This page collects the technical terms used across the audio and video archiving guides on this site. It’s alphabetical, with each entry deep-linkable — you can refer to a specific term from anywhere with a URL like /glossary-of-terms/#tbc. Where two terms mean the same thing, you’ll find a cross-reference rather than two copies of the same definition.
0-9 · A · B · C · D · E · F · G · H · I · J · K · L · M · N · O · P · Q · R · S · T · U · V · W · X · Y
0-9
3-2-1 backup — A standard data-redundancy discipline: three copies of every file, on at least two different storage media, with at least one copy stored off-site. The minimum sensible posture for an archive that is meant to survive longer than the original storage hardware.
3-head cassette deck — A cassette deck with separate dedicated heads for erase, record, and playback. The third (playback) head is widely considered to improve playback fidelity because it can be optimised for one job rather than compromised between record and playback duties.
3.5mm jack (1/8″) — The smaller of the two common headphone connectors. Found on phones, MP3 players, camcorders, and most consumer audio gear. Unbalanced.
4fsc — A sampling rate of four times the colour subcarrier frequency. For NTSC, 4fsc is 14.318 MHz; for PAL, 17.734 MHz. Used by digital composite video formats (D-2, D-3) and by vhs-decode’s intermediate .tbc files, which sample at 4fsc to represent the composite signal faithfully.
6.35mm jack (1/4″) — The bigger of the two common headphone connectors. Found on broadcast, radio, and studio gear; almost universal on studio headphones. Can be wired as unbalanced (TS) or balanced (TRS).
A
ADC (Analog-to-Digital Converter) — A circuit that converts a continuous analog signal into a series of digital samples. The chip at the heart of any analog capture card or audio interface.
AGC (Automatic Gain Control) — A circuit that automatically adjusts signal levels to a target range. In VHS playback, AGC can be triggered (intentionally or otherwise) by Macrovision-style pulses, producing rolling or darkening pictures.
AIFF (Audio Interchange File Format) — Apple’s uncompressed audio file format from 1988. Functionally identical to WAV for archival purposes — both store raw PCM audio with no compression, both are lossless.
Archival master — The lossless preservation copy of a capture, kept untouched so future tools and workflows can be applied to it. Distinct from a viewing copy, which is the playable derivative.
Audio channel — A single, mono stream of audio. Two channels make stereo; six make 5.1 surround.
AV1 — A modern royalty-free video codec, more efficient than H.265, slowly gaining hardware support. Not yet typical for archival use.
AVC (Advanced Video Coding) — Another name for H.264.
Azimuth — The angle of a tape head relative to the tape’s direction of travel. Misalignment between record-time azimuth and playback-time azimuth causes loss of high frequencies in audio, and chroma in video.
B
Balanced audio — A wiring scheme that reduces noise on long audio cables. Uses three conductors per signal (signal, inverted signal, ground); the receiving equipment subtracts the inverted from the non-inverted, cancelling any noise picked up equally on both. Found on XLR and 3-pole TRS connectors.
Baseband — A signal at its native frequency, near zero hertz, not modulated onto a higher-frequency carrier. Baseband audio is the waveform at a deck’s RCA outputs after the deck has decoded whatever was on the tape — whether the source on the tape was the linear analog audio track along the edge of the cassette, or the FM-modulated HiFi carriers that the deck demodulated internally. Composite video at the yellow RCA is also baseband. Contrast with a modulated signal, where the original has been encoded onto a high-frequency carrier — VHS HiFi RF at around 1.4 megahertz on the helical drum, the FM-modulated luma carrier in the video RF, or a radio broadcast. The term refers to frequency, not voltage; “baseband” doesn’t mean zero volts. See: Capture hardware in 2026.
Bit depth — The number of bits used to represent each sample. 16-bit gives 65,536 possible values per channel (CD quality); 24-bit gives ~16.7 million; 32-bit float covers an even wider range with finer detail. Higher bit depth gives more headroom for grading or restoration.
Bit rate — The number of bits used per second. For audio, calculated by multiplying bit depth × sample rate × channels. For video, depends on codec and content complexity.
Bit rot — Slow corruption of stored data over time, due to magnetic decay, flash-cell leakage, optical-layer breakdown, or simple medium failure. The reason archives need integrity checks and redundant copies.
bwdif (Bob Weaver Deinterlacing Filter) — An open-source FFmpeg deinterlacing filter. Produces decent results on most material at low computational cost. QTGMC produces noticeably better output but is much slower. Worth knowing about because Topaz Video AI uses bwdif internally as its deinterlacing pre-processor — the AI branding in Topaz is on the upscale or denoise stage that runs after deinterlacing, not on the deinterlacer itself.
C
Capstan — The motor-driven shaft inside a tape deck that pulls the tape past the heads at a controlled, constant speed. Capstan health is essential for stable playback; a worn capstan or pinch roller will produce speed instability and wow.
Capture card — A device (PCIe, USB, FireWire, or Thunderbolt) that takes an analog or digital video signal and writes it to disk as a file. Quality varies enormously between brands and tiers.
Cassette deck — A device that plays (and often records) audio cassette tapes. Comes in many forms — portable Walkman-style, rack-mount studio decks, separates for home Hi-Fi systems. The deck handles the mechanical transport of the tape past the playback heads.
CCD (Charge-Coupled Device) — A type of image sensor common in high-quality scanners and older digital cameras. Light hits the sensor and gets converted to an electrical charge that’s read out row by row. CCD sensors generally have better quality — wider dynamic range, less noise — than the cheaper CIS sensors used in printer-scanners.
Chroma — The colour information in a video signal, as distinct from luminance (brightness). Stored separately in nearly all video formats because the eye is less sensitive to chroma detail.
