Gonjiam: Haunted Asylum — The Korean Found-Footage Hit

Found footage has one structural wound and it has never healed. Twenty minutes into any entry in the form, something happens that would cause a real human being to drop the camera and run, and the character does neither, and the film’s credibility bleeds out through that hole for the remaining hour. Every good found-footage film is really a proposed answer to the question why are you still filming? — The Blair Witch Project answered it with denial, [REC] with a professional reflex and a locked door, Noroi with an obsessive’s completed dossier. In 2018 a Korean film called Gonjiam: Haunted Asylum answered it with a number on a screen, and the answer was so obviously correct that it is faintly embarrassing nobody got there sooner.

The number is the live viewer count. Ha-joon (Wi Ha-joon) runs a web horror channel called Horror Times. He has assembled six young people, offered them a fee, and taken them into an abandoned psychiatric hospital to broadcast the exploration live. He has publicly staked his reputation on hitting one million concurrent viewers. That target sits in the corner of the frame, ticking up, all night. Nobody in this film keeps filming out of denial or professionalism. They keep filming because they are being watched, and because stopping would be a public humiliation with a real-time audience, and because the money only lands if the stream lands. It is 2018’s answer to a 1999 problem, and it is airtight.

The hospital that sued

Gonjiam Psychiatric Hospital is real. It stands in Gwangju, in Gyeonggi Province, closed since the mid-1990s, and it has spent thirty years accumulating the standard Korean urban-legend apparatus: a director who supposedly hanged himself, patients who supposedly died en masse, a persistent rumour about a room that cannot be opened. CNN once put it on a list of the world’s freakiest places, which did the legend no harm at all. It is, in short, a location with a pre-installed audience, which is precisely what the film exploits.

The production did not shoot there. The hospital’s owner went to court seeking an injunction to stop the release, arguing the film would damage the property’s value — the application was rejected, and the interiors were built and shot elsewhere, largely around Busan. The film’s most effective marketing was therefore a lawsuit about a building it never entered. Jung Bum-shik made a found-footage film about the commodification of a haunted place, and a real owner of a real haunted place sued him over the value of the commodity. You could not stage it better.

Craft: the rig

The mechanics here are worth being specific about, because Gonjiam solved a second problem the form has always had, which is that found footage is mostly the back of someone’s head.

Jung’s crew mounted cameras on rigs extending from the actors’ bodies — a short arm holding the lens out in front of the face, pointed back at it. The result is a shot that no other film language produces: a continuous, close, unbroken reaction shot of a person’s face, in which the darkness behind them is fully visible over their shoulder, and in which the actor cannot cheat. There is no cut to safety. There is no cutaway to the corridor. You are looking at a face and, behind it, the depth of the room, at the same time, for minutes on end.

That framing changes where the dread lives. In conventional found footage the horror is in front of the camera and the tension is will it appear. Here the horror is behind the face and the tension is has it appeared already, and does she know. Jung compounds it by running multiple feeds — the face rigs, a handheld, a phone, the drone shot over the building — and cutting between them the way a livestream director would, which means the audience is put in Ha-joon’s chair, choosing which face to watch. Every time you are looking at one, you are failing to look at the others.

The film also understands that a livestream has latency and chat. Comments crawl over the image. The viewers, thousands of them and climbing, react before the characters do. There is a specific, modern kind of horror in watching a stranger’s typed reaction arrive a half-second before the thing they reacted to comes into your frame.

The rules and the room

Ha-joon has salted the building. He has planted rigged scares, briefed some of the group and kept others in the dark, and instructed everyone to sell it for the audience. This is the film’s second good idea. For the first hour, every effect is under suspicion — the audience is invited to catalogue what is staged and what is not, and the characters are doing the same, laughing at what they can attribute and going quiet at what they cannot. The scepticism is doing the work that a slow burn usually has to do with atmosphere alone.

And then there is Room 402, which is the legend’s centrepiece: the door that has never been opened. Ha-joon’s stream is built around opening it. The film’s structure is a countdown to a door, and the reason the countdown works is that the audience has been trained for an hour to expect a fake, because everything else has been fake.

