{"id":150,"date":"2026-11-16T07:52:00","date_gmt":"2026-11-16T07:52:00","guid":{"rendered":"https:\/\/tothepowerof2.com\/blog\/?p=150"},"modified":"2026-11-15T08:33:42","modified_gmt":"2026-11-15T08:33:42","slug":"video-encoding-ladders-powers-of-two","status":"publish","type":"post","link":"https:\/\/tothepowerof2.com\/blog\/video-encoding-ladders-powers-of-two\/","title":{"rendered":"Video Encoding Ladders Built on Powers of Two"},"content":{"rendered":"<p><strong>TL;DR \/ Summary:<\/strong> Encoding ladders built on powers of two compress more efficiently because hardware and codecs are designed around those sizes. Choose 360, 720, 1080, and 1440\u2014not 400, 750, or 1100.<\/p>\n<h2>Why the numbers cluster<\/h2>\n<p>Video resolutions were standardised early around dimensions that divide evenly by small powers of two, and hardware encoders were built to those standards. That legacy persists in codecs, in GPU encoding blocks, and in the scaling algorithms that prepare frames.<\/p>\n<p>An encoder processing a frame whose dimensions are divisible by 16, or at least by 8, can work in aligned blocks. A frame with awkward dimensions forces padding or partial blocks, which costs efficiency.<\/p>\n<p>For a single stream the cost is small. Across a ladder of six renditions and thousands of viewers, it becomes measurable.<\/p>\n<h2>The standard ladder<\/h2>\n<p>A practical ladder for most content is 360p, 480p, 720p, 1080p, and 1440p, with the aspect ratio handled separately. The vertical resolution is not a power of two, but the frame dimensions themselves are convenient multiples.<\/p>\n<p>1920 by 1080 and 1280 by 720 both divide cleanly into macroblocks. A &#8220;1100p&#8221; stream at 1955 by 1100 does not, and the encoder will round and pad internally.<\/p>\n<ul>\n<li><strong>426 \u00d7 240<\/strong> \u2014 lowest useful rung for mobile<\/li>\n<li><strong>640 \u00d7 360<\/strong> \u2014 cheap and widely cacheable<\/li>\n<li><strong>1280 \u00d7 720<\/strong> \u2014 the workhorse<\/li>\n<li><strong>1920 \u00d7 1080<\/strong> \u2014 the default for desktop<\/li>\n<li><strong>2560 \u00d7 1440<\/strong> \u2014 the top rung for most libraries<\/li>\n<\/ul>\n<p>Adding a rung between 720 and 1080 rarely helps. The bitrate difference matters more than the height difference.<\/p>\n<h3>Why not add every resolution<\/h3>\n<p>Each rung costs storage, encoding time, and cache complexity. A ladder with nine rungs does not produce better quality; it produces more switching decisions for the player and more objects for the CDN.<\/p>\n<p>Players select by bandwidth and buffer, and the gaps between rungs are tolerable when the bitrate steps are sensible. Four to six rungs covers almost every viewer.<\/p>\n<h2>Bitrate steps and perceptual quality<\/h2>\n<p>Defining a ladder by resolution alone is the common mistake. Two streams at the same resolution with different bitrates serve different network conditions, and the perceptual gain from a resolution bump is often smaller than the gain from a bitrate bump.<\/p>\n<p>Set bitrates so that each rung is roughly thirty to fifty percent above the one below. That produces a smooth adaptation curve rather than a cliff.<\/p>\n<p>For 1080p at thirty frames per second, a bitrate around 4.5 to 5 Mbps is a reasonable starting point for general content. Higher motion needs more; static content needs less.<\/p>\n<blockquote><p>The ladder should be smooth in bitrate, not just tidy in resolution.<\/p><\/blockquote>\n<h2>Demonstration: the scaling chain<\/h2>\n<p>If your source is 1440p and you need 1080p, 720p, and 360p, each derived from the source rather than from the rung above. Deriving 360p from 720p compounds softening.<\/p>\n<p>Modern encoders are fast enough to scale from source for every rung, and the quality difference is visible on detailed content such as text or foliage.<\/p>\n<p>Keep the source at the highest resolution you have. A 1440p master gives you flexibility; a 1080p master limits the top rung forever.<\/p>\n<h2>Frame rate alignment<\/h2>\n<p>Keep frame rate constant across the ladder. Varying frame rate between rungs makes switching visible, because motion suddenly changes character.<\/p>\n<p>If you need a low-frame-rate rung for very low bandwidth, use it as the single lowest rung rather than inserting it in the middle of the ladder.