Chroma subsampling — A method of reducing the amount of data needed to represent chroma, by storing chroma at lower resolution than luminance. Common ratios are 4:4:4 (full chroma), 4:2:2 (half horizontal chroma), 4:2:0 (half horizontal and vertical chroma) and 4:1:1 (quarter horizontal chroma). See: Understanding 4:2:2 chroma subsampling.
Cineform — A line of professional intermediate codecs from GoPro (originally CineForm Inc.). Visually lossless at the higher quality grades but not mathematically lossless — discards data, just not enough for the difference to be visible at one generation. Used widely as a Windows-side editing intermediate in colour-graded film and video work. Not an appropriate archive format for the same reason ProRes isn’t: the decoded pixels are close to but not bit-identical to the input, and the format is proprietary. Treat as a viewing or delivery codec, not a master.
CIS (Contact Image Sensor) — A simpler image sensor type used in multifunction printers, sheet-feed scanners and other compact devices. The sensor sits right against the glass — hence “contact” — which is cheaper and faster than CCD but produces flatter images with less dynamic range and a very shallow depth of field. Anything thicker than a flat photo loses detail quickly.
Clockgen Mod — A small daughter-board modification for CX-card based RF capture rigs. It ties the audio and video sampling clocks together, fixing a problem where independent clock crystals drift slightly against each other over a long capture and produce audio-video sync issues you only notice afterwards. See: Capture hardware in 2026.
Colour-under — The method of recording colour information on tape used by VHS, Betamax, U-matic and similar consumer/prosumer formats. The colour signal is shifted down to a low frequency (around 629 kHz for NTSC, 627 kHz for PAL) before being recorded onto the tape, then shifted back up to the standard subcarrier frequency on playback. This is why VHS chroma has so little bandwidth, and why colour quality is the format’s biggest weakness. See: How vhs-decode actually works.
Closed captions — Text overlays carrying dialogue or descriptive audio for accessibility. On NTSC, transmitted on Line 21 of the VBI; on PAL broadcasts, typically through teletext. Discarded by most consumer capture workflows but preserved by RF capture.
Comb filter — A filter that separates the luminance and chroma components of a composite signal, used inside many VCRs and capture chains. Quality varies enormously; a bad comb filter produces chroma noise and rainbow artefacts on saturated edges.
Component video — A video signal split into three separate channels (typically Y, Pb, Pr) carried on three cables. Higher quality than composite or S-Video. Not used on consumer VHS but appears on broadcast and professional gear.
Composite video — A video signal where luminance and chroma are combined into a single waveform on one wire — the yellow RCA connector on most consumer gear. The simplest and lowest-fidelity of the common analog video formats.
Compression — A process for making files smaller. Lossy compression discards information considered unlikely to be noticed; lossless compression discards nothing and decodes back to the bit-identical original.
Container format — A file wrapper that can hold one or more video, audio, subtitle, and metadata streams. Common containers are MKV, MP4, MOV, and AVI. Distinct from the codec, which is how the data inside is compressed.
CRF (Constant Rate Factor) — An encoding setting in modern video codecs (x264, x265, AV1) that targets a perceptual quality level rather than a fixed file size. Lower numbers mean higher quality and bigger files; higher numbers mean smaller files and more visible artefacts. CRF 18 is generally considered visually lossless for SD analog material; CRF 20-23 is typical for delivery copies.
CRT (Cathode Ray Tube) — The original electronic display technology, using an electron beam to excite phosphors on a glass screen. Standard for video and computer displays before LCDs; still used for some calibration work because of how cleanly they handle motion.
cvbs-decode — A software-decode project for composite (CVBS) video sources captured directly at the source. Sister project to vhs-decode and ld-decode. Still maturing as of 2026, but covers the use case where the source is already composite and capturing it with high sample rates is more useful than running it through a conventional capture card. See: How vhs-decode actually works.
CXADC — A Linux driver and capture mode for certain Conexant CX2388x-based capture cards, allowing them to record raw uncompressed sample streams. Used in vhs-decode workflows for capturing baseband or RF signals. See: Capture hardware in 2026.
D
DAC (Digital-to-Analog Converter) — A circuit that converts a digital signal into an analog one. Found in CD players, audio interfaces, and the output stages of video equipment.
DAW (Digital Audio Workstation) — Software for recording, editing, and mixing audio. Examples include Adobe Audition, DaVinci Resolve’s audio side, OcenAudio, Audacity, Pro Tools, and Reaper.
DCT (Discrete Cosine Transform) — The mathematical transform at the heart of most lossy video codecs (JPEG, MPEG-2, H.264, ProRes). The codec converts each block of pixels into frequency components and quantises the less-visible high-frequency components more aggressively, which is how it discards data.
ddd-capture-toolkit — Open-source batch workflow software for the full vhs-decode pipeline: capture, decode, compress, export, audio align, mux, validate, archive. Grew out of fixing the DomesdayDuplicator’s audio-sync problem and now coordinates the entire chain — queuing, prioritising and tracking the multi-step work across many tapes — for any setup that can produce a raw capture file, with or without DdD hardware. Runs in a terminal on a headless Linux box, so the typical use is a dedicated machine working through a collection overnight. Cryptographically validates that the compressed archive copies match the original captures. See: Capture hardware in 2026.
Diffusion model — A category of generative AI that synthesises content by progressively removing noise from a random starting point, guided by the input. Diffusion models can produce strikingly realistic images and video and are the technology behind several recent video tools including Topaz Starlight and ByteDance’s SeedVR2. For analog video restoration they share the structural property of all super-resolution work — they produce plausible detail rather than recovered detail.
