Glossary
Explore commonly used video streaming terms. All things video! Learn about codecs, streaming protocols, DRM, and more.
#
1080p
1080p is a video resolution of 1920 x 1080 pixels displayed with progressive scan, also known as Full HD. The number refers to 1,080 rows of pixels, and the p means every row is drawn in each frame. Shown in a 16:9 widescreen frame, 1080p is the default HD setting for most streaming services, TVs and online video.
2K resolution
2K resolution is a picture about 2,000 pixels wide. In cinema it means 2048 x 1080 pixels. For monitors, phones and game streams, 2K is often used for 2560 x 1440, also called 1440p or QHD. Both sit between Full HD (1080p) and 4K, giving a sharper picture than 1080p with far less data than 4K.
360p
360p is a low video resolution 360 pixels tall with progressive scan, typically 640 x 360 pixels for widescreen 16:9 video. It needs very little bandwidth, so streaming players use it as a fallback when networks are slow. 360p is unrelated to 360-degree video, which captures a full surrounding view.
3GP
3GP is a video container format defined by the 3GPP standards group for third-generation (3G) mobile phones. It was built to keep files small enough for slow mobile networks and limited phone storage, often using H.263 or H.264 video with AMR or AAC audio. 3GP is structurally similar to MP4 but is now largely obsolete.
480p
480p is a video resolution 480 pixels tall with progressive scan, usually 854 x 480 pixels for widescreen 16:9 video or 640 x 480 for older 4:3 video. It is called standard definition (SD) and is similar in sharpness to DVD. In streaming, 480p serves as a lower-quality fallback for slow or unstable connections.
4K resolution
4K resolution is a video resolution of about 4,000 pixels across, most commonly 3840 x 2160 pixels for TVs and streaming, also called Ultra HD (UHD). It has four times as many pixels as 1080p Full HD. Cinema 4K, used in film production, is slightly wider at 4096 x 2160 pixels.
720p
720p is a video resolution of 1280 x 720 pixels with progressive scan, often called HD. It uses a 16:9 aspect ratio and has fewer than half the pixels of 1080p. 720p is a common middle step in streaming, offering clear video at a much lower bitrate than Full HD.
8K resolution
8K resolution is a video resolution of 7680 x 4320 pixels, also called 8K UHD, with four times the pixels of 4K and sixteen times the pixels of 1080p. It offers extreme detail, mainly visible on very large screens viewed up close. Today 8K is used more in production, for cropping and reframing, than for everyday streaming.
A
AAC
AAC (Advanced Audio Coding) is a lossy audio codec that compresses sound into small files while keeping it clear for listeners. AAC is the default audio format in most streaming video, including HLS and MP4 files. AAC generally sounds better than MP3 at the same bitrate. Phones, browsers, smart TVs and streaming devices almost universally support it.
AC-3 (Dolby Digital)
AC-3, also known as Dolby Digital, is an audio codec that carries surround sound, typically 5.1 channels, alongside video. AC-3 was introduced in the early 1990s and became standard on DVDs, Blu-ray discs and broadcast TV. Most home theater systems, soundbars and smart TVs can decode it. Compared with newer codecs, AC-3 needs higher bitrates for the same quality.
AES-128 encryption (HLS)
AES-128 encryption (HLS) is a method of protecting HLS video by encrypting each segment with a 128-bit AES key, which the player downloads from a key URL before playback. It is built into the HLS standard and works in most HLS players without a DRM system. Because the key is delivered to the player directly, protection depends on securing that key URL.
AI captions
AI captions are on-screen text versions of a video's dialogue, generated automatically by speech recognition instead of typed by a person. The model transcribes the audio, a segmentation step splits the words into short timed lines, and the result is saved as a caption file such as WebVTT or SRT that the player loads beside the video. Closed AI captions can be switched on and off; open captions are burned into the picture. Either way, they make video usable for deaf and hard-of-hearing viewers and for anyone watching on mute.
AI dubbing
AI dubbing is the automated process of replacing a video's original spoken dialogue with speech in another language, using AI for transcription, translation and synthetic voice generation. Advanced systems try to keep each speaker's voice character and timing, and some adjust lip movements to match the new audio. The result is a version that viewers can listen to rather than read.
AI video analysis
AI video analysis is the automated process of running machine learning models over a video to extract useful information from its frames, audio and on-screen text. Typical outputs include detected objects, faces, logos, scene changes, transcripts and safety labels, each linked to a timestamp. The results turn an opaque video file into data that software can search, filter and act on.
AI video clipping
AI video clipping is the automatic creation of short clips from a longer video or a live stream, using AI to find the moments worth sharing. Models read signals such as the transcript, changes in audio energy, scene and shot changes, faces and on-screen action, then score candidate segments and pick start and end points that land on clean sentence and shot boundaries. Many clipping systems then render each clip ready to post: reframed for vertical screens, captioned for viewing on mute and given a title.
AV1
AV1 is an open, royalty-free codec from the Alliance for Open Media that squeezes video smaller than H.264 and usually a little smaller than HEVC at the same quality. It was released in 2018 by the Alliance for Open Media, a group that includes Google, Netflix, Amazon, Apple and Microsoft. AV1 plays in most modern browsers. Many newer phones and TVs also have hardware chips that decode it.
AV2
AV2 is the next-generation royalty-free video codec from the Alliance for Open Media, designed as the successor to AV1. AOMedia released the AV2 specification in June 2026. AV2 aims to deliver higher quality at the same bitrate, or the same quality using fewer bits, and adds support for multi-view and stereoscopic video. Hardware and software support is still at an early stage.
AVI
AVI (Audio Video Interleave) is a video file format, introduced by Microsoft in 1992, that stores video and audio together in a single container. The name means audio and video data are interleaved, or woven together, for smooth playback. AVI is an older format with limited support for modern codecs, streaming and subtitles, so today it is mostly found in legacy archives.
AVOD
AVOD (advertising-based video on demand) is free on-demand video paid for by ads that play before, during or after the content. Technically, each video carries ad break positions, an ad decision server picks the ads for each break and answers in the VAST format, and the ads are either played by an ad SDK in the player or stitched into the stream on the server. Impressions are reported back as tracking beacons, so a viewer who leaves before a break never generates one.
AVPlayer
AVPlayer is Apple's native media playback class, part of the AVFoundation framework, used to play audio and video in iOS, iPadOS, macOS and tvOS apps. AVPlayer plays HLS streams natively, including low-latency HLS and FairPlay DRM. Developers combine it with AVPlayerViewController or custom views to add controls, picture-in-picture and AirPlay.
Adaptive bitrate streaming (ABR)
Adaptive bitrate streaming (ABR) is a delivery technique where one video is encoded at several quality levels and the player switches between them during playback. The set of versions is called a bitrate ladder, and each version, or rendition, is split into short segments that start at the same points in time. Before each new segment, the player checks its measured download speed and how much video it has buffered, then picks the rendition that fits. The aim is steady playback at the highest quality the connection and screen can use.
AirPlay
AirPlay is Apple's wireless technology for streaming video, audio and screen content from an iPhone, iPad or Mac to an Apple TV, compatible smart TV or speaker. With video, the receiver often fetches the stream directly, while the Apple device acts as the remote. Apps using AVPlayer with HLS get basic AirPlay support with little extra work.
Anti-piracy
Anti-piracy is the set of technical controls and processes used to prevent, deter, detect and respond to unauthorized copying and redistribution of video. Prevention covers signed, expiring access, encryption and DRM. Deterrence covers visible watermarks and limits on account sharing. Detection uses fingerprinting to find copies on other sites and forensic watermarks to trace them to a source. Response covers takedown notices, revoking access and closing accounts. Because any video that can be watched can be recorded, anti-piracy aims to make copying harder, slower and traceable.
Anycast
Anycast is a network routing method in which many servers in different locations share the same IP address, and each user is automatically routed to the nearest or best-performing one. CDNs and DNS providers use anycast to send traffic to a nearby data center. If one location fails, traffic flows to the next closest one.
Apple ProRes
Apple ProRes is a family of high-quality video codecs made by Apple for editing and post-production rather than streaming. ProRes compresses lightly, so files keep near-original detail and are easy for editing software to work with. Versions range from ProRes 422 Proxy to ProRes 4444 XQ. Files are very large, often many times the size of an H.264 version.
Aspect ratio
Aspect ratio is the proportional relationship between a video's width and height, written as two numbers such as 16:9. Widescreen 16:9 is standard for TVs, laptops and YouTube, 9:16 is vertical video for phones, 1:1 is square, and 4:3 was common on older TVs. Aspect ratio describes shape, not resolution or quality.
Audio description
Audio description is an extra narration track that describes important visual details in a video, such as actions, scene changes and on-screen text, for people who are blind or have low vision. The narration is placed in natural pauses in dialogue. Audio description is usually offered as an alternative audio track that viewers can select in the player.
Automatic speech recognition (ASR)
Automatic speech recognition (ASR) is the branch of AI that recognizes spoken language in audio and converts it into text. ASR models learn the link between sound patterns and words from large collections of recorded speech. In video, ASR is the engine behind transcripts, captions, voice search and spoken-word indexing, and its accuracy is often measured as word error rate (the share of words it gets wrong).
Autoplay
Autoplay is a video player setting that starts playback automatically when a page or screen loads, without the viewer pressing play. Most browsers block autoplay with sound and only allow it when the video is muted or the user has interacted with the site. Autoplay is common for background videos, previews and social feeds.
Average bitrate
Average bitrate is the typical amount of data per second a viewer actually received during a view, weighted by how long each quality level played. Because adaptive streaming switches renditions as the network changes, this is the honest measure of the quality delivered, rather than the top quality on offer.
B
B-frame
A B-frame (bidirectional predicted frame) is a compressed video frame that can borrow picture information from both earlier and later frames. Using two directions lets B-frames achieve the best compression of any frame type. The encoder must send frames out of display order so the later reference arrives first. B-frames add a small amount of encoding and decoding delay.
BVOD
BVOD (broadcaster video on demand) is on-demand video offered by traditional TV broadcasters through their own streaming apps and websites. Examples include BBC iPlayer, ITVX and 9Now. BVOD services carry the broadcaster's shows after they air, and are often free with ads or funded by a license fee.
Bandwidth (video)
Video bandwidth is the amount of data a network connection can carry each second, measured in megabits per second (Mbps). It is different from bitrate, which is how much data a particular video stream needs each second. Playback stays smooth only while the viewer's available bandwidth stays above the bitrate of the rendition being played, with some margin to spare. On the publisher's side, bandwidth also describes the total data sent to viewers, which rises with every extra viewer, every extra minute and every step up the ladder.
Beacon (player beacon)
A player beacon, in video streaming, is a small data message that a video player sends to an analytics server to report playback events, such as play, pause, errors, quality changes and stalls. Beacons are sent automatically in the background while a video plays. Video analytics tools combine them to rebuild each viewing session and measure quality and engagement.
Bit depth
Bit depth in video is the number of bits used to store each color value for every pixel, which sets how many shades of color can be shown. Each color channel gets 256 possible shades in 8-bit video and 1,024 in 10-bit. Higher bit depth produces smoother gradients and less banding. HDR video requires at least 10-bit.
Bitrate
Bitrate is how much data goes into each second of video or audio, written in kbps (thousands of bits per second) or Mbps (millions of bits per second). Higher bitrate generally means better quality and larger files, but also more bandwidth needed to stream. Think of it like the width of a pipe: more data flows, but delivery costs more.
Brand safety
Brand safety is the set of practices that keep an advertiser's ads from appearing next to video or other content that could harm its reputation, such as violence, adult material, hate speech or misinformation. In video, brand safety relies on analyzing the content and context of each video before ads are placed. Advertisers and platforms often define custom categories that a brand wants to avoid.
Breakdown (video analytics)
A breakdown is one metric viewed across every value of one dimension, such as video startup time by country or error rate by browser. It is the standard way to find where a problem lives: the overall number looks fine, but the breakdown shows one device or region far worse than the rest.
Buffer fill (rebuffer duration)
Buffer fill is the total time, in seconds, that a viewer spent waiting while the player loaded more video after playback had already started. It is the raw number behind buffer ratio. A 30 second buffer fill on a 5 minute video is a bad experience; the same 30 seconds across a 3 hour match is barely noticed.
Buffer frequency
Buffer frequency is how often playback paused to buffer per minute of watching. Two viewers can have the same total buffering time, but one long pause feels very different from ten short ones. Frequency captures the second case, which is often the more annoying one.
