Analogue and digital video looks, explained

A picture made of lines

A video camera scans the image in horizontal lines, top to bottom, and turns each line into a changing voltage. The tape or the broadcast carries that voltage, and the television draws the lines back again. Nothing in the chain stores a frame the way a strip of film does. That is why video faults look so different from film faults: they run along the lines, they live at the edges of the frame where the machine changed its mind, and they follow the timing of the signal rather than the grain of a chemical.

Composite color and color bleed

When color came to television it had to fit into the black-and-white signal that already existed. The solution was to send the brightness at full detail and squeeze the color into a narrow extra band. Home video recorders squeezed the color even further to fit it on cheap tape. The brightness of a VHS picture is soft, and its color is much softer.

You see it as color that bleeds sideways. A red jumper smears a little to the right across a white wall, lipstick spreads past the lips, and fine colored detail turns into a blur, while the edges in brightness stay comparatively sharp. The app does exactly that, in the recording-medium stage: it blurs the color, and only the color, along the line and shifts it slightly late. That is why the bleed runs left to right and never up and down.

Tape noise, tracking and head switching

• Tape noise. Magnetic tape is never completely even, and a cheap tape at slow speed shows it as fine streaks of noise running along the lines, different on every frame.
• Tracking. The heads that read a video tape spin on a drum and have to follow narrow diagonal tracks. When they drift off the tracks, bands of noise tear across the picture. The tracking knob on old players was there to fix it. The app tears a band near the bottom of the frame, which is where it showed up most on a home tape.
• Head switching. A VHS drum has two heads that take turns, and the hand-over happens just before the end of each field. On a television the overscan hid it. On a modern screen it shows up as a ragged strip of noise along the bottom edge of the frame. If you have ever digitized a family tape, you have seen it.
• Timebase jitter. Tape stretches and the motor speed wavers, so each line arrives a fraction early or late. Vertical edges shiver slightly sideways, line by line.

All of these are in the analogue tape technologies of the 1960s to the 1990s, strongest in the home formats and in Second-Generation VHS Copy and Worn Analog Tape.

Scanlines and tube bloom

Early television cameras used vacuum tubes instead of sensors. A tube struggled with bright lights: they bloomed, and smeared into short comet tails. On a home set the picture was drawn in visible lines with dark gaps between them.

The app draws both in its television looks. Black-and-White Television in the 1950s and Colour Television Broadcast in the 1960s get the tube's bloom and a vertical comet tail on the brightest lights, and a few looks, including the 1990s surveillance tape, get the dark scanlines.

CCD smear and autofocus hunting

The first camcorders with solid-state sensors used CCDs. Point one at a bright lamp or the sun and a vertical streak runs through the whole frame, because the sensor's charge spilled down the column while it was being read. Consumer CCD Smearing in the 1980s shows a shorter version: a streak above and below the brightest lights.

Autofocus arrived in consumer camcorders and brought a new habit: in low light or on plain surfaces the lens searched, going soft and sharp again in a slow pulse. Consumer Autofocus Hunting in the 1990s does that on purpose.

DV and MPEG compression

Digital tape removed the tape noise but brought a new kind of fault. A digital camcorder couldn't store every pixel, so it compressed the picture in small square blocks. Mini DV compressed each frame on its own and did it lightly, so the picture is clean, with a slightly crunchy texture in fine detail. DVD camcorders and later memory-card cameras used MPEG compression, which leans on the frames before and after. When the scene moves fast or the light is low, the blocks show.

The app breaks the picture into blocks and flattens the detail inside them, more strongly the more compression the recipe has. A handful of looks also show macroblocking, the larger, coarser blocks of a starved stream: Early Internet Video in the 1990s and the live-stream and social looks of the 2010s. Social Platform Compression in the 2020s gets them from DEGREE 8, and DVD Camcorder MPEG-2 in the 2000s from DEGREE 10.

Sensor noise

A digital sensor counts light, and in dim light the count is uncertain. The result is noise: a fine, colored speckle tied to individual pixels, worst in the shadows. It looks different from film grain, smaller, more regular and often tinted. Small sensors in phones and pocket cameras had more of it. The app adds sensor noise to the digital and low-light video looks, and to the 1980s gain-up camcorder that pushed its signal to see in the dark.

Rolling shutter

Most phone and CMOS camera sensors don't capture the whole frame at once. They read it row by row from top to bottom. If the camera or the subject moves during that read, vertical lines lean and the picture wobbles like jelly. The app bends the frame row by row in the 2010s, 2020s and 2030s looks, and in the 2000s Early DSLR Video.

Oversharpening

Small consumer cameras sharpened their pictures heavily to look crisp on small screens. Every edge got a thin bright or dark outline, and fine texture turned brittle. Phone cameras still do some of this. The app adds that edge halo to most of the digital looks, from the late 1990s Mini DV onwards. In the photo styles of the 2000s to 2030s it also adds the opposite fault, noise reduction, which smooths away some of the fine texture in skin and walls.

A few parameters in the recipes are not rendered yet, among them the tone mapping, log profiles and computational HDR of the newest decades. Those looks show their compression, noise, sharpening and color, but not those processing steps.