Why an upload is not played as it is
A creator might upload a high-quality file from an editing program. That file may be too large for most connections, use settings some devices cannot decode, or be packaged in a way unsuited to streaming. Even if it played everywhere, a single version would suit only some viewers.
So services process each upload into a set of versions designed for different screens and connection speeds. That processing is called transcoding, and the resulting set is commonly described as an encoding ladder.
Encoding versus transcoding
Encoding means compressing video with a codec. Transcoding means taking video that is already compressed and converting it into another form: a different codec, resolution, bitrate, frame rate or container. In streaming, the upload is first decoded back to raw frames, then encoded again for each target.
Since the source has already been compressed once, each conversion loses a little more quality. For that reason services prefer a good-quality source, and creators who control their export settings are usually better off uploading a high-quality file than one already squeezed heavily.
What a ladder is
An encoding ladder is the list of renditions produced from one source, each pairing a video resolution with a video bitrate, and often a codec. Each entry is a rung. A simplified, illustrative ladder might run from a small, low-bitrate rung suitable for a weak mobile signal up to a large, higher-bitrate rung intended for a television. The numbers differ widely among services and are not standardised, so there is no single correct ladder.
The purpose of the ladder is to let adaptive bitrate players choose the highest rung their current conditions can sustain, moving up or down as those conditions change.
What makes a good ladder
Designing a ladder involves several linked decisions.
How many rungs. More rungs let a player pick something closer to its available speed, but each extra rung costs processing and storage and may split viewers across more files, which reduces how effectively caches are used. Too few leave big gaps, in which a viewer forced to drop a rung loses a lot of quality.
Spacing between rungs. Steps should be big enough to be worth switching but not so big that quality jumps feel jarring. Perceived quality tends to improve less and less as bitrate rises, so the steps at the top of a ladder are often larger in bitrate terms than the steps at the bottom.
The lowest rung. A light bottom rung keeps playback going on poor connections, even if it looks soft. It also helps startup, as discussed in startup delay.
Which resolutions to include. Offering a resolution that few viewers’ screens can show is costly. Offering fewer than devices can display wastes the viewers’ capacity. The topic of resolution is explored in more detail under video resolution, including why higher pixel counts are not always visible on small screens.
Frame rates and aspect ratios. A source recorded at a high frame rate may be offered at that rate on upper rungs only. Vertical clips need ladders designed around tall frames; see aspect ratios.
Fixed ladders and tailored ones
The simplest approach applies the same ladder to every video. It is easy to operate, but wasteful: a static talking-head video needs far less bitrate to look good than a fast sports match or a video game recording, so a fixed ladder may over-spend on some titles and under-serve others.
Many services therefore tailor the ladder to the content. Analysis of each title’s complexity, sometimes called per-title encoding, can decide how much bitrate each resolution truly needs. More granular approaches adjust settings within a title, scene by scene. These approaches usually rely on variable bitrate encoding rather than a flat rate, and on quality measures that estimate how a viewer would judge the result. Their benefits are real but depend on the content and on how much extra processing a service is willing to pay for.
Codecs in the ladder
A ladder may also vary by codec. The H.264 codec has very broad device support and often forms the base. More efficient codecs such as the HEVC codec, the VP9 codec or the AV1 codec can deliver comparable quality at lower bitrates, saving delivery costs and helping weak connections, but they take longer to encode and not all devices can decode them. A service might offer newer codecs only on popular titles or higher rungs. Hardware encoding can reduce the time and energy needed, but may produce somewhat different efficiency from slower software settings.
Cost and timing
Transcoding takes computing power, and the cost grows with the number of rungs, the length of the video and the complexity of the codec. For on-demand video, jobs can be queued and handled in parallel, with long videos often split into chunks and processed concurrently. For live video, the work must keep pace with the incoming feed, which limits how elaborate the ladder can be and affects live delay, a point developed in stream latency.
Because storage and delivery also scale with the number of versions, ladder design is closely tied to streaming costs. Platforms often decide how much processing a particular upload deserves, which may be why a new upload is sometimes available in lower quality first, with higher rungs appearing later.
After transcoding
Once the rungs exist, they are packaged into segments with manifests, in the formats discussed in HLS and DASH, with encryption applied if the content is protected. From there the files are copied out through content delivery networks, ready for any player to choose among.
What creators should take from this
Creators rarely build ladders themselves; the platform does. What they control is the source: upload at a good quality, at the intended frame rate and aspect ratio, using appropriate export settings, and leave the downscaling to the service. Heavy pre-compression, mismatched settings or low-resolution sources limit what any ladder can deliver.