Buffer ratio
Buffer ratio is the share of a viewer's watch time spent waiting for video to load, shown as a percentage. If someone watched for 100 seconds and the video paused to buffer for 3 of them, the buffer ratio is 3 percent. It is the most common single measure of how smooth streaming felt.
Buffering
Buffering is the player downloading video ahead of the playback position and holding it in memory, so short dips in the network do not interrupt the picture. That stored stretch of video is the buffer. When people say a video is buffering, they usually mean a rebuffer, or stall: the buffer has run empty, playback stops and a spinner appears until enough video arrives to resume. Stalls happen when segments arrive more slowly than they are played, for network, CDN, encoding or player reasons.
C
CBR (constant bitrate)
CBR (constant bitrate) is a video encoding mode that keeps the data rate roughly the same throughout a stream, whatever the scene complexity. CBR makes bandwidth use predictable. Simple scenes may waste bits, while complex scenes such as fast sports action may look worse. CBR is widely used for live streaming ingest.
CDM (content decryption module)
A CDM (content decryption module) is the software component inside a browser, operating system or device that decrypts DRM-protected video and passes it on for playback. The web page talks to the CDM through Encrypted Media Extensions (EME), but never sees the decryption keys. Examples include the Widevine CDM in Chrome and the FairPlay CDM in Safari.
CDN (content delivery network)
A CDN (content delivery network) is a network of servers spread across many locations that stores copies of content close to users so it loads faster. Instead of every viewer fetching a video from one central server, each viewer gets it from a nearby edge server. It works like local warehouses rather than a single central depot.
CEA-608 / CEA-708
CEA-608 and CEA-708 are US broadcast standards for closed captions that travel inside the video signal itself rather than in a separate file. CEA-608 is the older analog-era format with limited styling, while CEA-708 is the digital TV version with more fonts, colors and languages. In streaming, both are often carried inside HLS or DASH video so captions survive from broadcast to online.
CENC (common encryption)
CENC (common encryption) is an MPEG standard that sets one shared way to encrypt video, so a single encrypted file can be unlocked by different DRM systems. Its cbcs mode is the version that works with Apple FairPlay, Google Widevine and Microsoft PlayReady together. Without CENC, a service would need a separate encrypted copy of every video for each DRM.
CMAF
CMAF (Common Media Application Format) is a streaming standard that defines one shared fragmented MP4 format for video segments used by both HLS and MPEG-DASH. Before CMAF, services often stored the same video twice, once in TS for HLS and once in MP4 for DASH. CMAF also supports chunked delivery, which helps cut live streaming latency.
CRF (constant rate factor)
CRF (constant rate factor) is a video encoding setting that targets a consistent perceived quality level rather than a fixed bitrate. Busy, fast-moving scenes get more data under CRF, while still or simple shots get less. Lower CRF values mean higher quality and bigger files. CRF is common in x264, x265 and FFmpeg workflows.
CSAI (client-side ad insertion)
CSAI (client-side ad insertion) is a method where the video player itself requests and plays ads, pausing the main content to show them. The player fetches ad details, usually through a VAST or VMAP tag, then loads and plays the ad file before returning to the show. Ads and content come from separate sources.
Cache hit ratio
Cache hit ratio is the percentage of requests that a CDN serves directly from its cache, without fetching the content from the origin server. A high ratio means most viewers get video from nearby edge servers. It is calculated as cache hits divided by total requests.
Chroma subsampling
Chroma subsampling is a video compression technique that stores color information at lower resolution than brightness, because human eyes notice brightness detail far more than color detail. Chroma subsampling is written as ratios such as 4:4:4 (full color), 4:2:2 or 4:2:0. Most streaming video uses 4:2:0, which keeps a quarter of the color samples.
Chromecast (casting)
Chromecast (casting) is Google's technology for sending a video from a phone, tablet or browser to a TV, where the TV device fetches and plays the stream itself. The phone acts as a remote control, so battery use stays low. Casting works with Chromecast devices, Google TV and many smart TVs with Google Cast built in.
Chunked transfer encoding
Chunked transfer encoding is an HTTP/1.1 feature that lets a server send a response in pieces as they become ready, without knowing the total size in advance. In video streaming, it allows a CDN to forward parts of a live segment to the player while the encoder is still producing the rest. This is a key building block of low-latency DASH and some low-latency HLS setups.
Closed captions
Closed captions are on-screen text that shows a video's dialogue plus important sounds, such as music, laughter or a door slamming, and can be turned on or off by the viewer. Closed captions are written mainly for people who are deaf or hard of hearing. They are delivered as a separate text track, such as WebVTT, or embedded in the video stream as CEA-608 or CEA-708 data.
Cloud playout
Cloud playout is running a scheduled, always-on video channel from cloud servers instead of a broadcast control room full of hardware. An operator builds a rundown, a timed list of programmes, live feeds, breaks and fallback content, and the playout system plays it out as one continuous live stream. That stream can feed your own app, smart TV apps, FAST platforms or social channels. The schedule, switching and monitoring that used to need dedicated servers and staff in one room run as software you control from a browser.
Codec
A codec is software or hardware that compresses video or audio into a smaller file for storage or sending, then decompresses it for playback. The name joins the words coder and decoder. Common video codecs include H.264, HEVC and AV1, while AAC and Opus are common audio codecs. The choice of codec affects file size, picture quality and which devices can play the content.
Color space
A color space in video is a defined range of colors plus rules for how numbers in the file map to real colors on screen. Common video color spaces include Rec. 709 for standard HD and Rec. 2020 for HDR and wide color. When the file, encoder and display disagree on the color space, colors look washed out, too dark or oversaturated.
Container (video container)
A video container is the wrapper format that holds the video track, audio tracks, subtitles and descriptive data together as one file. Think of it as a box: the codec decides how each stream is compressed, while the container decides how those streams are stored and kept in sync. Common containers include MP4, MOV, MKV and WebM. Changing the container does not change the picture quality.
Content moderation
Content moderation is the practice of screening what users post, including text, comments, images, audio and video, and removing or restricting anything that breaks a platform's policies or local law. It is a system with four parts: written community guidelines, automated screening with filters and AI classifiers, human reviewers for borderline cases and reports, and an appeals path. Video moderation is the part of content moderation that deals with moving pictures and their soundtracks, and it is the most expensive part to do by hand.
Content protection
Content protection is the set of controls that decide who can watch a video and what can be done with it once delivered. It runs from access control, such as signed URLs and expiring tokens, through encryption and DRM, which lock the stream to licensed players, to output protection such as HDCP on external displays, watermarks that mark the picture, and monitoring for leaked copies. No single control covers everything. A protection setup is chosen layer by layer, matched to what a leak would cost and to what rights holders require.
Content-aware encoding
Content-aware encoding is an approach that analyzes what is in a video, such as motion, detail and scene changes, and adjusts compression settings to match. Easy scenes get fewer bits and demanding scenes get more. The goal is consistent visual quality at the lowest possible file size.
Custom dimension
A custom dimension is a label your own app attaches to each view, such as subscription plan, course name, A/B test group or creator ID, so playback metrics can be grouped by things only your business knows. Built-in dimensions cover devices and networks; custom ones cover your product.
D
DNxHD / DNxHR
DNxHD and DNxHR are Avid's professional video codecs designed for editing and post-production rather than delivery to viewers. DNxHD covers HD resolutions, while DNxHR extends the same approach to 2K, 4K and higher. Both compress lightly to keep quality high and make editing smooth. Files are usually stored in MXF or MOV containers and are very large.
DRM (digital rights management)
DRM (digital rights management) is the system that encrypts video and releases the decryption key only to approved viewers on approved devices. The video is encrypted once when it is packaged. When a viewer presses play, a secure module inside the device asks a license server for a license, which carries the key and rules such as an expiry time or whether offline playback is allowed. The three DRM systems in use today are Google Widevine, Apple FairPlay and Microsoft PlayReady.
DRM license server
A DRM license server is the service that checks whether a viewer is allowed to play protected video and, if so, sends the player a license containing the decryption key. The license can include rules, such as how long the video can be watched or whether it can be played offline. Each DRM system, like Widevine, FairPlay or PlayReady, has its own license protocol.
Deepfake detection
Deepfake detection is the use of AI and forensic techniques to identify videos, images or audio that have been synthetically generated or manipulated, such as swapped faces or cloned voices. Detectors look for clues like lighting mismatches, unnatural facial movement, compression artifacts and audio inconsistencies. Some systems also check provenance data, such as content credentials, that records how a file was created.
Dimension (video analytics)
A dimension, in video analytics, is any attribute you can group or filter views by: country, device type, browser, CDN, video title, player version. Metrics say how good playback was; dimensions say for whom. Slicing a metric by a dimension is how a vague problem becomes a specific one.
Dolby Atmos
Dolby Atmos is an immersive audio format that places sounds in three-dimensional space, including overhead, rather than in fixed speaker channels. In streaming video, Dolby Atmos is usually delivered inside an E-AC-3 (Dolby Digital Plus) audio track, and sometimes in Dolby AC-4. Compatible TVs, soundbars and headphones then render the sound for each listener's setup.
Dolby Vision
Dolby Vision is a proprietary HDR video format from Dolby that uses dynamic metadata to adjust brightness and color scene by scene for each display. Dolby Vision supports up to 12-bit color and very high peak brightness. Many TVs, streaming devices and phones support it, including iPhones, which record in Dolby Vision. Devices must be licensed to decode it.
Downscaling
Downscaling, in playback analytics, is when a video rendition is larger than the player showing it, so the device shrinks the picture. It wastes bandwidth and money without improving what the viewer sees. A 1080p stream playing in a small embedded player is a common example.
Dropped frames
Dropped frames are video frames the device failed to display in time and skipped, which viewers see as stutter or jerky motion. The network can be perfect and frames still drop, because the cause is usually a device or browser that cannot decode the video fast enough.
Dynamic ad insertion
Dynamic ad insertion (DAI) picks ads at playback time for each viewer and places them into marked breaks, instead of editing ads permanently into the video file. Two viewers watching the same episode can see different ads in the same break, chosen by location, device or audience. The breaks are marked with cue points or SCTE-35 markers, an ad decision server chooses the ads, and insertion happens either on the server (SSAI) or in the player (CSAI).
Dynamic watermarking
Dynamic watermarking is a technique that adds a watermark generated for each viewer or session, such as a username, email or ID, which can change position or appear at random times during playback. Because the mark is personal and moves around, a leaked recording points back to the viewer who shared it. Dynamic watermarks can be visible or invisible.
dash.js
dash.js is the open-source JavaScript reference player for MPEG-DASH, maintained by the DASH Industry Forum. dash.js uses Media Source Extensions to play DASH streams in web browsers, choosing quality levels automatically as network speed changes. It supports DRM through Encrypted Media Extensions, captions and low-latency DASH.
E
E-AC-3 (Dolby Digital Plus)
E-AC-3, also called Dolby Digital Plus, is an audio codec that delivers surround sound at lower bitrates than the original AC-3 (Dolby Digital). It supports more channels than AC-3, and streaming services commonly use it for 5.1 and 7.1 audio. It is also the most common way to carry Dolby Atmos in streaming. Most modern smart TVs, streaming sticks and soundbars support it.
Edge server
An edge server is a CDN server located close to end users that stores cached copies of content and delivers it to nearby viewers. When a viewer presses play, the video segments usually come from the nearest edge server rather than from the distant origin. This shortens the travel distance and reduces delay and buffering.
Egress (video delivery)
Egress, in video delivery, is data leaving a provider's network: from cloud storage, an origin server or a CDN out to viewers or to other systems. Cloud and CDN providers usually meter it per gigabyte and bill it separately from storage. Every minute of video watched creates egress, and so does every byte a player downloads ahead and never shows. For a streaming service it is often the line that grows fastest, because it scales with audience size, watch time and bitrate at once.
Encoding ladder / bitrate ladder
An encoding ladder (also called a bitrate ladder) is the set of resolution and bitrate pairs that one video is encoded into for adaptive streaming. A typical ladder might run from 360p at a low bitrate up to 1080p or 4K at a high bitrate. The player climbs up or down the ladder as the viewer's network speed changes.
Encrypted Media Extensions (EME)
Encrypted Media Extensions (EME) is a web browser API that lets a video player play DRM-protected content by connecting to a content decryption module built into the browser. EME handles the exchange of license requests and keys between the player and a DRM license server. It allows protected streaming without plugins, using systems such as Widevine, PlayReady and FairPlay.