Deck — Shorthand for a video or audio tape player/recorder. “VCR” is the consumer term; “deck” is the serious-user equivalent.
Demagnetiser — A tool that removes unwanted residual magnetism from tape heads and ferrous components. Stray magnetism causes noise and gradual erasure of magnetic media.
Digital audio interface — A device that handles the conversion between analog audio (from a tape deck, microphone or instrument) and digital audio (in your computer). It contains an ADC for input and a DAC for output, and connects to the computer over USB, Thunderbolt or PCIe. The same kind of device you’d use for recording a podcast or producing music.
Dolby Digital — A multi-channel audio compression standard used on DVDs, broadcast, and some streaming. Not the same thing as Dolby NR (noise reduction).
Dolby NR (Noise Reduction) — A family of analog noise reduction systems used on cassette and reel-to-reel tape. Dolby B, C, S, and HX-Pro each reduce tape hiss in slightly different ways; they must be decoded on playback with the same setting they were recorded at.
Domesday Duplicator (DdD) — A hardware device built by the vhs-decode community that captures the raw RF signal from a VCR head, allowing software-based decoding instead of relying on the VCR’s own playback electronics. See: Capture hardware in 2026.
DPI (dots per inch) — A resolution unit, originally from printing where one “dot” was one ink dot per inch of paper. For scanners it refers to how many image samples are taken per inch of the original. Higher DPI means more detail captured — but only up to the limit of the sensor’s actual physical resolution (see optical resolution).
DPX (Digital Picture Exchange) — A 10-bit log image format used in film post-production. Each frame of a scanned film negative is typically stored as a DPX file, with the colour values encoded logarithmically so that shadow and highlight detail is preserved without crushing. Common in cinema-grade workflows.
Dropout — A momentary loss of signal during tape playback, caused by oxide loss, debris in the head path, or a brief loss of head contact. On VHS the FM carrier briefly disappears; the signal defaults to peak white, producing a white horizontal flash.
DSLR (Digital Single-Lens Reflex) — A camera body design where you look through the same lens that takes the picture, via a mirror that flips up out of the way during the exposure. The dominant serious-photography category from roughly 2000 to 2015, before mirrorless cameras took over. DSLRs and mirrorless cameras (which work the same way for archival purposes) are increasingly used as makeshift film scanners — the camera shoots a backlit negative through a macro lens, and software inverts the image.
DV (Digital Video) — A consumer digital video format from the late 1990s, intra-frame compressed at 25 Mbit/s. Transferred to computer over FireWire as a bit-identical copy of what’s on the tape; the encoding was done at record time.
DVCAM — Sony’s prosumer variant of DV, using a slightly faster tape transport. Same DV25 codec; plays back in any DV-compatible deck.
dvgrab — A Linux command-line tool for capturing DV (digital video) from MiniDV camcorders and similar devices over FireWire. The Linux equivalent of WinDV; often more reliable than the Windows tools on modern systems.
Dynamic range — The difference between the brightest detail and the darkest detail a system can capture without losing information. A camera or scanner with wide dynamic range can record both bright highlights and deep shadows in the same image; a narrow-dynamic-range device crushes one or both into uniform black or white. Measured in stops (for cameras) or decibels (for audio).
E
Ektachrome — Kodak’s consumer reversal film (slide film) from the 1960s onward. Unlike Kodachrome, its dye layers are not stable over decades; magenta-layer collapse causes cyan or pink shifts on old Ektachrome reels. Transfer Ektachrome sooner rather than later.
EDID (Extended Display Identification Data) — The handshake protocol used over HDMI and DisplayPort, where a display tells the connected device (computer, capture box) its supported resolutions and refresh rates. Sometimes an HDMI capture device needs special “EDID emulation” to convince a player that it’s a TV and to keep outputting video.
Encoding format — The codec used to compress video or audio inside a container. Often confused with container format; the same container (e.g. MP4) can hold many different codecs.
EXR (OpenEXR) — A high-dynamic-range image format developed at Industrial Light & Magic in 1999. Supports 16-bit and 32-bit floating-point per channel — much higher precision than standard 8-bit or 16-bit integer formats. Standard for digital cinema and high-end visual effects work.
F
FFV1 — A mathematically lossless video codec, open and patent-free, used as the standard archival codec by many national archives. Stores built-in integrity checks (checksums) for each segment of every frame, so if part of the file corrupts the codec can tell you exactly which segment failed.
Field — Half a frame of interlaced video — either the odd-numbered or even-numbered lines, taken at one moment in time. NTSC has 59.94 fields per second; PAL has 50. Two fields interleave to make one displayed frame.
FFmpeg — A free, open-source command-line tool that handles almost every audio and video codec, container format, filter and stream operation imaginable. Powers most other audio/video software underneath. If you ever do anything beyond a single click in a media app, you’ll eventually hit a problem that FFmpeg can solve.
FireWire (IEEE-1394, i.Link) — A high-speed serial connection used for DV, DVCAM, HDV, and Digital8 capture. Mostly absent from modern PCs, which has become the main practical bottleneck in MiniDV digitisation.
FLAC (Free Lossless Audio Codec) — A lossless, open audio codec. Technically both a codec and a container, but normally referred to as a codec. The de facto archival format for digital audio.
Flatbed scanner — A scanner where the original sits face-down on a glass surface, scanned by a sensor head that moves underneath. Standard form factor for paper photos, books, and (with a transparency adapter) film negatives and slides. Contrast with sheet-feed scanners, which pull the original through past a stationary sensor.
FM (Frequency Modulation) — A way of encoding a signal by varying the frequency of a carrier wave. Used for HiFi audio on VHS (also called AFM) and for video itself (VHS colour-under, LaserDisc composite).
