Adaptive bitrate streaming is a video-delivery technique in which a service provides the same programme in multiple encoded versions, usually at different bitrates and often different resolutions, and the playback device dynamically switches between those versions according to current network and device conditions. Instead of selecting one fixed-quality stream for the entire viewing session, the player can request higher-bitrate segments when sufficient bandwidth is available and lower-bitrate segments when network capacity falls.
The objective is to maximise viewing quality without causing unnecessary buffering. Common adaptive streaming technologies include HTTP Live Streaming and MPEG-DASH. The important concept is that the content exists as several alternative representations, divided into corresponding segments or fragments, allowing a client to switch representation at defined boundaries while continuing playback.
Preparing adaptive streaming begins with encoding the source video into a bitrate ladder. Each rung represents a different combination of resolution, frame rate, codec settings and target bitrate. A service might provide low-resolution versions for constrained connections and higher-resolution versions for capable networks and displays. The same timeline is segmented consistently across the representations so that a player can move from one quality level to another without losing the programme's temporal continuity.
The service publishes a manifest or playlist describing the available representations and the locations of their segments. The player reads this metadata before and during playback and uses it to determine which media segment to request next. Careful encoding alignment is important because switching works most reliably when alternative representations contain compatible segment boundaries and appropriate random-access points.
The player's adaptation algorithm estimates available throughput by observing how quickly previously requested segments arrived. It may also consider buffer occupancy, playback rate, device processing capability, screen resolution, measured network conditions and recent switching history. If the player estimates that it can sustain a higher bitrate while maintaining a sufficient buffer, it may request a higher-quality representation for subsequent segments. If throughput falls or the buffer approaches a dangerous level, it can select a lower representation. A good algorithm avoids oscillating rapidly between qualities because constant switching can be distracting and may not accurately reflect sustained network conditions. Different streaming systems and players use different algorithms, so adaptive bitrate behaviour is not governed by one universal formula.
Buffer level is particularly important because throughput alone does not determine whether playback will stall. Suppose a player has several seconds of buffered video: a temporary throughput reduction may be harmless because existing buffered content continues to play. Conversely, if the buffer is nearly empty, even a brief reduction can cause an interruption. Adaptation algorithms therefore trade quality against rebuffering risk.
A conservative player may reduce quality early to preserve continuity, while a more aggressive player may pursue higher quality and accept greater buffering risk. Startup is another special case: the player may initially choose a relatively low representation so that playback begins quickly, then move upward after it has gathered evidence about available bandwidth.
Adaptive bitrate streaming is particularly useful because internet throughput is variable. A user's available capacity can change as they move between wireless coverage areas, as other devices use the same connection, as household traffic changes, or as congestion affects the route to the content server. A fixed high-bitrate stream may look excellent when conditions are good but repeatedly stall when conditions deteriorate. A fixed low-bitrate stream avoids some stalls but wastes available capacity when the network is capable of more. Adaptive streaming addresses this tension by treating quality as a variable that can change during playback. It therefore improves robustness across heterogeneous networks rather than assuming that every viewer has a stable connection with a predetermined capacity.
The technique involves more than bandwidth adaptation. Device capability can limit which representations are practical because decoding high-resolution or computationally demanding video may exceed the capabilities of a low-power device. Display resolution also affects the usefulness of higher-resolution representations. Codec support can determine which representation a client can decode. Content delivery network performance matters because a theoretically adequate internet connection can still experience poor throughput if the selected delivery server is congested. The player may consequently select a lower bitrate even when the user's nominal broadband package is very fast. Adaptive streaming is therefore a client-side decision informed by the complete playback environment rather than simply a direct measurement of the subscriber's advertised internet speed.
Quality switching normally occurs at segment boundaries, so a player does not arbitrarily change the bitrate in the middle of an encoded segment. Shorter segments allow more frequent adaptation and can reduce latency, but they increase request overhead and can complicate encoding and distribution. Longer segments can improve efficiency and reduce overhead but make adaptation slower and can increase live-stream latency. The encoding ladder must also be designed carefully because too many representations increase storage and processing requirements, while poorly chosen bitrate steps can produce inefficient quality changes. Modern codecs and perceptual encoding techniques can improve the amount of visual quality obtained per unit of bitrate.
Adaptive bitrate streaming is therefore best understood as a feedback-control process connecting content preparation, network measurement, buffering and representation selection. The server supplies a set of possible qualities; the client observes playback conditions; the adaptation logic chooses what to request; and the resulting buffer state and download performance provide new information for the next decision.
Its main benefit is not that it makes the network faster, but that it allows the video service to respond intelligently to a network whose available capacity changes over time. It is now a foundational technique for large-scale internet video because it permits one service to accommodate a wide range of devices, connection qualities and usage environments.