Entropy coding
Entropy coding is the final step of video compression, where the encoder packs data into the fewest bits by giving common patterns short codes and rare patterns longer ones, much like abbreviating frequently used words. It is lossless, so it never lowers quality. H.264 High profile uses CABAC, a stronger entropy coder than the simpler one in the Baseline profile.
Exit before video start (EBVS)
Exit before video start, or EBVS, counts viewers who pressed play and then left before the first frame appeared. It is the clearest sign that startup was too slow. Unlike an error, nothing broke; the viewer simply ran out of patience.
ExoPlayer / Media3
ExoPlayer is Google's open-source media player library for Android apps, now part of the AndroidX Media3 project. ExoPlayer plays HLS, MPEG-DASH, Smooth Streaming and progressive files, with support for Widevine DRM, captions, ads and background audio. Developers prefer it to Android's older built-in MediaPlayer because it is more flexible and gets updates outside operating system releases.
F
FAST (free ad-supported streaming TV)
FAST (free ad-supported streaming TV) is a streaming model where viewers watch scheduled linear channels for free, and the service earns money from ads. FAST services look like a classic TV channel guide, but they are delivered over the internet to smart TVs, streaming devices and apps. Individual scheduled streams in these services are called FAST channels.
FAST channels
FAST channels (free ad-supported streaming television) are linear channels delivered over the internet that play a fixed schedule, the way a cable channel does. Viewers do not pay or pick a title; they tune in from a channel guide on a smart TV or app and watch whatever is on. Technically, each channel is a 24/7 live stream generated from a library and a schedule by a playout system, with ad breaks marked in the stream so ads can be inserted, and the ad revenue pays for the channel.
FFmpeg
FFmpeg is a free, open-source command-line tool and set of libraries for recording, converting, editing and streaming audio and video. It can read and write almost every codec and container in use. Many video services, editing apps and media servers run FFmpeg behind the scenes to transcode, trim, resize and package video.
FFprobe
FFprobe is FFmpeg's companion inspection tool: you point it at a media file from the command line and it lists the file's technical details without altering anything. It shows codec, resolution, frame rate, bitrate, duration, audio tracks and container information. Output can be printed as readable text or as JSON for scripts to use.
FLAC
FLAC (Free Lossless Audio Codec) is an open audio format that compresses sound without throwing away any information. A FLAC file usually takes up around half to two thirds of the space of the raw original, yet decodes back to an exact copy. FLAC is popular for music archives and high-resolution audio. FLAC is rarely used as the audio track in streaming video.
FLV
FLV (Flash Video) is a video container format created for Adobe Flash Player to deliver video over the internet. FLV files were the standard for web video in the 2000s, including on early YouTube. Since Flash reached end of life in 2020, browsers no longer play FLV, though the format lives on inside RTMP live streaming.
Face detection
Face detection is a computer vision task that finds human faces in images or video frames and marks where each one appears, usually with a bounding box. It answers the question "is there a face, and where?" without identifying who the person is. Face recognition is a separate step that matches a detected face against known identities, and it carries stricter privacy and legal obligations.
FairPlay
FairPlay is Apple's DRM (digital rights management) system for protecting video streamed over HLS to Apple devices, including iPhone, iPad, Mac, Apple TV and the Safari browser. FairPlay Streaming encrypts video and releases keys only to authorized players. To use it, a business must request FairPlay credentials from Apple.
Forensic watermarking
Forensic watermarking is a technique that embeds an invisible, unique identifier into each copy of a video so that a leaked copy can be traced back to the specific viewer, device or session it came from. The mark is designed to survive compression, cropping and screen recording. It is mainly used to investigate leaks rather than to prevent them.
Fragmented MP4 (fMP4)
Fragmented MP4 (fMP4) is a version of the MP4 format that splits video into small, self-contained fragments instead of one continuous file. Each fragment holds a few seconds of media and its own timing information. This makes fMP4 well suited to adaptive streaming, because players can request and play pieces without downloading the whole file.
Frame
A frame is a single still image in a video, and many frames shown quickly in sequence create the impression of motion. A typical video shows 24 to 60 frames every second. In compressed video, frames come in types: I-frames store a full picture, while P-frames and B-frames store only changes from nearby frames to save space.
Frame interpolation
Frame interpolation is a technique that creates new in-between video frames by estimating motion between existing frames. It can raise a video's frame rate, for example from 30 to 60 frames per second, or produce smooth slow motion. Modern approaches use motion estimation or AI models to predict what the missing frames should look like.
Frame rate (FPS)
Frame rate (FPS, frames per second) is the number of still images, or frames, shown each second to create the illusion of motion in video. Film usually runs at 24 fps, TV and web video at 25 or 30 fps, and sports or gaming at 50 or 60 fps. Higher frame rates make motion look smoother but need more data.
G
GOP (group of pictures)
A GOP (group of pictures) in video encoding is a sequence of frames that starts with a keyframe (I-frame) followed by predicted P-frames and B-frames. GOP length, the number of frames between keyframes, shapes compression, seeking and streaming behavior. Longer GOPs give smaller files. Shorter GOPs make seeking faster and quality switching more responsive.
GPU transcoding
GPU transcoding is the use of graphics processors (GPUs) to decode and re-encode video, rather than relying only on a computer's main processor. Modern GPUs include dedicated video engines that can handle many streams at once. This makes transcoding much faster, which is especially useful for live video and large upload volumes.
GStreamer
GStreamer is an open-source multimedia framework for building audio and video pipelines by linking small processing blocks (called elements) together. A pipeline might capture a camera, encode the video, and send it out as a live stream. Developers use GStreamer inside apps, devices and servers where custom, real-time media handling is needed.
Geo-blocking
Geo-blocking is the practice of allowing or denying access to video based on the viewer's geographic location, usually worked out from their IP address. A sports match might be available in one country but blocked in another. It is mainly used to honor content licensing deals and local regulations.
Glass-to-glass latency
Glass-to-glass latency is the total delay between light hitting a camera lens and the matching image appearing on a viewer's screen. It adds up every step along the way: capture, encoding, upload, processing, CDN delivery, player buffering and decoding. Measuring it end to end shows the real delay a viewer experiences, not just one part of the chain.
H
H.264 (AVC)
H.264, also called AVC (Advanced Video Coding), is a video compression standard that plays on almost every phone, browser, smart TV and set-top box. It was published in 2003 and became the default codec for web and broadcast video. H.264 offers solid quality at moderate file sizes and fast encoding. Newer codecs compress better, but none match its device support.
H.265 (HEVC)
HEVC, short for High Efficiency Video Coding and also called H.265, is a compression standard that can match H.264 picture quality while using roughly half the data. It was published in 2013 as the successor to H.264. HEVC is widely used for 4K and HDR streaming. Its main drawbacks are complex patent licensing and patchy support in web browsers.
H.266 (VVC)
H.266, also called VVC (Versatile Video Coding), is a video compression standard finalized in 2020 as the successor to HEVC. H.266 aims to deliver the same quality as HEVC at roughly half the bitrate, which helps with 4K, 8K and 360 degree video. Encoding is very demanding. Device and browser support remains limited, and patent licensing is complex.
HDCP
HDCP (High-bandwidth Digital Content Protection) is a copy protection standard that encrypts video as it travels over a cable, such as HDMI or DisplayPort, from a playback device to a screen. It stops protected video from being captured by a recording device placed between the player and the TV or monitor. If a connected screen does not support HDCP, playback may be blocked or reduced in quality.
HDR (high dynamic range)
HDR (high dynamic range) video is video that shows a wider range of brightness and color than standard dynamic range (SDR), with brighter highlights, deeper shadows and richer color. HDR video typically uses 10-bit or higher bit depth and the Rec. 2020 color space. Common HDR formats include HDR10, HDR10+, Dolby Vision and HLG. Viewers need a compatible screen to see the benefit.
HDR10
HDR10 is an open, royalty-free HDR video format that uses 10-bit color, the Rec. 2020 color space and the PQ brightness curve. HDR10 carries static metadata, meaning one set of brightness information applies to the whole video. HDR10 is the most widely supported HDR format and serves as the baseline that HDR TVs and many streaming devices handle.
HDR10+
HDR10+ is an HDR video format that builds on HDR10 by adding dynamic metadata, which adjusts brightness and tone settings scene by scene or frame by frame. HDR10+ is royalty-free for content creators and was developed by Samsung with partners. It competes with Dolby Vision. It falls back to standard HDR10 on screens that do not support it.
HDS (HTTP Dynamic Streaming)
HDS (HTTP Dynamic Streaming) is Adobe's legacy adaptive streaming protocol that delivered video in fragments over standard web servers to Flash Player. It used an F4M manifest file and F4F fragment files. HDS worked much like HLS and DASH, but it depended on Flash, which Adobe stopped supporting in 2020, so it is now effectively obsolete.
HLG (hybrid log-gamma)
HLG (hybrid log-gamma) is a royalty-free HDR video format developed by the BBC and NHK for broadcast and live TV. HLG needs no metadata, and a single HLG signal looks acceptable on both HDR and standard dynamic range screens. This makes HLG well suited to live sports and events, where producing separate HDR and SDR feeds is costly.
HLS (HTTP Live Streaming)
HLS (HTTP Live Streaming) is a streaming protocol created by Apple and published as RFC 8216. It delivers video as short media segments over ordinary HTTP, so any web server or CDN can carry it. A text playlist with the .m3u8 extension tells the player which segments exist and in what order. A multivariant playlist lists several versions of the same video at different bitrates, so the player can switch between them as the connection changes. HLS plays natively on Apple devices and through JavaScript players elsewhere.
HLS encryption
HLS encryption is the protection of video streamed over HLS (HTTP Live Streaming) by encrypting its segments so only authorized players can play them. HLS supports simple AES-128 encryption of whole segments, and SAMPLE-AES, which encrypts the media inside each segment and is used with FairPlay DRM. Keys are delivered through a key URL or a DRM license server.
HTML5 video
HTML5 video is the native way to play video in a web browser using the <video> element, without plugins like Flash. The element provides basic playback, controls, captions through text tracks and a JavaScript API for custom players. On its own, HTML5 video plays simple files like MP4, while adaptive streaming in most browsers needs Media Source Extensions.
HTTP/3
HTTP/3 is the third major version of the protocol web browsers use to fetch pages and files, and it runs on QUIC over UDP instead of TCP. HTTP/3 connects faster and handles packet loss better, especially on mobile networks. For video, HLS and DASH segments can be delivered over HTTP/3 without changing the streaming format itself.
Hardware encoding
Hardware encoding is video encoding done by dedicated chips, such as GPUs or built-in media engines, rather than by general software running on a CPU. Phones, laptops and broadcast encoders all use hardware encoders for speed. It is much faster and uses less power, though software encoders often reach better quality at the same bitrate.
hls.js
hls.js is a free JavaScript player library that plays HLS streams in browsers without built-in HLS support. It uses Media Source Extensions to download segments, switch quality levels and feed video to the HTML5 video element. hls.js also supports captions, multiple audio tracks, low-latency HLS and DRM through EME.
I
I-frame (keyframe)
An I-frame (intra-coded frame), also called a keyframe, is a complete video frame that is compressed on its own without referring to any other frame. An I-frame in video is unrelated to an HTML iframe on a web page. Players can start decoding from an I-frame, which makes I-frames essential for seeking and starting streams. I-frames are much larger than P-frames or B-frames.
IDR frame
An IDR frame (instantaneous decoder refresh frame) is a special I-frame in H.264 and HEVC video that tells the decoder to discard all earlier reference frames and start fresh. No frame after an IDR frame may reference anything before it. This makes IDR frames clean starting points for playback, seeking and switching between stream qualities.
Ingest
Video ingest is the process of bringing video into a system, whether by uploading a file or receiving a live stream, so it can be stored, processed and delivered. For on-demand video it covers uploads and imports from URLs or storage buckets. For live video it means accepting a feed over protocols such as RTMP, SRT or WebRTC.
Interactive live streaming
Interactive live streaming is live video where viewers take part instead of only watching, through chat, polls, quizzes, reactions, purchases, predictions or joining on camera. Technically it is two systems running together: a live video stream delivered to every viewer, and a separate real-time data channel that carries the interactions. The two have to stay close enough in time that a poll, a question or a buy button matches what is on screen, which is why interactive streams need low latency and careful timing.
