The word came from the screen before the screen was full of ghosts. On analog television, “ghosting” meant a faint, offset copy of the picture, and most of the ghost-like shapes on tape and on digital video have names just as ordinary: VHS tracking and head-switching noise, compression macroblocking, rolling shutter, low frame rates, infrared backscatter, long exposure, HDR ghosting, frame interpolation, lens flare. Each one leaves a signature you can check.
This page goes through them one at a time: what each looks like on the monitor, and how to check for it. None of them proves that nothing happened in the room. They are what has to be ruled out before the question gets interesting.
Multipath ghosting on analog TV
What it looks like. A second, fainter copy of the whole picture, shifted a little to the side. Every edge has a pale twin: the presenter, the furniture, the lettering on screen.
How to check. Multipath ghosting comes from the same broadcast signal arriving twice, once directly and once after bouncing off a hill or a building, a fraction later. So the copy is offset by the same distance for everything in the frame, not only for the figure you are looking at. If the doorway has a ghost and so does the clock, the signal made both.
VHS: tracking, head-switching, dropouts, bleed and creases
What it looks like. Horizontal bands of noise that roll through the picture when tracking is off. A ragged band of torn picture along the bottom edge, which is head-switching noise. Brief white streaks where oxide has worn off the tape, called dropouts. Color that spreads sideways past the edges of objects, called chroma bleed. And where the tape was creased, a burst of distortion that tears across the frame.
How to check. Note where the artifact lives. Head-switching noise sits at the bottom of the frame for the whole tape. Tracking noise moves when you adjust tracking or play the tape on another deck. Dropouts and creases are damage to the tape itself, so they come back at the same moment on every playback, on every machine. A shape that appears in a rolling band of noise and nowhere else belongs to the band.
Compression: macroblocking and smears in the dark
What it looks like. Digital video saves space by describing the picture in blocks and reusing what did not change. When the bitrate is low, dark areas break into visible squares, and anything that moves leaves a smear. On a low-bitrate security camera a person walking through a dim hallway can become a blob, or a translucent shape the wall shows through.
How to check. Step through frame by frame and look at the edges of the shape. If they follow a grid of squares, the encoder drew them. Look at the rest of the dark areas too: they will be blocky even when nothing moves. Then ask for the original export from the recorder. Every re-upload compresses again, so a repost can be far worse than what the camera saved.
Rolling shutter
What it looks like. Most phone and security camera sensors do not capture a frame all at once. They read it out line by line, top to bottom, so anything that moves fast relative to the camera skews, bends, or appears cut into pieces. A person running past a doorway can lean at an impossible angle. A spinning fan turns into curved blades that do not exist.
How to check. Look for other fast things in the same clip. If a passing car leans the same way, or vertical door frames tilt whenever the camera pans, the sensor is doing it. The distortion always follows the direction of readout, so it is consistent across the frame in a way a figure would not need to be.
Low frame rates
What it looks like. Many security cameras record under 15 frames a second. A walking person jumps between positions instead of moving smoothly, appears suddenly in a doorway, or seems to be missing from a frame entirely.
How to check. Find the frame rate in the file’s properties or with a media inspection tool. Then step through and count. At a low frame rate, a person crossing a narrow doorway may only be caught in one or two frames, and the gap between positions is simply the time between exposures.
Infrared night mode: orbs and rods
What it looks like. In night mode a camera floods the scene with infrared light from LEDs around its own lens. Dust, drops of moisture and insects a few centimeters from the lens catch that light and bounce it straight back. They appear as bright, soft “orbs” that drift or dart across the frame, and as “rods” when an insect crosses during a single exposure and blurs into a streak.
How to check. Orbs of this kind are out of focus, because they are far closer than the camera’s focus. They sit in front of everything, move faster than anything in the room could, cast no shadow, and disappear when the camera switches to daylight mode. A cobweb or a spider near the housing is a frequent source. This is backscatter, and a flash on a still camera does the same thing.
Long exposure
What it looks like. In a long exposure, anyone who stands still for part of it and then leaves is recorded for only part of the time. They come out transparent, with the background showing through. Phone night modes gather light over a longer time and can do the same.
How to check. Look at the exposure time in the file’s EXIF data. Look at other moving things: water blurs, lights leave trails, leaves smear. The mechanism is the same one behind the double exposures of the 1860s, where a real, living person stood in for the spirit. How an AI ghost photo gives itself away sets the two side by side.
Phone HDR ghosting
What it looks like. In HDR mode the phone takes several exposures in quick succession and merges them. Anything that moved between them, a person, a pet, a curtain, appears as a faint, partial double: HDR ghosting.
How to check. Look at everything else that could have moved. Leaves, hands and the edges of people will show the same doubling. Take the same shot again with HDR off and compare.
Frame interpolation on TVs
What it looks like. Many televisions invent in-between frames to smooth motion. When the guess is wrong, moving objects grow halos, warp at their edges, or leave fragments behind for a frame. Filming a TV screen with a phone captures the invented frames as if they were on the tape.
How to check. Watch the source file on a computer, or with motion smoothing turned off. If the shape is gone, the television drew it.
Motion-triggered cameras
What it looks like. A camera that only records when it detects motion starts recording mid-motion. A door is already swinging in the first frame. A figure is already in the hallway, with no footage of it arriving.
How to check. Look at the first frame of the clip and at the camera’s event log. Ask whether the camera keeps a pre-record buffer and how long it is. A clip that begins on an open door shows what triggered the camera, not what opened the door.
Lens flare, reflections and insects
What it looks like. A bright shape or a pale figure that hangs in the air. Lens flare comes from a strong light hitting the lens. Glass windows and screens reflect whatever is behind the camera, including the person filming. An insect walking on the lens becomes a huge, blurred shape crossing the room.
How to check. Lens flare moves when the camera moves, in line with the light source. A reflection moves with things behind the camera, so find out what was there. An insect on the lens is completely out of focus and usually shows legs in some frame if you step through it.
What ruling out means
Put a clip through all of this and it usually lands somewhere on the list. When it does, that is an answer about the camera, not a verdict on the people who were in the room or what they felt there. Generated clips have their own set of tells, covered in how AI ghost videos are made, and the rest of the On Camera section takes single clips through the same questions.
If you have footage that survives the list, send it to us. The ones worth keeping are the ones where every check comes back clean, and the doorway is still not empty.