A man walks down a street. A woman walks down a different street.
Cut together, they appear to be walking toward each other.
Cut to a wide shot of a white building with columns. Cut back to the two of them, arriving.
Cut together this way, the footage makes one claim: two people set out toward each other and met at a landmark.
Almost none of that happened. The man and the woman had been filmed on separate streets in Moscow, in footage that had nothing to do with each other. The building was not in Moscow. It was the White House, in Washington, thousands of miles away, spliced in from footage shot somewhere else entirely. No one walked toward anyone. No one arrived anywhere. A place had been built that never held all of its own pieces at once.
A Place Built From Separate Streets
This was the work of Lev Kuleshov, a young filmmaker running experiments at a Moscow film workshop. By the time he staged the meeting at the White House on purpose, he was testing something he had already discovered by accident, while cutting an earlier film called The Project of Engineer Prite (1918). He needed a shot of his actors looking at a row of electric poles. He didn’t have one. What he had was footage of the actors looking off toward nothing in particular, shot in one part of the city, and a shot of some poles, shot somewhere else. He cut them together. On the screen, the actors were simply looking at poles. The accident made the principle unusually visible: for a place to feel single and continuous on screen, its pieces did not need to have been true together.
Kuleshov’s workshop turned that accident into a method. Two performers, filmed on unconnected streets, walking. A cutaway to a monument neither of them had stood in front of. A cut back to the two of them, arrived. The method became known as creative geography, and the phrase is a fair description of what it had actually done: not recorded a place, but built one, out of parts that had never shared a city, let alone a corner of it.
Kuleshov is more often remembered for a different demonstration, the one usually taught first: a shot of an actor’s blank face, intercut with a bowl of soup, a child in a coffin, a woman on a divan, said to make the same unchanging face read as hunger, then grief, then desire, depending only on what sat next to it in the sequence. The footage is gone. What survives are Kuleshov’s own accounts, which changed over time, and a longer version written years later by his collaborator Vsevolod Pudovkin. Nobody today can check the popular version against the reel it claims to describe. What modern researchers have found, using real footage and careful controls, is smaller than the legend: a neutral face does drift toward the mood of whatever surrounds it, a measurable bias, not a transformation.
The White House material does not depend on any of that. Its claim is plainer and harder to soften: the sense that what you are watching is happening in one shared place can be manufactured entirely after the fact, out of footage that was never in that place to begin with. The camera recorded a man on one street and a woman on another, and a building in Washington. None of that footage recorded a meeting. The edit did. A shot holds what the camera recorded. An edit can hold something no shot ever held: a relationship between pieces of film that were never in the same place, the same day, or the same city.
The Shot That Never Arrives
An editor decides which pieces of footage stand next to each other. An editor can just as easily decide which piece never arrives at all, and let the gap do the work instead.
Jaws is built around a piece of footage that mostly does not arrive. The missing piece is a mechanical shark, and the whole picture waits for it.
Bad weather worked against the shooting schedule from outside. From inside, the mechanical sharks worked against it too. The pneumatic hoses that moved their jaws and fins took on salt water. Salt water corroded their skins and, in time, weakened their internal frames until they broke. The neoprene meant to keep them from absorbing water did not do that, and the props swelled with it over the course of the shoot. The sled that towed one of the models kept tangling in seaweed. None of this was a single dramatic breakdown. It was a slow, grinding unreliability that ate the schedule from the inside, day after day.
This is what awaited Verna Fields, the editor, once the material reached her. Spielberg later summarized the production simply. On a good day they shot five scenes. On an average day, three. On a bad day, none.
The film that resulted was not the film that had been planned. Spielberg has described the difference himself, more than once and in close to the same words each time: the picture had been conceived nearer to a monster movie, with a shark on frequent, visible display, and it became something closer to Hitchcock, a film that withholds. He has called the shark’s failure a godsend for that reason. That is a director’s later account of what the constraint produced. It is not proof that anyone chose withholding as a strategy from the first day of shooting. Constraint and craft are not the same event. The cutting room is one of the clearest places to watch craft respond to constraint.
Richard Zanuck, one of the film’s producers, gave Fields a specific credit that is easy to miss inside the larger legend of the broken shark. He said she reconstructed some scenes Spielberg had shot for comedy into scenes that played as frightening, and reconstructed others the other way, footage meant to frighten becoming, in the cut, comic. That is a narrower and stranger claim than “the shark broke, so the editor invented suspense out of nothing.” What a shot was made to feel like and what it ends up feeling like on screen are not always the same thing. Something happened to that footage between the camera and the screen. It happened in the edit.
Watch the film with this in mind and a pattern becomes visible. The shark is rarely given to you whole. What arrives instead is the water itself, disturbed in a way that is not quite explainable by wind or current. A point of view low against the surface, aimed at legs kicking above, with nothing in the frame to confirm whose point of view it is. A face, someone else’s face, changing before you understand why. None of these images are neutral filler standing in for a shot the production could not get. Each one is specific enough to organize your attention around a shape you never fully see. The absence has an outline, built by everything placed around it.