Interlaced video
Interlaced video is a display method that draws each frame in two halves, called fields, one with the odd lines and one with the even lines, shown one after another. It was designed for older TV broadcasts to make motion look smooth while saving bandwidth. Formats such as 1080i and 480i use interlacing.
iframe
An iframe (inline frame) is an HTML element that loads one web page inside another, and it is the most common way to embed a hosted video player on a website. The video player runs in its own isolated frame, served by the video host. Attributes on the iframe control size, fullscreen permission and autoplay.
K
Key rotation
Key rotation, in video streaming, is the practice of changing the encryption keys that protect a stream at regular intervals, instead of using one key for the whole video or event. If one key leaks, it only unlocks a short part of the content. Key rotation is most common in live streaming and long-running linear channels.
Keyframe interval
A keyframe interval is the setting that controls how often a video encoder inserts a full keyframe (I-frame), measured in seconds or frames. A keyframe interval of 2 seconds at 30 fps means one keyframe every 60 frames. Live streaming services often require a fixed interval, commonly 2 seconds, so they can cut the stream into even segments.
L
LCEVC
LCEVC (Low Complexity Enhancement Video Coding), standardized as MPEG-5 Part 2, is a video coding method that adds a thin enhancement layer on top of an existing codec such as H.264, HEVC or AV1. The base codec carries a lower-resolution picture. The enhancement layer adds back fine detail. This boosts quality or cuts bitrate without replacing devices or decoders already in the field.
LL-HLS (low-latency HLS)
LL-HLS (low-latency HLS) is an extension of Apple's HTTP Live Streaming protocol that cuts live stream delay to around 2 to 5 seconds. It splits each segment into smaller parts that the player can download before the full segment is finished. The server can also hold a playlist request open until new content is ready, so the player does not waste time asking repeatedly.
Latency
Latency is the delay between something happening in front of a camera and a viewer seeing it on their screen. If you hear your neighbor cheer a goal before you see it on your own TV, the gap you are noticing is latency. In live streaming it is measured in seconds, from capture to playback.
Letterboxing
Letterboxing is the technique of adding black bars above and below a video so a wide image fits a narrower screen without cropping or stretching. It happens, for example, when a 2.40:1 cinema film plays on a 16:9 TV. Adding bars on the left and right instead, for narrow video on a wide screen, is called pillarboxing.
Linear TV
Linear TV is television that plays programs on a fixed schedule, where every viewer sees the same content at the same time. Viewers cannot pick a show on demand. They tune into a channel and watch what is airing. Linear TV started with broadcast and cable, and now also runs as scheduled streaming channels over the internet.
Live DVR
Live DVR is a feature that lets viewers pause, rewind and replay a live stream while it is still running, like a TV recorder for online video. The streaming service keeps a rolling window of recent video, for example the last hour, available to the player. Viewers can then jump back to a missed moment and return to the live edge.
Live captions
Live captions are text versions of speech shown on screen while a live stream or broadcast is happening. They are produced by streaming speech recognition, which processes audio as it arrives, or by a human stenographer typing in real time. Because the words are recognized on the fly, live captions trail the speaker by a few seconds and usually contain more errors than captions made after the event. They are delivered as a timed text track alongside the video or embedded inside the video signal, and the player shows them in sync with the picture.
Live edge
The live edge is the newest point of a live stream that a player can show, the moment closest to real time. In HLS or DASH it is the latest segment the server has published. A viewer watching at the live edge has the smallest possible delay, while a viewer who has paused or rewound is behind it.
Live encoder
A live encoder is software or hardware that compresses a camera feed in real time and sends it to a streaming server over a protocol like RTMP or SRT. It turns raw video into a smaller, compressed stream that can travel over a normal internet connection. Examples include OBS Studio, vMix and dedicated hardware boxes.
Live stream latency
Live stream latency is the delay between something happening in front of the camera and a viewer seeing it, measured in seconds. Standard HLS live streams often run 10 to 30 seconds behind, low-latency HLS around 2 to 5 seconds, and WebRTC under a second.
Live streaming
Live streaming is delivering video over the internet while it is still being captured, so viewers watch an event as it happens. A camera or screen feeds a live encoder, which compresses the picture in real time and sends it to a streaming service over an ingest protocol such as RTMPS or SRT. The service transcodes the feed into several qualities, cuts it into short segments and pushes them through a CDN to players, which keep switching quality as each viewer's network changes.
Live streaming API
A live streaming API lets developers broadcast live video from an encoder or app to viewers on any device through code. The API creates a stream with a stream key and ingest URL, receives the feed over a protocol such as RTMP, RTMPS or SRT, transcodes it into a bitrate ladder as it arrives, packages it as HLS or LL-HLS and delivers it through a CDN. Most also record the broadcast for replay and report stream health while it runs.
Live-to-VOD
Live-to-VOD is turning a live stream into an on-demand video that viewers can watch after the broadcast ends. The streaming service records the stream as it is ingested, closes the recording when the stream stops, and publishes it as a normal on-demand asset with its own playback link, often trimmed or clipped first. Done in the platform rather than by hand, the replay is ready shortly after the event finishes, with no download and re-upload, and keeps the same access rules as the live stream.
Logo detection
Logo detection is a computer vision task that finds brand logos in images or video frames and reports which brand appears, where it is on screen and for how long. Models are trained on examples of known marks so they can spot them on shirts, signs, products or screen overlays. The output is usually a list of brands with timestamps and confidence scores.
Lossless compression
Lossless compression is a way to shrink files so they can be restored exactly, bit for bit, to the original. Lossless compression removes only redundant data, not real detail, so quality never drops. FLAC for audio and PNG for images are common examples. For video, lossless files remain very large, so they are used for editing and archiving rather than streaming.
Lossy compression
Lossy compression is a way to shrink video, audio or image files by permanently removing details people are unlikely to notice. Lossy compression produces far smaller files than lossless methods, which makes streaming practical. H.264, HEVC, AV1 and AAC are all lossy. Higher compression saves more space but can cause visible blur, blockiness or audio artifacts.
Low latency
Low latency in live streaming means a short delay between something happening in front of the camera and a viewer seeing it on screen, known as glass-to-glass latency. Standard HLS and DASH live streams often run tens of seconds behind real time because players wait for several long segments. Low-latency setups such as LL-HLS and low-latency DASH send video in much smaller pieces and let players sit closer to the live edge, which brings the delay down to a few seconds over ordinary CDNs.
Low-latency DASH
Low-latency DASH (LL-DASH) is a mode of MPEG-DASH that reduces live streaming delay to a few seconds by sending video in small chunks while each segment is still being encoded. It relies on CMAF chunks and HTTP chunked transfer encoding, so the player receives frames almost as soon as they are produced. The player also keeps a small buffer and adjusts playback speed slightly to stay close to live.
M
M3U8
An M3U8 file is a text-based playlist that tells a video player where to find the segments of an HLS (HTTP Live Streaming) stream. It lists segment URLs and their durations and, in a master playlist, the available quality levels. The player reads the M3U8 file first, then downloads the video pieces it needs. M3U8 files contain no video themselves.
M4S
An M4S file is a media segment in fragmented MP4 format, holding a few seconds of video or audio from an adaptive stream. M4S segments are used by MPEG-DASH and by HLS streams that use fMP4 or CMAF. A player downloads them in order after reading a manifest, which tells it where each segment lives.
MKV
MKV (Matroska Video) is an open, free video container format that can hold many video, audio and subtitle tracks in a single file. It supports almost any codec, plus chapters and attachments such as fonts. MKV is popular for archiving films with several audio languages and subtitle options, but browsers and mobile devices support it less widely than MP4.
MOV
MOV is a video container format created by Apple for its QuickTime framework, used to store video, audio, timecode and metadata in one file. iPhones and Mac editing tools save video as MOV by default. MOV is closely related to MP4, and it often carries H.264, H.265 (HEVC) or Apple ProRes video.
MP4
MP4 (MPEG-4 Part 14) is a video container format that stores video, audio, subtitles and metadata in one file. It usually holds H.264 or H.265 video with AAC audio. MP4 plays on almost every phone, browser, smart TV and editing tool, which makes it the most common format for sharing and uploading video.
MPD file
An MPD file (Media Presentation Description) is the XML manifest that describes an MPEG-DASH video stream. It lists the available video and audio quality levels, languages, segment locations and timing. A DASH player reads the MPD first, then picks which segments to download based on the viewer's network speed and screen size.
MPEG-2
MPEG-2 is a video and audio compression standard from the mid-1990s that powered DVDs and digital broadcast television. MPEG-2 is far less efficient than H.264 or HEVC, so files are much larger at the same quality. Many cable, satellite and over-the-air TV systems still use it. The MPEG-2 standard also defines the transport stream format used in broadcasting.
MPEG-DASH
MPEG-DASH, or Dynamic Adaptive Streaming over HTTP, is an international streaming standard that breaks video into short chunks delivered from ordinary web servers. A manifest file called an MPD describes the available quality levels and where each segment lives. The player reads the MPD and switches quality as network conditions change. Unlike HLS, which Apple defines, DASH is an open ISO standard not controlled by a single company.
MPEG-TS
MPEG-TS (MPEG transport stream) is a standard container format for sending video, audio and data over broadcast and network links that may drop or corrupt packets. It splits media into small fixed-size packets of 188 bytes, with error-handling and timing information. MPEG-TS powers digital TV broadcasting and is a long-standing segment format for HLS streaming.
MXF
MXF (Material Exchange Format) is a professional video container format used in broadcast and film production to store video, audio, timecode and rich metadata together. It was standardized by SMPTE (the Society of Motion Picture and Television Engineers). Broadcast cameras, editing systems and playout servers use MXF to move high-quality footage between tools without losing information.
Macroblock
A macroblock is a small square block of pixels, typically 16 by 16, that older video codecs such as H.264 and MPEG-2 use as the basic unit for compressing each frame. The encoder predicts and compresses each macroblock separately. Newer codecs such as HEVC and AV1 use larger, flexible blocks instead. Over-compressed video often shows visible blocky edges along these boundaries.
Manifest file
A manifest file in video streaming is a small text file that tells the player which versions of a video exist and where to find each segment. It lists quality levels, audio and subtitle tracks, segment URLs and timing. HLS uses M3U8 manifests and MPEG-DASH uses MPD manifests. The player reads the manifest before downloading any video.
Mean opinion score (MOS)
Mean opinion score (MOS) is a rating of video or audio quality calculated by averaging scores given by human viewers, usually on a scale from 1 (bad) to 5 (excellent). MOS comes from controlled viewing tests where people rate sample clips. It remains the reference that automated metrics such as VMAF try to predict.
Media ID
A media ID is the internal identifier a video platform assigns to an uploaded video or recording. The API uses it to manage the asset: encode it, add subtitles, create clips, delete it. Viewers never see it; they see a playback ID instead.
Media Source Extensions (MSE)
Media Source Extensions (MSE) is a web browser API that lets JavaScript feed video and audio data into the HTML5 video element piece by piece. MSE makes adaptive streaming possible in browsers that cannot play HLS or DASH natively. A player library downloads segments, chooses the right quality and appends them to a buffer that the browser then decodes and plays.
Media over QUIC (MoQ)
Media over QUIC (MoQ) is an emerging IETF protocol for live and on-demand video that runs on QUIC, the fast transport behind HTTP/3. MoQ aims to combine the delay of under a second of WebRTC with the large-scale delivery of HLS and DASH. Media flows through relays that can fan out a stream to many viewers, much like a CDN, while letting late or less important data be dropped.
Microdrama
A microdrama is a scripted drama series told in very short episodes, often one to three minutes long, filmed in vertical format for mobile phones. Stories usually move fast and end each episode on a cliffhanger. Microdrama apps typically offer the first episodes free, then charge per episode through coins or subscriptions, or show ads.
Model Context Protocol (MCP)
Model Context Protocol (MCP) is an open specification that gives AI assistants and agents one shared way to plug into outside tools, data and services. An MCP server exposes actions, such as uploading a video or fetching playback analytics, that any compatible AI client can discover and call. MCP works like a universal adapter, so one integration serves many AI applications.
Motion compensation
Motion compensation is the video compression technique of building a frame by moving blocks from a reference frame according to motion vectors, then storing only the small difference that remains. Motion compensation lets codecs avoid resending unchanged parts of a scene. It is why P-frames and B-frames are far smaller than full I-frames. The decoder repeats the same process during playback.
Motion estimation
Motion estimation is the step in video encoding where the encoder searches previous or future frames to find where each block of pixels has moved. Motion estimation produces motion vectors (small arrows describing each movement). The encoder can then describe a new frame mostly by reference to earlier ones. Motion estimation is usually the most computing-intensive part of encoding.
