Modern video codecs only store full pictures occasionally. Most frames are described as "the previous frame, but with these differences" — which compresses extremely well because adjacent frames are usually 95% identical. The full frames are called keyframes (or I-frames); the partial ones are P-frames and B-frames.
A GOP is everything between two keyframes. Inside a GOP, every frame depends on earlier frames in the same GOP. Across GOPs, there is no dependency — you can start decoding cleanly at any keyframe. This is why video players can't seek to arbitrary timestamps; they jump to the nearest keyframe and play forward.
A short GOP (1-2 seconds) gives fine-grained seeking and faster recovery from packet loss, at the cost of compression efficiency (more keyframes = more bytes). A long GOP (10 seconds) compresses better but seeks coarser. YouTube sits at 5 seconds — a reasonable balance.
Common questions
Why do clips of YouTube videos start at slightly different timestamps than I asked?
Is GOP the same as a keyframe interval?
Related terms
Keyframe (I-frame)
A keyframe (I-frame) is a video frame that stores a complete picture, independent of any other frames.
FPS (frames per second)
FPS (frames per second) is the number of distinct still images displayed per second of video.
Codec
A codec is the algorithm that encodes (compresses) and decodes raw audio or video into a smaller stream.
Bitrate
Bitrate is the amount of data a video or audio stream carries per second, measured in bits per second (bps) or kilobits (kbps) and megabits (Mbps).
VidPickr is a free, browser-based YouTube downloader. Every term in this glossary either describes how YouTube delivers video or why your downloads behave the way they do. Try the downloader →