MPEG

In short: A family of standards for compressing video and audio data, named after the “Moving Picture Experts Group” that develops them.

In more detail: Uses lossy compression — video frames aren’t stored individually, but as differences relative to each other, which brings huge size advantages over uncompressed video. Well-known offshoots: MP3 (audio) and MP4 (video container).

In Depth

The key trick of video compression is that consecutive frames in a video usually differ only slightly — instead of storing every frame individually like a standalone JPEG, MPEG only records a complete reference image (“keyframe” or I-frame) at regular intervals and, in between, only stores the differences from the previous (and sometimes following) frame — this saves enormous space, because in most video scenes only a small area of the image actually changes per frame.

It’s important to distinguish between the codec (the actual compression algorithm, e.g. H.264/AVC, H.265/HEVC, AV1) and the container format (the wrapper that bundles compressed video/audio together with metadata, e.g. MP4, MKV, WebM) — an MP4 container can theoretically contain various codecs, even though H.264 is the most common in MP4 files. Newer codecs like H.265 or AV1 achieve considerably smaller files at the same perceived quality than older MPEG generations, but need more computing power to encode/decode.

The MPEG standard family

The Moving Picture Experts Group has published several numbered standards since the early 1990s: MPEG-1 (1993, among other things the basis for MP3 audio and VideoCDs), MPEG-2 (1995, the basis for DVDs and digital television), MPEG-4 (from 1998, a very broad family of standards, which produced, among other things, the H.264/AVC video codec and the MP4 container format), and newer developments like MPEG-H (including H.265/HEVC). Each generation considerably improved compression efficiency — H.265 needs about half the data rate of its predecessor H.264 at comparable quality.

Streaming and adaptive bitrate

Modern video streaming services also use adaptive bitrate methods (e.g. HLS or MPEG-DASH): a video is encoded into several quality levels, and the player switches live between these levels depending on available network bandwidth — automatically dropping to lower resolution on a poor connection, without interrupting playback. This segmentation into short, independently playable chunks (typically 2-10 seconds) is one reason modern streaming usually manages without a full buffer-before-start, despite fluctuating internet connections.

Licensing questions

An often-overlooked aspect: many MPEG codecs (especially H.264 and H.265) are patent-protected and require licence fees for commercial use in hardware/software, managed through patent pools like MPEG LA. This has driven the development of royalty-free alternatives like AV1 (from the Alliance for Open Media, including Google, Netflix, Mozilla), which is now supported by most major browsers and platforms and often produces even smaller files than H.265 at the same quality.

See also: JPEG, Compressing, Decompressing