Multi-CDN
Multi-CDN is a delivery setup that serves the same video through two or more content delivery networks and decides, for each viewer or request, which one to use. The decision can rest on measured performance, region, cost or a live outage. Each CDN has different server locations and different connections to internet providers, so the fastest network in one country is often not the fastest in the next. If one CDN slows down or fails in a region, new requests move to another and playback keeps going.
Multi-DRM
Multi-DRM is protecting the same video with several DRM systems, usually Widevine, FairPlay and PlayReady, so it plays securely on almost every phone, browser and TV. No single DRM system runs everywhere, so each device uses the one it supports. With common encryption (CENC), one set of encrypted files can serve all three systems. Each system still has its own license server, so a shared entitlement service decides who may watch and each license server delivers the same key in its own format.
Multi-audio tracks
Multi-audio tracks are several alternative audio streams attached to one video, letting the viewer pick a language, commentary or accessibility version in the player. The video is stored once, while each audio track is a separate rendition listed in the HLS or DASH manifest. This saves storage and lets audiences switch audio without restarting playback.
Multimodal AI
Multimodal AI is artificial intelligence that can process and connect more than one type of data, such as images, video, audio and text, within a single model or system. For video, this means the model can relate what is said to what is shown on screen. A multimodal model can, for example, read a slide, hear the speaker and summarize both together.
Multipart upload
A multipart upload is a method of sending a large file by splitting it into smaller parts that are uploaded separately and then joined back together on the server. Parts can be sent in parallel to speed things up, and a failed part can be retried alone. Cloud object storage services widely support multipart uploads for large media files.
Multistreaming / restreaming
Multistreaming, also called restreaming, is sending one live video feed to several streaming platforms at once, such as YouTube, Twitch, Facebook Live and your own website. In the usual setup the encoder sends a single stream to a cloud service, which duplicates it and pushes a copy to each platform's ingest URL with that platform's stream key. Compared with pushing a separate output from the encoder for every platform, this keeps upload bandwidth and encoder load at the level of one stream.
Muxing / demuxing
Muxing (multiplexing) is the process of combining separate video, audio and subtitle streams into a single container file, and demuxing is the reverse process of splitting them apart. An MP4 file, for example, holds a muxed video track and audio track that stay in sync. Players demux the file so each stream can go to the right decoder.
N
NSFW detection
NSFW detection is the use of AI classifiers to identify content that is "not safe for work", such as nudity, sexual material or graphic violence, in images and video frames. Each frame or clip receives a confidence score for categories like explicit or suggestive. Platforms then compare those scores to their own thresholds to decide whether to block, blur, age-gate or send content for human review.
Named entity recognition
Named entity recognition is a language AI technique that finds and labels the names of real-world things in text, such as people, organizations, places, products and dates. In video workflows, named entity recognition runs on transcripts, captions and on-screen text, so a video can be tagged with who and what it mentions. Each entity is linked to the timestamp where it appears.
O
OBS Studio
OBS Studio (Open Broadcaster Software) is free, open-source software for recording video and live streaming from a computer. It captures cameras, screens, microphones and media files, mixes them into scenes, and sends the result to a streaming service using a server URL and stream key. OBS Studio runs on Windows, macOS and Linux.
OTT
OTT (over-the-top) is the delivery of video and audio content directly to viewers over the internet, bypassing cable, satellite and broadcast TV providers. Viewers watch OTT content through apps on smart TVs, phones, computers and streaming devices. Netflix, Disney+ and YouTube are well-known examples of OTT services.
OVP (online video platform)
An OVP (online video platform) is a hosted product for uploading, managing, publishing and streaming video without running your own video infrastructure. It bundles encoding, storage, CDN delivery, an embeddable player, a media library with metadata, and viewing analytics behind a web dashboard. Most OVPs are built for a person who uploads a video and pastes an embed code, and many also expose an API for developers. The platform owns the pipeline; you configure it rather than build it.
Object detection
Object detection is a computer vision task in which an AI model finds specific things in an image or video frame, such as people, cars, animals or products, and marks each one with a label and a bounding box (a rectangle showing its position). In video, detection runs across many frames, so software can tell when and where each object appears.
Opus
Opus is an open, royalty-free audio codec designed for both speech and music over the internet. Opus works well at very low bitrates and adds very little delay, which makes it ideal for real-time communication. It is the default audio codec for WebRTC and is common in WebM files. Opus was standardized by the IETF in 2012.
Origin server
An origin server is the main server that holds the original copy of video files and manifests, from which a CDN fetches content when it does not already have a copy cached. Viewers rarely talk to the origin directly. Instead, edge servers request content from the origin once and then serve many viewers from their cache.
Origin shield
An origin shield is an extra caching layer that sits between a CDN's edge servers and the origin server, so that edges ask the shield for content instead of the origin. Many edge requests for the same segment are combined into one request to the origin. This protects the origin from overload and lowers data transfer costs.
oEmbed
oEmbed is an open format that lets a website turn a plain link to a video, image or post into a rich embedded player automatically. The site sends the URL to the provider's oEmbed endpoint and receives back embed HTML, a title, a thumbnail and dimensions. This is how pasting a video link into many CMS editors or chat apps produces a playable preview.
P
P-frame
A P-frame (predicted frame) is a compressed video frame that stores only the changes from an earlier frame instead of the full picture. P-frames use motion compensation to describe how parts of the image moved. Because they reference earlier frames, P-frames are much smaller than I-frames. A decoder needs the earlier reference frame before it can show a P-frame.
PSNR
PSNR (peak signal-to-noise ratio) is a video and image quality metric that measures how much a compressed frame differs from its original, pixel by pixel, expressed in decibels (dB). Higher PSNR means less difference from the source. PSNR is fast and simple to calculate, but it does not always match what human eyes notice.
PVOD
PVOD (premium video on demand) is a form of transactional video where viewers pay a higher price to watch new movies or shows early, often while they are still in cinemas or soon after. It is usually a rental for a set period at a premium price. Studios use PVOD to earn more from new releases.
Page load time
Page load time, in video analytics, is how long the web page holding the player took to finish loading. It matters because a viewer cannot press play until the page is ready, so a slow page shows up as a slow video even when the stream is fast.
Pay-per-view
Pay-per-view (PPV) sells access to a single live event or title for a one-time payment, instead of an ongoing subscription. Technically, a purchase creates an entitlement, a record that this buyer may watch this event. At playback, your backend turns that entitlement into a short-lived signed token, and the video platform plays the stream only when the token is valid. High-value events add DRM, and a replay window is enforced by when tokens and licenses expire.
Per-title encoding
Per-title encoding is a way of building a separate bitrate ladder for each video, based on how hard that video is to compress. A screen recording or simple animation can look clean at a low bitrate, while handheld sports footage with grass, crowds and fast motion needs far more data at the same resolution. Instead of applying one fixed table of resolutions and bitrates to every upload, the encoder analyzes each title and picks the rungs that give the most visible quality for the fewest bits.
Picture-in-picture
Picture-in-picture (PiP) is a viewing mode where a video keeps playing in a small floating window on top of other apps or browser tabs. Viewers can move and resize the window while they browse, message or work. PiP is supported in major desktop browsers and on iOS and Android through their native player APIs.
PlayReady
PlayReady is Microsoft's DRM (digital rights management) system for protecting video, used in the Edge browser, Windows apps, Xbox and many smart TVs and set-top boxes. PlayReady encrypts content and issues licenses that set playback rules, such as rental windows or output restrictions. It supports both on-demand and live streaming.
Playback ID
A playback ID is the public identifier a video platform puts in a stream URL so a player can request that video, separate from the internal ID used to manage the file. Keeping the two apart means a platform can make a video public or private, or rotate access, without touching the asset itself.
Playback error
A playback error is any failure the video player reports while trying to load or play a video, from a missing file or expired token to an unsupported codec or a DRM license refusal. Each error carries a code, a short message and technical context so engineers can trace the cause.
Playback score
A playback score rates whether a view succeeded at all. A view that played counts fully, a view the viewer abandoned before start counts partly, and a view that hit an error counts as a failure. Averaged across many views, it shows what share of play attempts actually turned into watching.
Playback speed
Playback speed is a video player setting that changes how fast a video plays, such as 0.5x for slow motion or 2x for double speed. Good players keep audio pitch natural at higher speeds, so voices do not sound squeaky. Playback speed is supported by the HTML5 video element and by native players on iOS and Android.
Player initialization time
Player initialization time is how long the video player takes to load and get ready after the page asks for it, before any video is requested. It is separate from video startup time. Heavy player bundles, slow scripts and third-party plugins all add to it.
Pre-roll / mid-roll / post-roll
Pre-roll, mid-roll and post-roll are video ads named by when they play: a pre-roll runs before the content starts, a mid-roll interrupts it partway through, and a post-roll plays after it ends. Each position trades viewer patience against attention. Pre-rolls reach the most people, while post-rolls are often watched by fewer viewers.
Private video hosting
Private video hosting is a way of hosting video so that only approved viewers, such as signed-in users, employees or paying students, can play it. Videos are not listed or searchable, and every playback request is checked against a rule your own system controls, usually through a signed, expiring link or token. Private hosting is about who can watch. Secure hosting adds controls that make copying harder, such as encryption, DRM and watermarks, so a private library can still be recorded by someone who has access.
Progressive scan
Progressive scan is a video display method that draws every line of each frame in order, from top to bottom, in a single pass. It produces cleaner motion than interlaced video, which splits each frame into two halves. The p in 1080p and 720p stands for progressive, and phones, computers and modern TVs all display video this way.
Q
QUIC
QUIC is a modern internet transport protocol that sends data over UDP with built-in encryption and faster connection setup than TCP. Originally developed at Google and standardized by the IETF, QUIC carries several independent streams in one connection, so one lost packet does not stall everything else. QUIC is the foundation of HTTP/3.
Quality of experience (QoE)
Quality of experience (QoE) describes how good playback feels from the viewer's seat: how quickly the video begins, whether it pauses to load, and how sharp the picture stays. QoE focuses on what people actually see, not on network statistics. A stream can run on healthy servers and still deliver poor QoE on a viewer's phone.
Quality of experience score (QoE score)
A quality of experience score is a single number, usually 0 to 100, that sums up how good video playback felt to a viewer. It blends how fast the video started, how often it stalled, and how sharp it looked. One score makes it possible to compare days, regions, devices or releases at a glance instead of reading a dozen separate metrics.
Quality of service (QoS)
Quality of service (QoS) is the measurement and management of the network and delivery performance that carries video to viewers, such as throughput, packet loss, latency and server errors. QoS looks at the pipes and servers rather than the viewer's screen. Good QoS makes good quality of experience likely, but it does not guarantee it.
Quantization (video)
Quantization in video compression is the step where an encoder rounds detailed picture information into fewer possible values so it takes fewer bits to store. Quantization is the main place where lossy codecs discard detail. A higher quantization level means smaller files but more blur or blockiness. A lower level keeps more detail but produces bigger files.
R
RTMP
RTMP (Real-Time Messaging Protocol) is a protocol for sending live video and audio from an encoder to a streaming server over a steady TCP connection. Originally built by Macromedia for Flash, it is now used mainly for ingest (the first mile from the broadcaster to the platform). The platform then converts the stream into HLS or DASH for viewers, since modern browsers no longer play RTMP directly.
RTMPS
RTMPS is RTMP (Real-Time Messaging Protocol) sent inside an encrypted TLS connection, the same security layer that protects HTTPS websites. It carries live video from an encoder to a streaming platform exactly like RTMP, but nobody between the two can read or alter the stream or its stream key. It usually runs on port 443, which firewalls rarely block.
RTP
RTP (Real-time Transport Protocol) is a network protocol that carries audio and video packets in real time, adding timestamps and sequence numbers so the receiver can play them in order. It usually runs over UDP for speed. RTP is the delivery layer underneath WebRTC, RTSP and many VoIP systems, while a companion protocol called RTCP reports on quality.
RTSP
RTSP (Real Time Streaming Protocol) is a network protocol for controlling live video streams from cameras and media servers, with commands like play, pause and teardown. RTSP acts like a remote control: it sets up the session, while the video itself usually travels separately over RTP. It is the standard output of most IP security cameras (CCTV) and many surveillance systems.