Frame — A complete picture in a video signal. In interlaced video, a frame is made up of two fields shown 1/50th or 1/59.94th of a second apart.
Frame rate — How many frames per second a video runs at. NTSC video is 29.97 fps (or 59.94 interlaced fields); PAL is 25 fps (50 fields). Film is typically 24 fps.
Frame synchroniser — A device that locks an incoming video signal to a stable internal reference, stabilising its timing enough for downstream equipment to track it cleanly. Often confused with a TBC — many devices marketed as “TBCs” are actually frame synchronisers with a proc-amp attached. See: TBC vs Frame Sync.
G
Gamma — The curve that maps stored signal values to actual screen brightness. Designed to roughly match how the human eye perceives brightness — the eye is more sensitive to changes in shadows than in highlights, so gamma concentrates more detail down at the dark end. Different gamma standards (Rec. 601, Rec. 709, sRGB, BT.1886) describe slightly different curves; mismatched gamma between encoding and display causes brightness and saturation shifts.
Gamut — The range of colours a system can represent. Different colour spaces (Rec. 601, Rec. 709, P3, Rec. 2020) cover different gamuts. VHS has a relatively narrow native gamut.
Gate weave — Slight horizontal or vertical wobble of film as it passes through a projector or scanner gate, visible as the image moving subtly within the frame. Removable in post with stabilisation tools.
Generative AI — Software that produces new content (text, images, audio, video) by sampling from patterns learned during training, rather than retrieving or transforming existing content. In image and video work, generative AI underlies most “AI restoration” and “AI upscaling” tools, including Topaz Video AI’s newer models and the wider landscape of diffusion-based tools. Generative output is plausible but not evidentially faithful to a specific source; for archival purposes that distinction matters.
Generational loss — Cumulative quality loss when content passes through multiple encoding or copying steps. Especially visible with lossy codecs, repeated chroma-subsampling cycles, or repeated RGB ↔ YCbCr conversions.
GOP (Group of Pictures) — In inter-frame codecs (MPEG-2, H.264, H.265, AV1), a sequence of frames that begins with a keyframe and continues with predicted frames depending on the keyframe. Long-GOP codecs are more efficient but harder to edit than intra-frame codecs.
H
H.264 / MPEG-4 AVC — A widely used video codec with broad hardware support. Used for streaming, broadcast, web video, and Blu-ray. Less efficient than H.265 but more universally compatible.
H.265 / HEVC — A more efficient successor to H.264, achieving similar visual quality at roughly half the bitrate. Broad hardware support today, though the rollout was historically slow due to patent licensing complications.
H.266 / VVC — The next-generation codec after H.265, offering a further 30-50% bitrate reduction at the same quality. Very limited hardware support as of 2026.
HDV (High-Definition Video) — A high-definition format recorded on MiniDV-shaped tapes using MPEG-2 Long-GOP compression at 25 Mbit/s. Physically identical cassettes to MiniDV, but requires HDV-aware capture software.
Head alignment — The physical positioning and angle of a tape head. Misalignment causes loss of high frequencies in audio and chroma in video. Professional tape transfer often involves checking and adjusting head alignment for each tape.
Head amplifier — A small PCB inside a VCR sitting near the rotating head drum. It’s the first stage of electronics the heads’ output passes through, taking the very small RF signal coming off the spinning heads and amplifying it for the rest of the deck. The test points on the head amplifier expose the raw FM signal coming off the tape, which is what a vhs-decode capture rig taps into. See: How vhs-decode actually works.
Head switching — The point in a VHS frame where the helically scanned drum hands over from one head to the other. Appears as a band of noise at the very bottom of each frame. CRTs hid it in overscan; flat panels show it.
High definition audio — A consumer term for audio above CD quality. CD is 16-bit, 44.1 kHz; HD audio typically starts at 24-bit, 96 kHz.
Hi-Fi audio (AFM) — The high-quality audio track on VHS and Betamax tapes, encoded as a frequency-modulated FM signal on the same head as the video. Distinct from — and significantly better than — the linear audio track that also exists on the tape. Most archivists want to capture the Hi-Fi track because the linear track is usually noisy and limited.
hifi-decode — A software-decode project for the FM-modulated HiFi stereo audio tracks on VHS, Betamax and 8mm tapes. Reads the captured RF — either from the same capture as the video, or from a separate audio-head tap — and produces a digital audio file. Sister project to vhs-decode. See: How vhs-decode actually works.
HuffYUV — An older mathematically lossless video codec, 8-bit. Largely superseded by FFV1, which is more efficient and supports higher bit depths.
I
IEEE-1394 — See FireWire.
Inter-frame compression — A class of video compression that encodes some frames as differences from neighbouring frames. More efficient than intra-frame but harder to edit and less resilient to packet loss. Examples: MPEG-2, H.264, H.265, AV1.
Interlacing — A video scanning method that draws odd-numbered lines first, then even-numbered lines, in two separate fields per frame. Used by NTSC, PAL, and all standard-definition broadcast formats. Modern displays are progressive and have to deinterlace interlaced material for display.
Intra-frame compression — A class of video compression where each frame is encoded independently. Less efficient than inter-frame but more editable and more resilient to packet loss. Examples: ProRes, DV, FFV1, JPEG sequences.
I-frame, P-frame, B-frame — In modern video codecs (MPEG-2, H.264, H.265, AV1), three kinds of frame:
I-frame (Intra) — a complete, self-contained frame, like a JPEG.
P-frame (Predicted) — only stores the differences from the previous frame.
B-frame (Bidirectional) — predicts from both earlier and later frames.
This pattern is what makes inter-frame compression efficient — and what makes such codecs less editable than intra-frame-only ones like ProRes or FFV1.