This is a different claim than “less is more,” and the difference is worth holding onto. An editor withholding an image at random would only produce confusion. What this film does instead is keep feeding you enough, water, a look, a consequence, to let you keep building the missing shot yourself, so that by the time you are finally shown something, your own imagination has already been recruited into making it frightening.
None of this erases the production trouble that made it necessary. A mechanical failure narrowed what the filmmakers had to work with, and that narrowing is real and documented. But narrowing the material is not the same act as deciding what to do with it. One of those things happened to a machine in the water. Another part of the response happened at a cutting table, where comedy could apparently be turned into dread, and where withholding an image turned out to be its own kind of decision, not an absence of one.
The Performance That Survives
In the seconds before Ki-taek picks up the knife in Parasite, his face does something specific. It is not a wide view of the whole room. It is close, and it changes, and the change is doing more work in that instant than anything else in the frame.
Yang Jin-mo, who edited the film, has said this was the hardest single moment for him to cut in the entire picture. Not the longest sequence to assemble, not the most complicated action to track. One face, at one moment, just before a killing. He has described being unusually deliberate about which expression, which piece of Song Kang-ho’s performance in that instant, would carry the audience through the turn.
That is a small, specific claim, and it is worth sitting with exactly as small as it is. It does not mean Yang invented the character’s turn, or that the performance in that scene belongs to the editing room rather than to the actor who gave it. It means that at the exact second where the film’s whole moral direction changes, more than one recorded version of that face existed, and a choice had to be made about which one the audience would actually see.
We tend to talk about “a performance” as if it were one continuous thing that happened in front of the camera and simply got recorded. What we are usually watching, even when it feels singular, is a performance that survived a choice. A scene may exist in more than one recorded version, an actor doing it again, differently, take after take. What ends up standing in that place in the finished film might be one of those takes whole, or it might be pieced together from more than one of them. Either way, what survives is a selection, not simply whatever happened to be recorded first. None of the versions that did not survive were wrong or false. They simply were not the material that survived into the finished film.
Bong Joon Ho does not shoot the kind of standard coverage many productions rely on, a wide shot and a matching close-up of everything, angle after redundant angle. Without that safety net, when a piece of a take does not quite work, the solution on Parasite has been to join parts of different takes together closely enough that the seam disappears. This is not a claim that every moment in the film was built this way. It describes what happens specifically when a single take does not fully hold, and it means that what plays as one uninterrupted shot may, in places, not be one piece of film at all.
An editor is not choosing from thin air. A take is a person doing something, on purpose, with their voice and their timing and their face, in front of other actors and a director who is watching for exactly this. Nobody can select an expression that was never given. What an editor can do is decide, among the expressions actually performed, which one belongs in the finished sequence, shaped by the shot before it, the shot after it, and what the scene needs the audience to feel next.
That is a stranger fact than it first sounds. There is no way to tell, from the outside, which second survived because it was the truest thing the actor did that day, and which survived because of what it needed to do next to everything around it. Every one of those possibilities was really performed, by a real person, in front of a camera. What survives in the finished film is not necessarily any single one of them whole; it may be assembled from more than one. Either way, only the selected material remains visible, and there is no way to know, from the finished film alone, what other performed possibilities existed.
When Sound Changes the Picture
A man is alone in a hotel room in Saigon, and the footage of him has a loose, immediate feeling, closer to something observed than something carefully arranged. He drinks. He moves in a way that does not look composed for the shot, unhurried and unpredictable the way an actual moment in a room usually is. Some of it was not originally meant to be in the finished film at all. Walter Murch, working without a script that told him what this opening was supposed to do, assembled it into something Francis Coppola felt could work as a kind of overture, setting up Willard’s state of mind before the story properly begins.
Watched silently, that footage shows a man behaving strangely in a room. What arrives over it is a voice, Captain Willard’s own, and the voice does not belong to any single photographed instant in the scene. It exists apart from the images, laid across them, and it is doing something the pictures alone cannot do: telling you what this man thinks of himself, from inside his own account of it. Murch found that narration in an earlier draft of the screenplay, where it had gone unused, recorded a temporary version of it himself, and placed it selectively across roughly the first half hour of the film. The experiment helped establish that Willard’s interior narration could work laid across the already assembled images; Michael Herr later wrote and revised the narration heard in the finished film. The pictures were already assembled. Willard’s spoken interior account arrived afterward, on a different track entirely.
Take the voice away and the room does not change. The man still drinks, still moves, still does what the camera recorded. What changes is what the moment means. Alone, the image is a record of behavior. With the voice added, the same footage becomes an account, a man narrating his own condition over pictures of himself in exactly that condition, and the two together produce something neither one contains by itself.