Rate-distortion optimization
Rate-distortion optimization (RDO) is the method a video encoder uses to choose, for each block of a frame, the option that gives the best quality for the number of bits spent. Rate means the bits used, and distortion means the visible quality loss. RDO weighs both and picks the best balance. Stronger RDO improves compression but slows encoding.
Rebuffering / rebuffer ratio
Rebuffering is a pause during playback when the video player runs out of downloaded video and must stop to load more. Rebuffer ratio is the share of total viewing time spent stalled in this way. For example, 3 seconds of stalling during 100 seconds of watching gives a rebuffer ratio of 3 percent. Lower is better.
Rec. 2020
Rec. 2020 (ITU-R BT.2020) is the international standard that defines a wide color space and picture settings for ultra HD video, including 4K and 8K. Rec. 2020 covers far more colors than Rec. 709, especially deep greens and reds. HDR formats such as HDR10, HLG and Dolby Vision use Rec. 2020 colors. Few displays can show the full range yet.
Rec. 709
Rec. 709 (ITU-R BT.709) is the international standard that defines the color space, brightness curve and other picture settings for HD video. Rec. 709 is the default for most standard dynamic range (SDR) content on the web, TV and streaming services. Its color range covers less than Rec. 2020, which is used for HDR. Most displays and players assume Rec. 709 when no other metadata is present.
Remuxing
Remuxing is the process of taking the audio, video and subtitle streams out of one container file and placing them into a new container without re-encoding. Converting an MKV file to MP4 with the same H.264 video inside is a common example. The picture and sound stay bit-for-bit identical, so remuxing finishes in seconds.
Render quality score
A render quality score measures how sharp the picture looked relative to what the screen could show. It drops when the player has to stretch a low resolution video to fill a large screen (upscaling) or when quality keeps switching. A high score means viewers mostly saw video at or above their screen's resolution.
Rendition
A rendition, in video streaming, is one encoded version of a video at a specific resolution and bitrate, such as 720p at 3 Mbps. Several renditions of the same video together form the encoding ladder used for adaptive streaming. The player picks the rendition that best fits the viewer's screen and connection at each moment.
Rendition shift
A rendition shift is a switch between quality levels during playback: an upshift when the player moves to a higher bitrate as the network improves, a downshift when it drops to avoid buffering. Counting shifts shows how stable the connection was and how aggressively the adaptive logic reacted.
Request latency
Request latency is the time between a player asking a server for a piece of video and the first byte of the reply arriving. High latency usually points at a far-away or overloaded CDN edge rather than a slow connection, and it adds directly to startup time and seek time.
Request throughput
Request throughput is the real download speed a viewer's connection achieved while fetching video segments, usually measured in kilobits or megabits per second. It is what the player actually got, not what the ISP advertises, so it shows whether the network can carry the bitrate being sent.
Resolution
Video resolution is the number of pixels in each frame of a video, written as width by height, such as 1920 x 1080. More pixels can show more detail, so higher resolution usually looks sharper on large screens. Common video resolutions include 480p, 720p, 1080p and 4K.
Resumable upload
A resumable upload is a file upload that can pause or fail partway and then continue from where it stopped, rather than starting again from zero. The file is sent in pieces and the server remembers which pieces have arrived. This matters most for large video files and for users on unstable mobile networks.
S
SCTE-35
SCTE-35 is a broadcast standard for placing cue markers inside a video stream that signal where ad breaks and program changes start and end. The markers travel with the video, for example in MPEG-TS or as tags in an HLS playlist. Downstream systems read them to know exactly when to swap in an ad or a local program.
SRT (Secure Reliable Transport)
SRT (Secure Reliable Transport) is an open-source protocol for sending live video over unpredictable networks like the public internet with low delay. It runs on UDP and resends lost packets quickly, within a time window the sender sets. SRT also supports AES encryption. Created by Haivision, it is widely used as a modern alternative to RTMP for contribution (sending feeds from the field to the studio or cloud).
SRT subtitle file
An SRT subtitle file (SubRip Subtitle) is a plain text file that stores subtitles as numbered lines with start and end timestamps. It is the most widely supported subtitle format across video editors, players and social platforms. An SRT file is unrelated to the SRT streaming protocol (Secure Reliable Transport), which sends live video over networks.
SSAI (server-side ad insertion)
SSAI (server-side ad insertion) stitches video ads into a stream on the server, so the player receives content and ads as one continuous stream. A manifest manipulator reads break markers such as SCTE-35, asks an ad decision server which ads each viewer should get, and rewrites that viewer's playlist to point at ad segments encoded to match the content. Because ads arrive from the same source as the show, transitions are smooth and ad blockers have little to remove. Tracking moves to the server.
SSIM
SSIM (structural similarity index) is a video and image quality metric that compares a compressed frame with its original by looking at brightness, contrast and structure, rather than raw pixel differences. Scores usually top out at 1, which means identical. SSIM tends to match human judgment better than PSNR, because people notice structural changes like blur and blockiness more.
SVOD
SVOD (subscription video on demand) gives subscribers access to a whole video catalogue for a recurring fee, and removes access when the subscription ends. For the video stack, that means every play checks an active subscription before a playback token is issued, licensed titles carry DRM whose licenses expire with the subscription, the same catalogue plays on phones, browsers and TVs, and playback quality is measured per subscriber so problems surface before viewers cancel.
Scene detection
Scene detection is the automatic process of splitting a video into scenes, meaning groups of shots that share the same location, time or storyline. It often builds on shot boundary detection, then groups related shots together based on visual similarity, audio and content. The result is a list of scene start and end times that software can use for navigation, editing or analysis.
Secure video hosting
Secure video hosting is video hosting that controls not just who can start a video, but what happens to it after delivery. Videos are stored privately and played only with signed, expiring access. Streams arrive in encrypted segments rather than as one downloadable file, and DRM can release decryption keys only to licensed players on trusted devices. Watermarks mark the picture so copies show their origin. Public platforms are the opposite case: anyone with a link can usually watch, embed or save the video.
Seek latency
Seek latency is how long playback takes to resume after a viewer jumps to a new point in the video. Long seek waits usually mean the CDN does not support byte-range requests well or the segments are long. It is also called jump latency.
Semantic video search
Semantic video search is a way of finding videos, or moments inside them, based on the meaning of a query rather than exact keyword matches. It converts both the query and the video content into embeddings (numbers that capture meaning), then returns the closest matches. A search for "how to reset a password" can find a clip where the speaker says "recover your login".
Shaka Player
Shaka Player is an open-source JavaScript video player library from Google that plays MPEG-DASH and HLS streams in web browsers. Shaka Player has strong support for DRM through Encrypted Media Extensions, offline downloads, multiple audio tracks and subtitles. It works on desktop browsers, Android, many smart TVs and casting receivers such as Chromecast.
Shoppable video
Shoppable video lets viewers buy products shown on screen without leaving the video, by tapping a product tag, card or link timed to the moment the item appears. Technically, it is a video stream plus a timeline of product cues: timestamps paired with product IDs, delivered as a sidecar file or as timed metadata inside the stream. The player reads the cues against the playhead, draws the overlay, and hands each tap to the store's product page or cart. It works for on-demand clips, vertical feeds and live shopping streams.
Short-form video
Short-form video is video content that typically runs from a few seconds up to about three minutes, designed to be watched quickly, often on mobile phones. It is usually vertical, fast-paced and viewed in a scrolling feed. TikTok, Instagram Reels and YouTube Shorts are leading examples of short-form video.
Shot boundary detection
Shot boundary detection is the automatic process of finding the exact points in a video where one camera shot ends and the next begins. It detects hard cuts, where the picture changes instantly, and gradual transitions such as fades, dissolves and wipes. Each detected boundary is reported as a timestamp or frame number, giving a precise map of how the video was edited.
Signed URL
A signed URL is a link that carries a cryptographic signature, and usually an expiry time, so a server can tell it was issued by you and has not been changed. For video, the server or CDN checks the signature before it returns a manifest or segment, and refuses the request if any part of the link was edited or the expiry has passed. Signed URLs let a video stay at a stable address while access to it is granted to one viewer, for a limited time.
Simulcast
Simulcast in live streaming is broadcasting the same live video to several destinations at the same time, for example your own app plus YouTube and Facebook Live. Usually the encoder sends one stream to a cloud service, which copies it to each platform's ingest URL. In WebRTC the word has a second meaning: a sender encodes the same camera at several qualities at once so a server can forward the right one to each participant. Either way, one live source reaches viewers through more than one path.
Simulive
Simulive is playing a pre-recorded video at a scheduled time as if it were a live broadcast. Instead of an encoder capturing a camera, a playout system reads the finished file and outputs it as a live stream, so every viewer joins the same moment and nobody can skip ahead. Chat, polls and Q&A can run alongside, and the session can switch to a real live feed at the end. Simulive pairs the reliability of recorded content with the shared timing of a live event.
Smooth Streaming
Smooth Streaming is Microsoft's adaptive bitrate streaming protocol, launched for Silverlight, that delivers video over HTTP in small fragments at several quality levels. It uses fragmented MP4 files and an XML manifest. Smooth Streaming helped prove that HTTP-based adaptive streaming works at scale, and its ideas shaped MPEG-DASH, but it is now a legacy format.
Speaker diarization
Speaker diarization is the process of splitting an audio or video recording into segments according to who is speaking, answering the question "who spoke when?". It labels each segment with a speaker tag such as Speaker 1 or Speaker 2, without necessarily knowing their names. Combined with a transcript, diarization turns a block of text into a readable conversation.
Speech-to-text
Speech-to-text, also called automatic speech recognition or ASR, is software that converts spoken words in audio or video into written text. A neural model turns short slices of sound into likely words and picks the most probable sentence, often helped by a language model, and the system adds punctuation and a timestamp for each word. In a video pipeline, speech-to-text produces the raw transcript that captions, search, chapters, summaries and translations are all built from, so any error it makes travels into every one of them.
Sprite sheet (storyboard)
A sprite sheet, also called a storyboard, is a single image that packs many small frames, taken at regular intervals through a video, into a grid. A small index file, usually WebVTT or JSON, maps each time range to one tile's position in the grid. When a viewer hovers over or drags along the seek bar, the player crops the matching tile and shows it as a preview. One image replaces hundreds of separate thumbnail requests, so seek previews appear as soon as the pointer moves.
Stability score
A stability score measures how smoothly a video played once it started. It is built from how many times playback paused to buffer and how much of the viewing time those pauses took up. A high score means the picture kept moving; a low score means viewers sat through spinners.
Startup score
A startup score grades how quickly a video began playing after the viewer pressed play. Short waits score close to 100 and the score falls as the wait grows. It turns a raw startup time in milliseconds into a number that can be compared across titles, devices and days.
Stream key
A stream key is a unique secret code that tells a live streaming server which channel or event an incoming video feed belongs to. You paste it, together with a server URL, into encoder software such as OBS Studio. Anyone who has the key can broadcast to that channel, so it should be kept private like a password.
Subtitles
Subtitles are on-screen text that shows a video's spoken dialogue, usually translated into another language, for viewers who can hear the audio but do not understand it. Unlike closed captions, subtitles typically leave out sound effects and speaker cues. Subtitles are stored as separate text files, such as SRT or WebVTT, and the player shows them in sync with the video.
T
TS file
A TS file is a video file that uses the MPEG transport stream format, a container built to carry video and audio over networks where data can be lost. In streaming, TS files are often the small segments, a few seconds each, that HLS players download one after another. TS files are also common in broadcast TV and DVR recordings.
TTML
TTML (Timed Text Markup Language) is an XML-based W3C standard for storing subtitles and captions with precise timing, styling and positioning. It supports rich formatting such as colors, fonts and placement on screen. Profiles such as IMSC and SMPTE-TT are widely used for broadcast, DASH streaming and professional subtitle delivery.
TUS protocol
The TUS protocol is an open, HTTP-based standard for resumable file uploads. A client creates an upload on the server, sends the file in pieces, and asks the server how much has arrived if the connection breaks. Ready-made TUS libraries exist for browsers, mobile apps and servers, so teams do not need to design resumable uploads themselves.
TVOD
TVOD (transactional video on demand) is a business model where viewers pay for individual titles, either to rent them for a limited time or to buy them to keep. Movie rentals on Apple TV or Amazon Prime Video are common examples. There is no subscription: customers pay only for what they choose to watch.
Thumbnail
A video thumbnail is a still image that stands in for a video before it plays. It appears in catalogue rows, search results and social link previews, and on the player itself as the poster. Most thumbnails are a frame captured from the video at a chosen time, then resized and compressed for each place it is shown. On a video API, thumbnails are usually generated on request from a URL, and for a private video they need the same access control as the stream.