IPS (In-Plane Switching) — An LCD panel technology common in modern monitors. Offers wider viewing angles and better colour accuracy than the cheaper TN (Twisted Nematic) panels it largely replaced.
IRE — A unit for measuring analog video signal level. Named for the Institute of Radio Engineers (now the IEEE). 100 IRE is reference white, 0 IRE is reference black. NTSC (except in Japan) traditionally has a 7.5 IRE “setup pedestal” — a black-level offset above zero — which is the source of the washed-out blacks you sometimes get on NTSC captures.
JEDEC — The Joint Electron Device Engineering Council, the standards body for solid-state memory specifications. JEDEC’s published standards include the data-retention specifications that explain why SSDs aren’t suitable for unpowered long-term storage: consumer-grade SSDs are specified to retain data for only around one year at thirty degrees Celsius when unpowered. Enterprise-grade flash has longer retention; in either case, archive use isn’t the design target. The relevant standard for retention is JESD218.
J
Jitter — Tiny variations in the timing of a signal. Time-base errors in VHS playback are a form of jitter; a TBC corrects them.
JPEG — The ubiquitous lossy image format from 1992. Great for sharing and viewing; not great as an archive because each re-save loses a little more detail. Use it for the viewer copy you hand around, not the master file you preserve.
K
Kodachrome — Kodak’s reversal film (slide film) from 1935 to 2009. Unlike Ektachrome, its dyes were added during processing rather than being built into the stock, giving exceptional long-term colour stability. 1950s and 1960s Kodachrome reels often still look excellent today.
L
LCD (Liquid Crystal Display) — The dominant flat-panel display technology, using liquid crystals modulated by an electric field to control light from a backlight.
ld-decode — The original software-decode project of the family. LaserDisc is an analog video format — the disc is optical, but the video signal on it is FM-modulated analog, not digital. ld-decode captures the raw RF coming off the player’s laser pickup, before the player’s internal demodulator gets to it, and decodes the video in software instead. (LaserDisc’s later PCM audio tracks are digital, and ld-decode handles those as a separate stream.) vhs-decode and cvbs-decode are sister projects that grew out of the same approach. See: How vhs-decode actually works.
Line TBC — A type of time-base corrector that operates line-by-line on the video signal as it plays. Found inside professional and high-end consumer VCRs like the Panasonic AG-1980 and JVC HR-S9000 series. The most effective TBC for VHS playback.
Linear interpolation — A method of filling in missing or damaged data by drawing a straight line between known good points on either side. In the context of analog video capture, vhs-decode uses linear interpolation between adjacent good lines to replace a line lost to dropout, which is a closer approximation to the missing content than the sample-and-hold method consumer VCRs use.
Lossless — Encoding that allows perfect reconstruction of the original signal. Mathematically lossless codecs (FFV1, FLAC, HuffYUV) decode to bit-identical data. Distinct from visually lossless, which is a weaker claim.
Lossy — Encoding that discards information considered unlikely to be perceived. Most consumer codecs (JPEG, MP3, AAC, H.264, H.265) are lossy.
LTO tape — Linear Tape-Open, a magnetic tape format used by institutions for cold archival storage. Capacities range from terabytes to tens of terabytes per cartridge. Drives are expensive, but tapes themselves are cheap per TB.
Macroblock — In modern video codecs (MPEG-2, H.264, H.265), the small square block of pixels — typically 16×16 for luma — that gets processed and compressed as a unit. When you see blocky compression artefacts in low-quality video, you’re looking at macroblock boundaries.
Luminance (Luma) — The brightness component of a video signal, as distinct from chroma. The eye is more sensitive to luminance detail than to chroma detail, which is why chroma subsampling works as well as it does.
M
M-Disc — An optical disc format claiming archival lifetimes of hundreds to thousands of years, using an inorganic data layer. Real-world testing supports decades of stability; rated lifetimes are unverifiable on those timescales. Useful for the irreplaceable-tier of an archive.
MFP (Multifunction printer) — A combined printer, scanner, and copier (sometimes plus fax) in a single unit. The scanner part uses a CIS sensor — fine for documents, limited for archival photo work because of the shallow depth of field and lower dynamic range.
MISRC — An open-hardware capture board designed by Stefan Ölsner for the vhs-decode community. It integrates the clock generation with sample capture in a single device, replacing the more cobbled-together CX-card + Clockgen Mod approach. Considered the longer-term replacement for that legacy combination. See: Capture hardware in 2026.
mkvpropedit — A command-line tool from the MKVToolNix suite that lets you edit the properties of an existing Matroska (MKV) file without re-encoding the streams. Adds or modifies tags, language metadata, track names, and segment information after the file has already been written. The standard tool for embedding archive metadata (tape ID, capture date, deck used, operator notes) into FFV1-in-MKV captures post-encode, without touching the underlying video.
Macrovision — A copy-protection scheme that inserts AGC-confusing pulses into the vertical blanking interval of a video signal. Designed to make a recording VCR misbehave while remaining invisible to a TV. Often misdiagnosed as the cause of rolling pictures on tapes that don’t actually have it.
MPEG-2 — A video codec used by DVD, broadcast video, and HDV. Inter-frame compressed. Considered legacy but well supported.
MP3 — A lossy audio codec from the late 1990s. The dominant consumer audio format for over a decade. Largely superseded by AAC (used by Apple and YouTube) for delivery and FLAC for archival, but still in widespread use because every device ever made plays it.
N
Noise floor — The lowest level of unwanted signal present in an analog (or sometimes digital) chain. On tape, noise floor includes tape hiss, head-amplifier noise, and other random components.
