Late Shift: the interactive crime film shown in cinemas

Late Shift is a film performed by live actors in which the viewer constantly decides what the protagonist does. Matt Thompson is a student working as a parking attendant in London: a crew steals a BMW i8 from the garage, takes Matt with it, and — after one robber breaks his arm — forces him into the heist of an auction house: a Chinese porcelain rice bowl of the Tchoi family, auctioned by the steel magnate Samuel Parr. From there the plot forks — abduction, chase, betrayal — but the setup is always that one.

The CtrlMovie technology comes from the studio of the same name; the director and co-writer is Tobias Weber (co-written with Michael R. Johnson), the publisher is Wales Interactive, the engine is Unity. Joe Sowerbutts plays Matt, Haruka Abe plays May-Ling. It is not a filmed visual novel and not cutscenes between gameplay: the whole picture is a filmed performance, and the choice is part of the edit.

A decision appears on screen about once a minute, and in a cinema you get roughly three seconds to tap your choice on the phone. The film never stops: a server collects the hall’s votes and immediately sends the next clip to the projection system. The home version keeps the same rule — the timer runs, nothing pauses, and even not choosing is a choice: the CtrlMovie tagline says “No decision is still an action.”

Why that matters for directing

The missing pause button is not a gimmick; it is the frame. First, it removes the branched film’s worst habit — the freeze-frame menu that reminds you every time that you are inside a program. Second, it dictates the economics of the shoot: every branch must rejoin the shared flow within seconds, so the script stands on rejoin nodes and waiting loops, not on a tree that scatters. Third, it changes the actor’s job: the performer plays a continuous scene, not a “stop before the choice”. This is the same film-or-game line I unpack in the dedicated essay — here the choice mechanism is deliberately made cinematic.

In a cinema, the audience installed an iOS/Android app and joined the hall’s special Wi-Fi network. A decision appeared on the phone with about three seconds to tap; a computer in the hall collected the answers and cut the next clip into the projector. The phone had to stay unlocked in your hand for the whole show — a detail that tells you how alive the format is. The UK release ran at Vue cinemas in March 2017, after the film had worked its way around festivals in Europe and North America.

The screenings also travelled outside regular distribution. On 16 October 2017 the film was shown at Paramount Pictures to roughly a hundred executives, film students, and interns: per Deadline, Paramount’s senior VP for new media Tom Hayes polled the room on paying a small up-charge for an interactive version of, say, Paranormal Activity — “virtually every hand shot up”. Producer Kino Industries, through its CtrlMovie unit, was shopping the “movies by film logic” further into the studios. The tech team admitted one number: 91 percent of the hall voted to kill the old man. An audience acting as a single character is the Kinoautomat 1967 idea continued — with an app instead of voting cards.

CtrlMovie’s formula is “1 Story · Countless Storylines · 7 Different Endings”: from one beginning the paths fork and rejoin, concluding in seven distinct endings. The film contains about 180 decision points, of which only about a dozen are forks that change the course; the rest are colour, tone, micro-choices. Deadline describes the shape better than I can: early on the branches stay narrow, and only in the final 40 percent of the runtime do they widen like a delta.

I am not writing a walkthrough of every ending — that is a wiki’s job, not a director’s reading. The shape itself is the point: narrowing early and widening late means the viewer carries the illusion of one continuous film for a long time, and the real consequences arrive only when there is something for them to collide with. That geometry — a nodal structure that keeps rejoining — is exactly what the branching-structures guide describes.

Late Shift has what the first interactive film lacked. Kinoautomat (1967) gave the hall a vote but snapped back to a pre-written “canonical” course after every decision — a joke about how illusory the audience’s power is. Late Shift puts the vote on a server that edits the film in real time: the hall genuinely sees the consequence of its choice within seconds. And Bandersnatch (2018) carried the same mechanism home, onto a streamer’s remote — after Netflix closed that catalogue, games like Late Shift remain the place where the format is alive.

That is why, for me, Late Shift is the boundary case on the film-game line. It is made like a film: a shoot, actors, an edit rhythm. It works like a system: a timer, a vote, seven endings. The experience is neither “a film with breaks” nor “a game with video”, but one continuous screening with a decision mounted inside it. A detailed analysis of this hybrid is in my article on the genre specificity of interactive cinema.

My criterion is the experience, not the file format: can it be lost? In Late Shift it can: a wrong tap takes Matt to an arrest, a bullet, or away from the ending you wanted. But the skill the film asks for is fast reactions inside staged circumstances, not mastery. That places it closer to the branching film than to the adventure game — the full criterion is in Interactive film or video game.

The home version arrived on 18 April 2017 for Windows, macOS, PlayStation 4 and Xbox One, and on 26 April 2018 for Nintendo Switch; the Steam page is still live. A standalone iOS/Android app, per CtrlMovie, lets you take every decision yourself — that is the version to recommend to anyone unhappy with obeying the hall. Check languages and region on the store page before buying.