A viewer watching a live stream sees events some seconds after they happen. The delay is assembled from several stages, and most of it exists to protect the picture from network problems.
Encoding takes time before anything is sent
Video must be compressed before transmission, and compression works by comparing frames against each other, which requires the encoder to hold several frames at once.
Higher compression quality means examining more frames and more options, so the settings that produce a better picture at a given bandwidth also add delay.
This portion is small, usually well under a second, but it establishes that the stream is already behind before it has left the broadcaster's machine.
Buffering absorbs network variation
Data does not arrive at a constant rate. Packets are delayed and occasionally lost, and a player that displayed frames the instant they arrived would stall constantly.
Holding several seconds of video in advance lets playback continue smoothly through interruptions, because the buffer supplies frames while the network recovers.
This is the largest and most adjustable part of the delay, which is why low-latency modes work primarily by shrinking it and accept more frequent interruptions.
Distribution adds hops
Streams are sent to an ingest server, transcoded into several quality levels, and copied to distribution servers near viewers. Each step adds time.
Transcoding is necessary because viewers have very different connections, and offering one quality would exclude a large share of the audience.
The copies also protect the broadcaster, since thousands of viewers pulling from one machine would fail. The delay is the cost of serving an audience of unpredictable size.
Delay is sometimes added deliberately
Competitive broadcasts insert additional delay so that viewers cannot relay information about opponents to players in real time.
The added interval must exceed the time it would take for a viewer to observe something and communicate it, which is why tournament delays are measured in minutes rather than seconds.
Individual broadcasters use smaller deliberate delays for the same reason when playing competitive games on stream.
Low latency changes what a stream is for
Reducing delay makes conversation with viewers feel immediate, which matters for broadcasts built on interaction rather than on watching.
The trade is stability: with a small buffer, a brief network problem becomes a visible stall rather than something the viewer never notices.
Broadcasters therefore choose settings according to what their stream is doing, and the same choice would be wrong for a broadcast where picture quality matters more than immediacy.