NLE (Non-Linear Editor) — Editing software that lets you arrange clips on a timeline and freely change the order. The name distinguishes it from older tape-based editing systems, where you had to record one clip at a time onto a master tape in order. Examples: DaVinci Resolve, Premiere, Final Cut, Vegas, Kdenlive, Avid.
nnedi3 — A neural-network-augmented edge-directed interpolator, used inside QTGMC for spatial gap-filling during deinterlacing. The “AI” component is small and narrow — nnedi3 was trained to predict missing pixels in interlaced video, and works at the per-line / per-block level rather than generating whole images. A foundational tool in the AviSynth and VapourSynth deinterlacing toolchain; an example of AI applied surgically to a narrow signal-processing task.
NTSC — National Television System Committee. The analog colour TV standard used in North America, most of South America, and Japan from the 1950s until digital TV took over in the late 2000s. 525 lines per frame, 29.97 frames per second, colour subcarrier at ~3.58 MHz. Distinct from PAL and SECAM, the European standards.
Nyquist sampling theorem — A foundational principle in signal processing: if you want to capture a signal containing frequencies up to X Hz, you must sample it at more than 2X times per second. Below that rate you lose information you can’t get back; above it, you can perfectly reconstruct the original signal (in theory). The basis of every digital-to-analog and analog-to-digital conversion. Often used to argue that 4:2:2 chroma sampling is enough for VHS — see the chroma article for the practical caveats.
O
OLED (Organic Light-Emitting Diode) — A self-emissive display technology where each pixel produces its own light. Offers perfect blacks and very high contrast; common in high-end TVs and phones. No backlight needed.
PAL — Phase Alternating Line. The European (plus Australia, NZ, parts of Asia and Africa) analog colour TV standard from the late 1960s until digital TV took over. 625 lines per frame, 25 frames per second, colour subcarrier at ~4.43 MHz. Generally considered technically superior to NTSC — higher resolution, more robust colour encoding.
Open format — A file format whose specification is freely available and not encumbered by patents. The right choice for archives because it can be decoded indefinitely without licensing or proprietary tooling. Examples: FLAC, FFV1, MKV.
Optical resolution — A scanner’s actual physical sensor density, measured in DPI. Distinct from the “interpolated” or “enhanced” resolution sometimes advertised, where the scanner takes the optical samples it actually has and uses software to invent values between them. Only the optical part adds real detail.
P
PCM (Pulse-Code Modulation) — Uncompressed digital audio — the native digital audio format on CD, DAT, and most professional capture pipelines.
PNG (Portable Network Graphics) — A lossless image format with built-in transparency support. Good for archival image storage if 8-bit or 16-bit per channel is enough for your source. For higher bit depth or floating-point data, use TIFF, DPX or EXR instead.
Proc-amp — A processing amplifier that adjusts brightness, contrast, chroma saturation, and hue of an analog video signal. Often bundled into frame synchronisers and consumer “TBC” boxes.
Progressive scan — A scanning method that draws every line of a frame in sequence, without interlacing. All modern flat-panel displays are natively progressive.
ProRes — A family of intra-frame codecs developed by Apple, used as a mezzanine codec for editing and grading. Visually lossless at the higher profiles, but not mathematically lossless — ProRes 4444 included.
Pulldown (3:2 pulldown / IVTC) — A technique for fitting 24-frames-per-second film into 60-field-per-second interlaced NTSC video, by repeating fields in a 3-2-3-2 pattern (one film frame becomes 3 video fields, the next becomes 2, and so on). When you want to recover the original 24p frames from such material, the process is called inverse telecine (IVTC). Common on NTSC store-bought movies on VHS and DVD.
Q
QLED (Quantum-dot Light-Emitting Diode) — An LCD variant using quantum dots in the backlight layer to produce more saturated colours. Not the same as OLED — QLED still has a backlight; OLED doesn’t.
QTGMC (Quasi-Temporal Gentle Motion Compensation) — The community-default deinterlacer in the AviSynth and VapourSynth ecosystems. Processes interlaced video field-by-field with motion compensation and temporal noise reduction, producing progressive output that’s significantly cleaner than simpler deinterlacers like yadif or bwdif. Uses nnedi3 internally for spatial gap-filling. Widely regarded as the most natural-looking deinterlacer for tape-grade analog video.
Quantisation — The step in digitising any signal where continuous values get rounded to the nearest available discrete value. Think of it like rounding 3.7 to 4 because you only have whole numbers to work with. Bit depth determines how many values are available — 8-bit gives 256 steps per channel, 16-bit gives 65,536. The more steps you have, the smaller the rounding error. Quantisation is also what lossy codecs use to throw data away — they map similar values to the same quantised level and save space by storing it once.
R
RCA — A common consumer audio and video connector, often colour-coded (red and white for audio, yellow for composite video). Unbalanced only.
Real-ESRGAN — An open-source AI image and video upscaler, widely used in the community for super-resolution of compressed or low-quality sources. Built around the ESRGAN architecture (Enhanced Super-Resolution Generative Adversarial Network) with training data tuned for “real-world” degradation rather than clean synthetic test sets. Available as a standalone tool, as a VapourSynth plugin, and through GUIs like ChaiNNer. Better on animation and clean digital content than on tape-grade analog material.
Rec. 601 — The colour space standard for standard-definition video, used by all NTSC and PAL broadcast and VHS material. Slightly different colour primaries and gamma to Rec. 709; untagged Rec. 601 content played back as if it were Rec. 709 produces visible colour shifts.
Rec. 709 — The colour space standard for HD video. Modern playback software assumes Rec. 709 by default, which is why tagging is essential when delivering Rec. 601 content from analog sources.