This is a different kind of relationship than a cut creates. Nothing about the picture itself changes when the voice arrives. There is no new shot, no new angle, no new piece of footage spliced in to signal that something has shifted. The image keeps running exactly as it was. What changes is entirely on the other track, and the audience experiences the two as one event anyway.
Murch has written about this kind of choice directly, describing what happens when a filmmaker deliberately withholds visual information rather than supplying it. “Starving the eye,” he wrote, “will inevitably bring the ear, and therefore the imagination, more into play.” He has also described the effect of placing a sound that does not quite match the picture as opening a kind of gap, a perceptual vacuum, that the audience’s own mind then has to fill. That principle helps explain what the opening is doing. The footage on its own gives you his body, his isolation, the shape of his distress. What it does not give you is his own account of what any of that means. The voice supplies that part.
The image does not become incidental. The footage still had to exist, still had to be shot, still had to be assembled into something a voice could be laid across. But whose experience that footage belongs to, and what it is actually telling you beyond what is visibly happening, changes through a relationship between two things that were never recorded together in the same room at the same time. A man drinking in a hotel and a voice describing his own mind exist in the finished film together because Murch laced one across the other.
The Hidden Cuts in Rope
At several points in Rope, the frame goes dark. Someone’s back fills it, or a piece of furniture, or the near side of a door, close enough that there is nothing left in the shot to see. For a second or two, the image gives you almost nothing useful to look at, no space, no action, nothing to place yourself by. Then the camera pulls back, or the person moves aside, and the scene continues.
That darkness is not a mistake and it is not incidental staging. It is where the film changes its mind about being one piece of footage and becomes another.
The physical film Hitchcock was shooting on could not simply run for as long as a scene needed. A load of Technicolor stock in the camera lasted roughly nine to ten minutes before it ran out, a limit set by the stock and the camera magazine, not by the story. So the film had to be built from segments about that length, and something had to happen at the end of each one. Some of those endings are visible, ordinary cuts. Five of them are not. Five times, one take ends by moving close enough to something dark or solid to fill the frame, and the next take begins the same way, close against a similar dark or solid surface, so that the two separately recorded pieces can be joined at a point where there is almost nothing to see. Joined there, the cut between them is hard to spot; darkness meets darkness, and the seam disappears into it.
Hitchcock called it a stunt. The film runs to roughly ten long takes, plus its opening credits as a segment of their own, eleven pieces in total joined by ten transitions. Five of those transitions are hidden this way. The other five are left as plain, visible cuts, some of them serving a practical purpose that has nothing to do with the story, placed where a theater’s projection reels would need to change over, giving that moment a clean, ordinary edit to work with instead of a hidden one.
That is worth being exact about, because Rope gets remembered, inaccurately, as a single unbroken shot. It is not. It is a film that goes to considerable, specific, physical trouble to make you feel that it is, while still containing real, countable boundaries between one piece of recorded film and the next.
Which raises a stranger question than how many cuts there are. The continuity you experience and the continuity the camera actually recorded are not the same thing. One is a sequence of physical segments, limited by how much stock a camera could hold at once, joined at points chosen partly to be invisible. The other is an evening that seems never to have been interrupted at all. Both are true of the same print. They are simply not the same fact.
Once you know to look for it, the trick stops being invisible, but it does not stop working. You can watch for the moment the frame goes dark and know, now, that something in the material is ending there. The film still gives you the feeling of an unbroken room. It has just stopped being able to hide, from you specifically, that the feeling and the film are two different things.
Why This Image Now
Watch an ordinary film tomorrow and the screen will look exactly the way it always has. Nothing about the picture itself will have changed. What has changed is what you now know to ask of it.
A camera records material. A finished film gives that material relationships, positions, and timing it did not have as recorded footage: which pieces stand next to which others, which piece arrives and which stays missing, how a sound relates to a picture it was never recorded with, and whether a boundary between one piece and the next stays visible or not. None of that decision-making replaces what an actor gave, what a director staged, what a cinematographer lit, what a composer or a sound recordist put into the world for someone else to use. The material still has to exist before anyone can do anything with it. But the material and the experience built from it are not the same object, and the difference between them is not a technicality. It is one of the places where a film actually happens.
That leaves a small, ordinary-looking question that turns out not to be small at all. Not which camera angle, though it can mean that. Why this recorded moment and not another one that was just as real. Why now, in this place in the sequence, rather than a few seconds earlier or later. Why arriving instead of the thing you expected. Why with this sound laid across it, or none at all. Why here, at this exact point, rather than somewhere the boundary could have stayed hidden.
Why this image now.
It is a small enough question to ask of a single cut. It does not get smaller the more you ask it. Put it to a film you already know well, one you could probably narrate from memory, and somewhere in it, at a moment you have watched a dozen times without once wondering why it landed exactly there, the question will still be waiting.