Time to first frame (TTFF)
Time to first frame (TTFF) is the time it takes from a playback request until the first video frame appears on the viewer's screen. TTFF is a precise, technical form of startup time measured inside the player. It is usually recorded in milliseconds and tracked both as an average and at slower percentiles, such as the 95th percentile.
Token authentication
Token authentication, in video streaming, is an access method in which your backend issues a short-lived signed token that the player must present before a stream will load. The token, usually a JSON Web Token (JWT), names the video it opens and when it expires, and may carry other claims such as a viewer ID. The video service checks the signature and expiry on each request and rejects anything missing, altered or out of date, so entitlement decided in your system is enforced at the stream.
Total startup time
Total startup time is the whole wait from a page starting to load to the first video frame appearing. It adds page load time, player initialization time and video startup time. It is the number closest to what a viewer actually feels, since they do not care which part was slow.
Transcoding
Transcoding is the process of decoding a video that is already compressed and encoding it again with a different codec, resolution, bitrate or frame rate. Streaming platforms transcode almost every upload, because files arrive already compressed, often as one large MP4 or an editing master such as ProRes. Transcoding turns that single file into a bitrate ladder: the same video at several qualities, so the player can choose the version each screen and connection can carry. It is the step between a raw upload and something that streams.
Transmuxing
Transmuxing is the process of moving audio and video streams from one container format into another without re-encoding them. For example, video arriving over RTMP can be transmuxed into HLS segments while the compressed picture stays exactly the same. Because nothing is decoded, transmuxing is fast, cheap and causes no loss in quality.
Transrating
Transrating is the process of changing a video's bitrate while keeping the same codec and, usually, the same resolution. A 10 Mbps H.264 stream might be transrated to 4 Mbps H.264 so it fits a slower network. Transrating is a narrow form of transcoding focused only on data rate.
Trick play
Trick play is the set of playback features that let viewers move through a video in ways other than normal play, such as fast forward, rewind, scrubbing and frame stepping. On streaming services it often includes thumbnail previews while seeking. Trick play relies on keyframes, special low-frame-rate tracks or image sprite sheets to respond quickly.
Two-pass encoding
Two-pass encoding is a way of compressing video where the encoder analyzes the whole video in a first pass, then uses that information to distribute bits more efficiently in a second pass. Because the encoder already knows where the hard scenes are, the second pass can meet a set file size or average data rate with better quality than one pass alone. It roughly doubles analysis time and cannot be used for live streams.
U
UGC (user-generated content)
UGC (user-generated content) video is video created and uploaded by everyday users of a platform rather than by the platform itself or professional studios. Examples include social clips, product reviews, gaming highlights and community posts. In video infrastructure, UGC means unpredictable uploads in many formats, sizes and quality levels that must be processed and checked at scale.
Ultra-low latency
Ultra-low latency is a class of live video delivery where the delay from camera to viewer is under about one second. Reaching this level usually requires real-time protocols such as WebRTC rather than standard segment-based HLS or DASH. The result feels close to a video call, where people can talk and react with almost no lag.
Upscaling
Upscaling, in playback analytics, is when the player stretches a video rendition to fill a screen larger than the video's own resolution. A 720p stream on a 4K TV is heavily upscaled and looks soft. Analytics report it as the percentage of watch time spent upscaled and the maximum stretch reached.
User engagement data
User engagement data is the record of how viewers interact with video content, including plays, pauses, seeks, watch time, completions, shares and drop-off points. It is collected from video players and apps, often through player beacons. Analyzed together, user engagement data shows what content people value and where they lose interest.
V
VAST
VAST (Video Ad Serving Template) is an IAB standard XML format that ad servers use to tell a video player which ad to play and how to track it. A VAST tag holds the ad's media file links, duration, click-through URL and tracking URLs for events like impressions and completions. It lets any compliant player work with any compliant ad server.
VBR (variable bitrate)
VBR (variable bitrate) is an encoding mode where the data rate rises for busy, detailed moments and drops for calm, simple ones. VBR delivers better overall quality than constant bitrate at the same average file size. Encoders usually cap VBR with a maximum bitrate so streams stay deliverable.
VMAF
VMAF (Video Multimethod Assessment Fusion) is a video quality metric, developed by Netflix with university researchers, that predicts how good a compressed video looks to a human viewer. VMAF compares the encoded video with its original source and gives a score from 0 to 100. Higher scores mean the encoded version looks closer to the original.
VMAP
VMAP (Video Multiple Ad Playlist) is an IAB standard XML format that tells a video player when to show ad breaks across a whole video. It lists each break, such as a pre-roll at the start or a mid-roll at ten minutes, and points to the VAST tag that supplies the ad for each slot. VMAP handles timing, while VAST handles the ads themselves.
VP8
VP8 is an open, royalty-free video codec that Google released in 2010 after acquiring On2 Technologies. VP8 offers compression roughly similar to H.264 and was the original video codec for the WebM format. VP8 is also one of the mandatory video codecs for WebRTC, the technology behind browser-based video calls. For streaming, VP9 and AV1 have largely replaced it.
VP9
VP9 is an open, royalty-free video codec developed by Google and released in 2013 as the successor to VP8. VP9 compresses video roughly as well as HEVC and much better than H.264. YouTube uses it widely, and it plays in Chrome, Firefox, Edge and on Android devices. VP9 is usually packaged in WebM or MP4 files.
VPAID
VPAID (Video Player-Ad Interface Definition) is an older IAB standard that let interactive video ads run code inside a video player and talk to it. It made clickable, game-like and measurable ads possible, but it also slowed players and raised security concerns. The IAB has deprecated VPAID in favor of newer approaches such as SIMID and OMID.
Vertical video
Vertical video is video filmed and displayed in portrait orientation, usually with a 9:16 aspect ratio, so it fills a phone screen held upright. It is the standard format for TikTok, Instagram Reels, YouTube Shorts and Stories. Vertical video avoids black bars on phones, unlike horizontal 16:9 video viewed in portrait mode.
Video AI
Video AI is the use of machine learning models to analyze, understand, edit or generate video content automatically. It covers tasks such as spotting objects and faces, turning speech into text, writing summaries, flagging unsafe material and creating new clips from prompts. Instead of a person watching every minute, software reads the pictures, sound and on-screen text and returns structured results.
Video API
A video API is a set of HTTP endpoints that lets developers add video to their own product through code: upload, encoding, storage, streaming, playback, live broadcasts, analytics and AI features. The app sends a request, such as "create an asset from this file", and the service runs the pipeline and returns IDs, URLs and webhook events. Instead of operating encoders, packagers and CDNs, the team writes integration code and keeps control of the user experience, access rules and viewing data.
Video CDN
A video CDN is a content delivery network set up for streaming video. It keeps copies of video segments and manifests on edge servers, called points of presence, close to viewers, so each request travels a short distance instead of going back to the origin server. Video traffic is heavy and continuous: a player asks for a new segment every few seconds for the whole view, and a live event sends its entire audience to the same segments at once. A video CDN is tuned for that pattern.
Video CMS
A video CMS (content management system) is the editorial layer for a video library. It stores each video's title, description, thumbnails, tags, captions, categories, release dates and access rules, and lets editors search, update, schedule and publish without writing code. A video CMS describes and organizes the content. The encoding, storage and streaming usually happen in a video platform or API underneath, which the CMS references by ID. Some products bundle both layers, and many teams keep them separate.
Video OCR
Video OCR (optical character recognition) is the process of detecting and reading text that appears on screen in a video, such as burned-in captions, slides, signs, scoreboards, name tags and product labels. The system samples frames, finds text regions and converts them to machine-readable text with timestamps. The result makes on-screen words searchable even when nobody says them aloud.
Video RAG
Video RAG (retrieval-augmented generation) is a technique in which an AI assistant answers questions by first retrieving relevant video segments, transcripts or frames from a library and then using them as context for a language model's answer. Grounding answers in real content reduces made-up responses. Good systems also cite the source video and timestamp so users can check the answer.
Video SDK
A video SDK (software development kit) is a package of code you install in an app or backend to add video features without writing them from scratch. Client SDKs run on a device, in a browser, on iOS or Android, or inside a cross-platform framework, and handle playback, uploads, live capture or analytics. Server SDKs wrap a provider's video API in a language such as Node or Python. An SDK runs inside your code; the API it calls runs on the provider's servers.
Video SEO
Video SEO is the practice of making videos easier for search engines to find, understand and show in results such as Google search, video tabs and AI answers. It covers titles, descriptions, transcripts, thumbnails, structured data and video sitemaps. Good video SEO also needs a crawlable page and a video file that search engines can access.
Video analytics
Video analytics, in streaming, is the measurement of how each video is delivered and watched. A small SDK in the player sends events such as play, playing, pause, seek, stall and error, and a backend groups them into views and turns them into metrics: startup time, rebuffering, bitrate, errors and watch time. Those metrics are broken down by device, country, network, CDN and title, so teams can see where playback fails and what people watch. The same phrase also names computer vision on camera footage, a separate field.
Video chapters
Video chapters are labeled sections of a video, each with a title and start time, that let viewers jump straight to the part they want. Chapters usually appear as markers on the player's progress bar or as a clickable list. Chapters can be written by hand or generated automatically by AI that detects topic changes in the transcript.
Video codec
A video codec is a method for compressing moving pictures into far fewer bits and then rebuilding them on the viewer's screen. Raw video is enormous, so a codec removes detail the eye barely notices and repeats between frames. H.264, HEVC, VP9 and AV1 are widely used video codecs. Each one trades file size, image quality, encoding speed and device support differently.
Video completion rate
Video completion rate is the percentage of video starts in which the viewer watches to the end, or close to it. If 1,000 people start a video and 400 finish it, the completion rate is 40 percent. Analytics tools usually count a view as complete at 90 or 95 percent of the length, because many viewers skip end credits.
Video compression
Video compression shrinks raw video into far fewer bytes so it is cheaper to store and quicker to stream. It works by removing information that repeats within each frame and between neighboring frames, plus fine details the eye barely notices. A codec such as H.264 or AV1 performs the compression. Good compression keeps quality high while cutting bandwidth and storage costs.
Video embed
A video embed is a short piece of code that drops a hosted player into a web page, app or document, so people can watch right where they are. Most embeds use an iframe or a JavaScript player tag. The video file stays on the hosting service, which handles streaming, while the page just displays the player.
Video embeddings
Video embeddings are lists of numbers (vectors) that an AI model produces to represent the meaning of a video or a segment of it, including its visuals, speech and text. Videos with similar content end up with similar vectors, so a computer can compare them by measuring the distance between them. Embeddings are typically stored in a vector database for fast lookup.
Video encoding
Video encoding is the process of compressing raw or lightly compressed video into a compact stream using a codec such as H.264, HEVC or AV1. The encoder stores only what changes between frames, discards detail the eye is unlikely to notice, and aims at a target bitrate or quality level. The result is wrapped in a container such as MP4, or split into HLS or DASH segments for streaming. Encoding settings decide how sharp a video looks, how much data each minute costs to store and deliver, and which devices can play it.
Video encryption
Video encryption is the process of scrambling video data with a secret key so that only players with the right key can decode and play it. Encryption can protect video while it is stored, while it travels over the internet and during playback. In streaming, common approaches include AES-128 encryption of HLS segments and DRM-based encryption using common encryption (CENC).
Video engagement
Video engagement is a measure of how actively viewers interact with video content, such as how long they watch, whether they finish, and whether they replay, share, comment or click. Engagement shows whether content holds attention, not just whether it was opened. It is usually tracked through several metrics together, since no single number captures it fully.
Video fingerprinting
Video fingerprinting builds a short, unique signature from what a video looks and sounds like, so copies can still be recognized after they are cropped, re-compressed or filmed off a screen. Unlike a watermark, a fingerprint is not added to the video. It is calculated from the content itself and compared against a database of known videos.
Video hosting
Video hosting is the infrastructure that takes a video file and makes it stream well to any device. It covers ingest, encoding into several qualities, packaging into streaming formats, storage, delivery through a CDN (a network of servers close to viewers) and playback. Without it, a site hands every viewer one large file and hopes the connection keeps up. Hosting can be public, where anyone can watch, or private, where your own app decides who plays each video.
Video indexing
Video indexing is the process of analyzing a video and building a structured, time-coded record of its contents, such as spoken words, on-screen text, objects, faces and scenes, so it can be searched quickly. Like the index at the back of a book, a video index tells software where each topic or item appears. Indexes are usually built when a video is uploaded.
