Every title is encoded as a set of renditions, and the player moves between them as the connection changes. A viewer on a train gets a softer picture rather than a stopped one, and never has to think about why.
Ask a viewer what quality they were watching at and most of them have no idea. Ask them whether it stopped and they will tell you exactly how many times. That asymmetry is the entire argument for adaptive bitrate streaming, and it is worth stating plainly because operators consistently optimise for the wrong one of those two things. Serving the highest quality a connection can theoretically manage produces a stream that stalls the moment anything changes, and a stall costs you more goodwill than an hour at 480p ever will. Adaptive delivery means the player is always choosing the best rendition it can sustain right now, not the best one available, and revising that choice continuously as conditions move.
Delivery uses the same adaptive formats every device already understands, which is why a stream plays on a browser, a phone and a television without three different implementations behind it.
The decision is remade continuously as the stream plays rather than once at the start, so a connection that recovers gets a better picture back without anybody reloading anything.
Automatic is right for almost everybody and wrong for the person on a metered connection who would rather choose. The quality control in the player lets them pin it.
Most viewers should never touch this, and a minority genuinely need to. Somebody on a capped mobile plan may prefer to watch at a lower rendition all evening rather than let the player use their allowance on a picture they cannot see the benefit of on a phone. Somebody on a large television with good bandwidth may want to pin the top rendition rather than let the player be cautious during the first few seconds. Both are reasonable, and both are one control in the player settings alongside playback speed and subtitles. Leaving it on automatic is the default because that is correct for the overwhelming majority, and offering the override costs nothing. The rest of what that panel does is covered under the video player.
Adaptive delivery is often discussed as though it were something the player does on its own, and it is not. The player can only select among the renditions that exist, which means the quality of the experience is decided earlier, when you chose your encoding ladder. A ladder with 1080p and nothing below it gives the player no way to help somebody on a weak connection, and the result is exactly the stalling that adaptive streaming is supposed to prevent. This is why operators serving markets with uneven connectivity usually gain more from adding lower rungs than higher ones. The rungs themselves are chosen under video encoding settings, and produced when a file is ingested through video uploading and transcoding.
Choosing the right rendition helps nothing if every segment travels halfway around the world to reach the viewer. Adaptive streaming and edge delivery solve two halves of the same problem, one by reducing what has to be sent and the other by shortening how far it goes, and a service that gets one right and the other wrong still feels slow. Segments are served from the global CDN so the connection the player is measuring is a short one. On live channels the same mechanism applies with less room for error, because there is no buffer to fall back on, which is covered under live streaming. What viewers actually received, as opposed to what you made available, shows up in reports and analytics.
Have questions about taking advantage of this limited-time offer? Check out the FAQ for answers.