The real ancestor here is not Blair Witch. It is Grave Encounters (2011), the Canadian film in which a ghost-hunting TV crew with a producer’s cynicism enters an abandoned asylum that turns out to be architecturally hostile — same building type, same fraudulent-crew premise, same asylum-that-rearranges idea. Gonjiam is the version with better technique and a firmer grasp of why the crew cannot leave. If you are assembling the lineage, ten found-footage films that actually work has the map, and why found footage refuses to die argues the economics.

The light discipline

One more piece of craft deserves naming, because it is the hardest thing to do in this form and almost nobody does it. Gonjiam keeps its light honest.

Found footage routinely cheats. A crew supposedly carrying two torches will be lit from a third direction by a source that exists nowhere in the fiction, because the alternative is an unreadable frame and a distributor’s note. Jung mostly declines the cheat. The building is lit by what the characters are holding — phone torches, a camera light, the sick grey wash of a night-vision mode — and when they turn away from something, that something goes black. The film is prepared to give you a frame in which the useful information is simply absent, and to hold it.

That produces the picture’s best texture: the sweep. A torch beam moving across a room is an act of editing performed by a frightened person. They choose what to reveal, in what order, at what speed, and they always sweep too fast past the thing you wanted a second look at. The audience ends up shouting at the light. Conventional coverage cannot manufacture that, because in conventional coverage the director controls the reveal; here the reveal is delegated to a character with a shaking hand and every incentive to look away.

Wi Ha-joon holds all of this together, a couple of years before Squid Game made him internationally recognisable. He plays Ha-joon as a man whose showmanship is a job rather than a personality, and his slow slide from producer to participant is the only character arc the film properly funds.

The case against

The characters are thin. This is defensible — a livestream crew assembled by fee is meant to be a set of strangers — but the film asks you to care about seven people it has given roughly one trait each, and by the last act the deaths are arriving without much purchase. The middle stretch also leans on the standard inventory: the door that slams, the shape at the end of the corridor, the possession-adjacent behaviour change. Jung’s execution is far above the material, which is a way of saying the material is ordinary.

And the ending, which I will get to below, makes a choice that the film’s own rules do not entirely earn. It is a good film that occasionally acts like a great one and is caught out.

None of which stopped it. Made for a modest budget, Gonjiam took over two and a half million admissions in Korea, beat Tomb Raider at the domestic box office in its opening week, and became one of the most profitable Korean horror films ever made on a return-per-won basis. It streams widely. Watch it in the dark with the phone in another room, which is exactly the deprivation the film is about.

Spoilers below

The rigged scares stop being rigged. The film’s hinge is the moment the crew’s own equipment — the planted effects, the pre-agreed cues — starts producing results nobody planted, and the members who were in on the con realise the con has been joined by something else. Jung times this well: the audience’s catalogue of fakes, so carefully built, becomes useless in a single beat.

The group fractures the way they always do, and the picking-off runs through the standard set: the woman who becomes not-herself, the man who walks into the dark and comes back wrong, the pair who find a room they cannot explain. The strongest single sequence involves a character whose face-rig keeps rolling while their expression changes into something the actor is not performing for anyone in the room, only for us, which is the whole thesis of the rig paid off in one shot.

Room 402 opens, and the film’s final movement takes the crew inside — into a space that behaves less like a room than like a mouth. Ha-joon reaches his million viewers. He gets the number he staked everything on, which is the film’s bleakest joke: the stream is a triumph, the audience is enormous, and the currency he was paid in is his crew.

The last minutes push into an explicitly supernatural, geometrically impossible space, and this is where the film overreaches. Everything before it operates on the logic of a real building doing wrong things. The finale abandons that logic for a hallucination, and while it is startling, it cashes in the credibility the rig spent ninety minutes accumulating. A colder ending — the door open, the corridor beyond it perfectly ordinary, the count still climbing, the feed cutting — would have been the film this film had earned.