Reel-to-reel — An analog audio tape format with two separate physical reels — a feed reel and a take-up reel — loaded onto a deck, rather than enclosed in a cassette shell. Used from the 1940s into the 1980s by both home enthusiasts and professional studios. Wider tape and faster transport than cassette gave significantly better fidelity.
Regular 8 (Standard 8) — An amateur cine film format from 1932 to the mid-1970s. Made by exposing 16mm film on one side, then flipping and exposing the other, then slitting it in two after development. Pre-dates Super 8 and uses smaller frames.
RF capture — A capture method that records the raw radio-frequency signal coming directly off a VCR’s head, before the deck does any of its own decoding. The basis of vhs-decode and other software-decoded workflows. See: Capture hardware in 2026.
RF tap — A modification to a VCR that pulls the raw radio-frequency signal directly off the head before the deck’s internal electronics decode it. The captured RF is then decoded in software (vhs-decode), producing results that surpass what the VCR itself could output. The vhs-decode wiki maintains a “Tap List” of documented tap points across many decks. See: How vhs-decode actually works.
RGB — Red, Green, Blue. The colour space used by computer displays, image sensors, and most image formats. Each pixel is described by three numbers — how much red, green and blue light it should emit. Contrast with YUV/YCbCr, which separates brightness from colour and is used in video formats.
RIFE (Real-time Intermediate Flow Estimation) — An AI frame-interpolation tool that generates intermediate frames between existing ones, used for frame-rate conversion and motion smoothing. RIFE is the AI tool the analog-video community treats with least controversy — it solves a well-defined problem (given two frames, what would the frame between them look like?) where the authenticity concerns of super-resolution don’t apply in the same way.
S
S-VHS — Super VHS, a higher-resolution variant of VHS with separated luminance and chroma on tape and an S-Video output. The prosumer standard in the late 1980s and 1990s.
S-Video — A connector and signal type that carries luminance and chroma on separate wires within the same connector. Better quality than composite (which combines them on one wire); less than full component video.
Sample-and-hold — The dropout-compensation method used by most consumer VCRs. When the head loses contact with the tape and the signal disappears, the VCR’s dropout compensator pastes the previous good line over the bad one. Simple, fast, and visible — the same content is repeated for as long as the dropout lasts, which produces a smeared band rather than a clean replacement. Contrast with linear interpolation.
Sample rate — The number of audio samples taken per second. CD audio is 44.1 kHz; broadcast and video are typically 48 kHz. Higher sample rates (96/192 kHz) capture frequencies above the human hearing range, useful for processing headroom rather than for the listener.
SDI (Serial Digital Interface) — The professional video signal standard, carrying uncompressed video over a coaxial cable with a BNC connector. Standard in broadcast and post-production for connecting cameras, switchers and capture devices. Variants by bandwidth: SD-SDI (270 Mbit/s), HD-SDI (1.485 Gbit/s), 3G-SDI (3 Gbit/s) and so on.
SECAM — Séquentiel Couleur À Mémoire. The third major analog colour TV standard (alongside NTSC and PAL), used in France, Russia and parts of Africa. Same 625-line, 25 frames-per-second frame structure as PAL, but uses a different colour-encoding method — sending the two colour components on alternating lines rather than together, with the receiver remembering the previous line to fill in the missing one. Less common to encounter in archival work than NTSC or PAL.
Sheet-feed scanner — A scanner with an automatic feeder that pulls one sheet at a time past a stationary sensor — like the document feeder on a small office copier. Dedicated photo sheet-feed scanners (like the Epson FastFoto FF-680W) capture roughly one photo per second, making them the right choice for bulk paper photo digitisation.
Sinc filter — In mathematics and signal processing, the theoretical “ideal” filter for reconstructing a continuous signal from its samples — but only in theory. The sinc filter has infinite length, which is impossible to build in the real world. All real filters are finite-length approximations, which is why real-world digital signal processing is always slightly imperfect compared to what Nyquist’s theorem promises in the ideal case.
SNR (Signal-to-Noise Ratio) — The ratio of clean signal to noise in a chain, usually expressed in decibels. Higher is better. VHS chroma has only about 5 bits of usable SNR; the rest is noise.
Super 8 — An amateur cine film format introduced by Kodak in 1965, still manufactured today. Smaller sprocket holes than Regular 8, giving about 50% more picture area in the same 8mm width. Cartridge-loaded; far easier to use than Regular 8.
Super-resolution — A class of image and video techniques that take a low-resolution input and produce a higher-resolution output by inferring pixels that weren’t directly captured. Modern super-resolution is typically AI-based, with a trained neural network learning to map low-resolution to high-resolution. The fundamental issue is what’s sometimes called the one-to-many problem: many different high-resolution images can downscale to the same low-resolution one, so the model picks one. For analog tape, where the recording bandwidth is the hard limit on detail, any extra pixels produced by super-resolution are invented rather than recovered.
T
Tape hiss — The high-frequency random noise heard on analog magnetic tape. Caused by the discrete nature of the magnetic particles in the tape coating. One specific contributor to the broader noise floor.
TBC (Time-Base Corrector) — A device that corrects time-base errors in an analog video signal. A true TBC operates line-by-line on the signal; many devices marketed as “TBCs” are actually frame synchronisers with a proc-amp. See: Why a Time Base Corrector matters.
Telecine — The process of transferring film to video. Originally a real-time process using a film projector and a video camera; modern telecine is usually frame-by-frame digital scanning.
Teletext — A broadcast text-data service transmitted in the VBI of PAL signals, providing news, programme schedules, subtitles and similar information accessed via the TV remote. Largely defunct as a public service but the signal is still present on archival PAL recordings.
