attributed score completed the hybrid production. Luke McKernan’s account documents the source-footage lineage, while WIRED describes the actor and technical process.
Why the faces look broken
The flickering faces, distorted mouths, misplaced facial hair and unstable features reflect the limits of 2018 face-swapping technology. The production also had only 48 hours, leaving little time to correct artifacts, improve synthetic speech or rebuild scenes.
Those defects are not simply evidence that the concept failed. They make the boundary between a recorded performance and a digitally manufactured one visible. Middleditch’s face can appear to slide, deform or lose alignment with the body because the software was transforming footage rather than creating a clean, unified performance.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Rank #3
The 48-hour constraint changed the result
The short was made for Sci-Fi-London’s 48-Hour Film Challenge, which supplied requirements or prompts and allowed only two days for production. That deadline explains several features viewers might otherwise mistake for deliberate artistic choices:
- Face swaps remained visibly experimental.
- Some voice-generation work was incomplete.
- The team had little opportunity to restructure the story.
- The footage-selection system could not reliably find suitable shots.
- Editors had to make practical decisions under time pressure.
Where the automated editing broke down
The most revealing limitation concerned shot selection. Sharp told Ars Technica that the convolutional neural network did not have enough useful object descriptions in the screenplay, and the available films did not contain enough matching objects. Benjamin therefore could not reliably select all the footage needed to realize its script.
Sharp and editor Jono Chanin worked from the premise that Benjamin’s choices represented an artistic direction, while retaining the wording of its screenplay. This is the clearest reason the word “entirely” needs qualification: the experiment attempted full automation, but human judgment was required when the pipeline failed.
Who made the experiment?
- Oscar Sharp: the Los Angeles-based director who developed the filmmaking experiment with Goodwin.
- Ross Goodwin: the AI researcher and creative technologist behind Benjamin’s technical system.
- Thomas Middleditch: the actor whose face and performance material were digitally transformed.
- Elisabeth Gray and Humphrey Ker: actors who also contributed performance material.
- Jono Chanin: the editor involved in interpreting and assembling the final footage.
The project’s creator-linked page is Therefore Films’ Benjamin project page. WIRED’s making-of video provides additional production context.
How “directed by AI” should be understood
In ordinary filmmaking, directing includes interpreting a script, guiding performances, choosing shots, controlling pacing and making final creative decisions. Benjamin was designed to perform software analogues of some of those tasks. The filmmakers deliberately treated its outputs as decisions to be followed rather than merely suggestions.
That does not mean Benjamin understood the story, possessed intentions or operated without supervision. Humans created the code and workflow, chose the data and source films, supplied the actors, monitored the system and intervened when automation failed. A precise description is that Zone Out was an AI-led or AI-generated filmmaking experiment built and completed by people.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Where it fits in the Benjamin experiments
Zone Out followed earlier collaborations between Sharp and Goodwin:
- Sunspring (2016): Benjamin generated a screenplay that human actors performed.
- It’s No Game: a later experiment expanded the project’s use of Benjamin and included actor David Hasselhoff.
- Zone Out: the team attempted to extend machine involvement across writing, performance manipulation, visual assembly, editing and music.
The sequence matters because Zone Out was not an isolated stunt. It was a progressively broader test of how much of a film’s creative pipeline could be delegated to software. The project page for these films is hosted by Therefore Films.
Recommended Free Tools
Best Value
How successful is it?
As a conventional film
The story is difficult to parse, performances are frequently distorted, and the pacing is shaped by technical compromise. Viewers looking for a coherent science-fiction narrative may find it frustrating. Its melodramatic score, robotic delivery and visual glitches are more memorable than its plot.
As a technical experiment
Its value is clearer. The short demonstrates, in one constrained production, machine-generated writing, synthetic performance, face replacement, algorithmic music and attempted automated editing. It also records exactly where those systems needed human assistance.
WIRED reported that the film did not place in that year’s challenge. That result says little about whether the experiment was worthwhile: a competition judges a finished film, while Zone Out was also testing authorship, control and the limits of automation.
Why the film still matters
Zone Out is an early, visibly imperfect example of questions that later became central to AI media: who owns a synthetic performance, where an actor’s contribution ends, whether a machine can be called an author, and how audiences distinguish recorded reality from manipulated imagery.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →It should not be treated as proof that software can independently make a feature film. The project was short, deadline-driven, dependent on existing footage and built around a prepared pipeline. Nor should it be reduced to “just a deepfake”: face replacement was only one component alongside generated writing, voices, music and semi-automated assembly.
The verdict
Zone Out is more successful as a historical technology demonstration and cultural curiosity than as a polished science-fiction movie. Its headline is defensible only if “directed by AI” means that the filmmakers tried to let Benjamin make decisions normally associated with directing. The finished work still depended on human design, performances, infrastructure, supervision and editing.
That tension is precisely what makes the short worth watching. The bizarre dialogue and damaged faces are not distractions from the experiment; they show, frame by frame, what creative control looked like when an early machine-learning system was given an ambitious job and two days to do it.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