Video metadata
Video metadata is the descriptive and technical information attached to a video, such as its title, description, tags, duration, resolution, codec, creation date and captions. Some metadata is added by people, some is read from the file, and some is generated by AI. Metadata makes videos searchable, sortable and easy to manage.
Video moderation
Video moderation is the process of screening video for material that breaks a platform's rules or the law, such as nudity, sexual content, violence, hate symbols or abusive language, and then acting on it. Automated systems sample frames, transcribe the audio and run classifiers that score each signal by category and timestamp. The platform compares those scores with its own policy to publish, restrict, blur, remove or send a video to a human reviewer. Most systems combine AI screening for volume with people for borderline calls and appeals.
Video monetization
Video monetization is earning money from video through subscriptions, ads, one-off purchases or in-video shopping. Each model is enforced in the video stack, not just on the web page. Paid models need entitlement checks and signed playback tokens, with DRM on licensed content. Ad models need break markers and an ad insertion path, either in the player or on the server. Commerce needs a player that can show product overlays. All of them need per-view playback data, because a view that fails earns nothing.
Video on demand (VOD)
Video on demand (VOD) is video that viewers can watch whenever they choose, rather than at a scheduled broadcast time. The content is recorded, stored and streamed when a viewer presses play, with options to pause, rewind and skip. Streaming movies, online courses and recorded webinars are all examples of VOD.
Video packaging
Video packaging is the step that prepares encoded video for streaming by splitting it into segments, wrapping them in the right container and writing the manifest files a player needs. The same encoded video can be packaged as HLS for Apple devices and DASH for others. Packaging is also where encryption for DRM is usually applied.
Video paywall
A video paywall is access control that refuses to play a video until the viewer has the right entitlement: a purchase, an active subscription or an enrolment. A paywall holds only when it is enforced at playback, not just on the page. Your backend checks the viewer's rights and hands the player a signed token that expires quickly, and the video platform refuses any request without one. Premium titles add DRM on top. Many paywalls leave a free preview in front of the gate.
Video piracy
Video piracy is the unauthorized copying, sharing, streaming or selling of video content without permission from the rights holder. It includes downloading and reuploading files, screen recording, sharing login credentials and illegally restreaming live events. Video piracy takes revenue from creators and distributors, and pirated copies can carry malware or poor quality.
Video player
A video player is the software that fetches a stream, decodes it and draws it on screen, with controls such as play, pause, seek, volume and captions. On the web it usually wraps the browser's video element and feeds it segments through Media Source Extensions, the browser API that lets JavaScript hand video data to the decoder. Native apps build on platform players such as AVPlayer on iOS or ExoPlayer on Android. A streaming player also picks the quality level, handles DRM and reports what happened during each view.
Video quality metrics
Video quality metrics are measurements used to judge how good a video looks or how well it plays, covering both picture quality and playback performance. Picture metrics such as VMAF, PSNR and SSIM compare an encoded video with its source. Playback metrics such as startup time and rebuffer ratio describe what viewers actually experience.
Video search
Video search is the ability to find videos, or exact moments inside them, by typing a query. Basic video search matches titles, tags and descriptions. Deeper search also indexes what is said and shown: transcripts, on-screen text, people, objects and scenes, each with a timestamp, so a result can open the player at the right second. Search can match exact words, match meaning through embeddings, or combine both. Inside a product, it lets someone ask a question and land on the minute of a lecture, call or recording that answers it.
Video segment / chunk
A video segment (or chunk) is a short piece of a video, usually 2 to 10 seconds long, that a streaming player downloads one after another. Protocols such as HLS and DASH split each rendition into these segments and list them in a manifest file. Small pieces let the player switch quality between segments and start playback quickly.
Video sitemap
A video sitemap is an XML list that tells search engines which videos a site has and describes each one: its title, summary, thumbnail, length and where the file lives. It helps search engines like Google discover videos they might miss while crawling. A video sitemap can be a separate file or part of an existing sitemap.
Video start failure
Video start failure is a playback attempt where the viewer presses play but the video never begins, usually because of an error. Common causes include a broken or expired link, a DRM license error, a network failure or a format the device cannot play. Teams often track it as a percentage of all playback attempts, called the video start failure rate.
Video startup time
Video startup time is the wait between a viewer asking for playback and the first frame appearing on screen. It covers fetching the manifest, which lists the available renditions, any DRM license request, downloading the first segments and decoding the first frame. Many teams also track the time to load the page and initialize the player, because viewers feel those waits too. Startup time is the first quality signal in every view: viewers who wait too long leave before anything plays.
Video streaming
Video streaming is the delivery of video over the internet in small pieces that play as they arrive, so viewers can start watching without downloading the whole file. The video is encoded, split into segments, sent through a CDN and reassembled by a player. Streaming can be live, as events happen, or on demand from a stored library.
Video streaming API
A video streaming API is a programming interface that delivers video to viewers as adaptive streams instead of single files. It takes a video, encodes it into several qualities, packages them into short segments with a manifest in formats such as HLS or DASH, and serves them through a CDN. The app gets back a playback URL, and the player uses it to start fast and switch quality as each viewer's network changes. Most streaming APIs cover on-demand video, and many add live.
Video summarization
Video summarization is the use of AI to condense a long video into something much shorter: a written summary of its key points, or a highlight reel of its most important moments. Text summaries are usually built from the transcript, with visual signals such as on-screen text and scene changes adding context. Good summaries keep timestamps, so each point can link back to the moment it came from. The goal is to let someone understand what a video covers, and decide whether to watch, without playing all of it.
Video tagging / metadata extraction
Video tagging, also called metadata extraction, is the process of adding descriptive labels to a video, such as topics, people, objects, locations, brands and moods, either automatically or by hand. AI models generate tags by analyzing frames, audio and on-screen text, often with timestamps. The tags become searchable metadata that helps organize, recommend and monetize a video library.
Video transcoding API
A video transcoding API converts video from one encoding to others through code. You send a source file or URL, and the service decodes it and re-encodes it into new codecs, resolutions, bitrates or containers, usually as a bitrate ladder for adaptive streaming and sometimes as MP4 downloads. Uploads from phones, cameras and editing tools arrive in every format; transcoding makes them compact and playable on every device. The API runs the encoder fleet, queues and retries, and reports results by webhook.
Video transcription
Video transcription is the process of turning the speech in a video into written text, with timestamps that show when each word or line was said. It can be done by AI speech-to-text, by human transcribers, or by AI followed by a human review pass. A finished transcript has punctuation, timestamps and often speaker labels. It can be read as a document, searched, or cut into caption files such as WebVTT and SRT. For a lecture or a news interview, the transcript is often a second product that sits under the video.
Video translation
Video translation is the process of converting a video's spoken words and on-screen text from one language into another so new audiences can understand it. It can produce translated subtitles, a dubbed audio track or both. Modern workflows usually transcribe the original speech, translate the text with machine translation and then either display it as subtitles or generate new speech.
Video understanding
Video understanding is the ability of an AI system to interpret what is happening in a video, including the actions, people, objects, speech and how they change over time. It goes beyond labeling single frames by tracking events across a sequence, much like following a plot rather than glancing at a photo. Modern systems often combine vision and language models to answer questions about a clip.
Video upload API
A video upload API lets an app move video files from a user's device, or from cloud storage, into a video service for processing. The backend requests an authenticated upload URL, the client sends the file straight to the service, often in resumable chunks, and the service reports progress and fires a webhook when the video is ready to play. A good upload API survives large files, weak mobile networks and dropped connections, and keeps the bytes off your own servers.
Video watermark
A video watermark is a mark placed on or in video frames to show ownership, keep a brand on the picture or trace where a copy came from. Visible watermarks are logos, text or labels drawn over the picture, usually in a corner, either burned into the encoded video or drawn by the player at playback. Invisible, or forensic, watermarks are patterns hidden in the picture that viewers cannot see but detection software can read later, sometimes unique to each viewer so a leak points to one account.
Video.js
Video.js is a popular open-source web video player framework that builds a customizable interface on top of the HTML5 video element. Video.js includes its own HLS and DASH playback engine, a theming system and a large plugin ecosystem for ads, analytics and quality menus. It gives developers consistent controls across browsers without designing a player from scratch.
View (video view)
A view, in video analytics, is one attempt by one viewer to watch one video, from pressing play to leaving. Everything else is measured inside a view: how long it took to start, how often it buffered, how far the viewer got. A viewer who watches three videos creates three views.
View session
A view session groups several views by the same viewer in one sitting, such as three episodes watched back to back. It links what happened in the player with what happened at the CDN and on the server, so a problem can be traced end to end rather than inside one video.
Viewer ID
A viewer ID is a label that identifies the same person across many views, so analytics can tell one viewer watching ten videos from ten viewers watching one. It is usually either an ID the app passes in from its own login system or an anonymous ID the analytics SDK stores in the browser.
Vision language model (VLM)
Vision language models (VLMs) are AI models trained on both images and text, so they can look at a picture or video frame and describe, answer questions about or reason over what they see in plain language. A VLM typically pairs an image encoder with a large language model. Given sampled frames from a video, a VLM can write captions, list objects or explain a scene.
W
WHEP
WHEP (WebRTC-HTTP Egress Protocol) is a standard way for a player to receive a live stream from a server using WebRTC, set up with a single HTTP request. It is the playback partner of WHIP, which handles sending video in. WHEP gives viewers delay of under a second while letting any compatible player work with any compatible server, instead of relying on custom code for each vendor.
WHIP
WHIP (WebRTC-HTTP Ingestion Protocol) is a standard way to send a live stream into a streaming platform using WebRTC, set up with a single HTTP request. Before WHIP, each platform had its own custom signaling (the handshake that starts a WebRTC session). WHIP lets any compatible encoder or browser publish to any compatible server, much like RTMP does for traditional streaming.
Watch time
Watch time is the total amount of time viewers spend actually playing a video or set of videos, usually measured in minutes or hours. Paused time and time spent buffering are typically excluded. Watch time can be totaled across an audience, or averaged per viewer or per session to show how well content holds attention.
WebM
WebM is an open, royalty-free video container format designed for playback in web browsers. It is based on Matroska (MKV) and usually carries VP8, VP9 or AV1 video with Opus or Vorbis audio. Most modern browsers play WebM natively, though support on older Apple devices is limited.
WebRTC
WebRTC (Web Real-Time Communication) is an open standard, built into modern browsers, for sending audio, video and data between devices with delay usually under one second. It powers video calls in the browser without plugins. WebRTC uses UDP-based transport, built-in encryption and techniques to connect devices across home routers and firewalls.
WebVTT
WebVTT (Web Video Text Tracks) is a text-based subtitle and caption format designed for HTML5 video in web browsers. Each cue has a start time, end time and text, with optional styling and positioning. WebVTT is also the standard subtitle format for HLS streaming, and it can hold chapter markers and thumbnail preview data.
Webhook
A webhook is an automatic message one system sends to another over HTTP when a specific event happens, such as a video finishing processing. Instead of an app repeatedly asking a server whether something is ready, the server notifies the app at a set URL. It is like getting a text when your order ships instead of checking the tracking page.
White-label video platform
A white-label video platform is video infrastructure that runs behind your own brand, so viewers see your logo, colors, domain and player design and never the vendor's. The provider operates encoding, storage, delivery and playback, and you present it as part of your product. White-label can mean a finished portal or app that the vendor runs with your branding applied, or a set of APIs and an unbranded player that you build into your own site and apps.
Widevine
Widevine is Google's DRM (digital rights management) system for protecting video, used in Chrome, Android, Android TV, Chromecast and many smart TVs and set-top boxes. Widevine has security levels, from L1, where decryption happens in protected hardware, to L3, which runs in software. Studios often limit HD and 4K playback to L1 devices.
Y
YCbCr
YCbCr is the digital color format used by almost all compressed video, storing one brightness value (Y) and two color-difference values (Cb for blue and Cr for red). YCbCr lets codecs apply chroma subsampling, keeping full brightness detail while reducing color detail. H.264, HEVC and AV1 video are encoded in YCbCr. It is closely related to YUV.
YUV
YUV is a way of describing video color that separates brightness (luma, the Y) from color information (chroma, the U and V). Splitting them lets codecs compress color more heavily than brightness, which saves space without obvious loss. In digital video, the term YUV is often used loosely to mean YCbCr. Raw, uncompressed video frames are commonly stored in YUV formats.