TIFF (Tagged Image File Format) — A versatile lossless image format. Supports 8-bit, 16-bit and 32-bit per channel, multiple channels, and various compression schemes (including no compression). The standard archive format for scanned photos and film, alongside DPX and PNG.
Time-base error — The variation in line and frame timing of a video signal as it plays back. Causes horizontal jitter and may prevent capture cards from locking. Corrected by a TBC.
Tracking — The mechanical alignment between a VHS deck’s head drum and the recorded track on tape. Mistracking shows as bands of noise or instability; the tracking knob lets the user fine-tune the alignment for a given tape.
TVL (Television Lines) — The standard measurement of horizontal resolution for analog video. Counts the number of distinct vertical line pairs that can be resolved across a width equal to the picture’s height. Approximate native TVL for common consumer formats: VHS ~240 (NTSC) / ~250 (PAL), S-VHS ~400, Video8 ~240, Hi8 ~400, MiniDV ~500. Broadcast SD sits around 500; LaserDisc CAV around 425. PAL versions of a given consumer tape format tend to sit slightly above their NTSC counterparts because the luma FM carrier is a little higher. TVL is a resolving-power measurement of what the recording format can actually carry, not a count of pixels in a digital file — a correct SD digitisation already captures whatever TVL the tape recorded; sampling at higher rates does not invent more TVL.
U
Uncompressed audio — Audio stored with no compression at all — larger file sizes, absolutely lossless. PCM in a WAV or AIFF container is the typical form.
V
v210 — A 10-bit 4:2:2 uncompressed video pixel format used by transfer houses, broadcast archives, and the BBC. Large files, but no codec processing in the chain.
VapourSynth — A frame-server framework with a Python scripting interface, used for video-processing pipelines. The modern successor to AviSynth in most post-production workflows. Supports a wide range of filters including QTGMC, neural-network-based filters, and AI tools wrapped as VapourSynth plugins. Most serious vhs-decode post-production chains run through VapourSynth at some stage.
VBI (Vertical Blanking Interval) — The brief period between video fields when the scanning beam returns from the bottom of the screen back to the top, leaving a “blank” interval in the signal. Used for non-picture data: closed captions on Line 21 (NTSC) or teletext (PAL), VITC timecode, copy-protection pulses (Macrovision), and other ancillary signals. Most consumer captures discard the VBI; vhs-decode and similar RF workflows preserve it.
VCR (Videocassette Recorder) — The consumer term for a VHS or S-VHS deck. Largely interchangeable with “deck” but more general.
vhs-decode — An open-source project for software-decoding raw RF captures from VCRs. Produces output equivalent to or better than the VCR’s own playback electronics, at the cost of a more involved capture setup. See: How vhs-decode actually works.
VHS recording speeds (SP, LP, EP/SLP) — VHS tapes can be recorded at different speeds to trade quality for runtime:
SP (Standard Play) — best quality, shortest runtime (2 hours per T-120 tape in NTSC).
LP (Long Play) — doubles the runtime, at a noticeable quality cost.
EP / SLP (Extended Play / Super Long Play) — triples the runtime, with significant quality degradation.
Most consumer VCRs can play back all three. The speed a tape was recorded at determines what quality you can capture from it.
Viewing copy — A delivery encode (typically H.264 or H.265 in MP4) that is easy to play and share. Regenerated from the archival master, never used as a replacement for it.
Vinegar syndrome — The chemical breakdown of acetate-based film (and some tapes), producing acetic acid as a by-product. Affected film smells distinctly of vinegar and becomes brittle and shrunken. Needs professional handling; do not run it through a household projector.
Visually lossless — A codec quality target where compression artefacts are below the threshold of casual perception. ProRes and DNxHR are visually lossless at higher profiles. Distinct from mathematically lossless, which is a stronger guarantee.
VITC (Vertical Interval Timecode) — Timecode information embedded in the VBI of a video signal. Allows precise frame-accurate addressing and synchronisation across multiple devices. Used in broadcast production and preserved by RF captures.
Vorbis comments — A simple key-value metadata format used inside FLAC and Ogg Vorbis files. Arbitrary fields can be embedded directly inside the file, so the metadata travels with the audio. The FLAC equivalent of Matroska tags; for the RF-capture workflow it’s the canonical way to embed capture-chain metadata (deck, date, operator, condition notes) into the RF FLAC file rather than relying on a separate sidecar that can be lost in a move.
vrecord — An open-source automated capture tool from the AMIA (Association of Moving Image Archivists) community. Pairs with Blackmagic and similar capture devices, providing real-time quality-check monitoring during capture (vectorscope, waveform, audio levels, dropped-frame counter). The preferred open-source workflow for analog video capture.
W
WAV (Waveform Audio File Format) — Microsoft’s standard container for uncompressed PCM audio, dating from 1991. Lossless, simple, universally supported. The de facto interchange format for audio.
WinDV — A tiny free Windows utility for capturing DV (digital video) from MiniDV camcorders and similar devices over FireWire. No editing features, just a Start button and a dropped-frame counter. The canonical tool for DV transfer on Windows for over 20 years.
X
XLR — A locking professional audio connector with a balanced 3-pin layout. Standard for microphones, line-level interconnect, and studio gear.
Y
Y/C — A short form for separated luminance and chroma. Also another name for the S-Video signal type.
YCbCr (Y’CbCr) — The digital video colour space, with Y as luminance and Cb/Cr as the two chroma components. Used in all common video codecs that don’t store full RGB.
YUV — A family of colour spaces used in video that separate brightness (Y) from colour information (U and V). The exact specifications vary — YCbCr is the digital variant used in modern codecs; YUV historically referred to analog component video. Separating brightness from colour is what enables chroma subsampling, which works because the human eye is far more sensitive to brightness detail than colour detail.