<\/p>\n<h2>Encoding efficiency in practice<\/h2>\n<p>Aligned dimensions matter most with hardware encoders, which process fixed block sizes. Software encoders are more flexible and still benefit from alignment.<\/p>\n<p>Check the output dimensions the encoder actually produces. Some tools silently adjust to the nearest multiple of eight, which is fine\u2014but you want to know, because the player&#8217;s reported resolution should match your manifest.<\/p>\n<p>Test one short clip at each rung and compare file sizes at matched quality. A rung that produces a disproportionate file size for its resolution is a sign that the dimensions are fighting the encoder.<\/p>\n<h3>Storage arithmetic<\/h2>\n<p>Doubling the number of rungs roughly doubles storage. For a large library, the storage cost of an extra rung is often more than the bandwidth saving.<\/p>\n<p>This is another argument for a compact ladder. Six well-chosen rungs beat twelve optimised ones for most libraries.<\/p>\n<h2>A practical ladder to copy<\/h2>\n<p>For a typical talking-head or tutorial library: 360p at 800 kbps, 480p at 1.2 Mbps, 720p at 2.5 Mbps, 1080p at 5 Mbps, 1440p at 9 Mbps. Adjust within these ranges rather than adding rungs.<\/p>\n<p>For high-motion content, raise each bitrate by roughly fifty percent. For slides and screen recordings, lower them; the content compresses well.<\/p>\n<p>If you want to test how a power-of-two layout behaves, <a href=\"https:\/\/tothepowerof2.com\/\">powers of 2<\/a> is a quick reference for the values that divide cleanly, but the ladder above is the practical starting point.<\/p>\n<p>Encoding ladders are a case where convention and efficiency agree. Following the established sizes costs nothing and avoids a class of problems most teams never diagnose.<\/p>\n<h2>Audio in the ladder<\/h2>\n<p>Audio bitrate does not need to scale with resolution, but it should not be identical across every rung either. The lowest rung can use a lower audio bitrate to keep the total below a bandwidth threshold.<\/p>\n<p>Keep the codec consistent across rungs. Switching codecs between adjacent rungs produces audible differences at the transition, which draws attention to the adaptation.<\/p>\n<p>Check that the lowest rung remains intelligible on a phone speaker. A rung optimised for file size at the cost of speech clarity defeats the purpose of having it.<\/p>\n<h3>Testing the switch<\/h3>\n<p>Throttle the connection during playback and watch the player adapt. The transition should be barely visible; a visible jump in quality means the rungs are too far apart.<\/p>\n<p>Test on a phone and on a laptop. Small screens hide artefacts that a larger display reveals, and the ladder has to serve both.<\/p>\n<p>Record the bitrate at which the player settles on each device. That number is more useful for tuning than any theoretical model.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>TL;DR \/ Summary: Encoding ladders built on powers of two compress more efficiently because hardware and codecs are designed around those sizes. Choose 360, 720, 1080, and 1440\u2014not 400, 750, or 1100. Why the numbers cluster Video resolutions were standardised early around dimensions that divide evenly by small powers of two, and hardware encoders were &#8230; <a title=\"Video Encoding Ladders Built on Powers of Two\" class=\"read-more\" href=\"https:\/\/tothepowerof2.com\/blog\/video-encoding-ladders-powers-of-two\/\" aria-label=\"Read more about Video Encoding Ladders Built on Powers of Two\">Read more<\/a><\/p>\n","protected":false},"author":1,"featured_media":59,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[3],"tags":[],"class_list":["post-150","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-binary-exponents-computing"],"_links":{"self":[{"href":"https:\/\/tothepowerof2.com\/blog\/wp-json\/wp\/v2\/posts\/150","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/tothepowerof2.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/tothepowerof2.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/tothepowerof2.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/tothepowerof2.com\/blog\/wp-json\/wp\/v2\/comments?post=150"}],"version-history":[{"count":1,"href":"https:\/\/tothepowerof2.com\/blog\/wp-json\/wp\/v2\/posts\/150\/revisions"}],"predecessor-version":[{"id":151,"href":"https:\/\/tothepowerof2.com\/blog\/wp-json\/wp\/v2\/posts\/150\/revisions\/151"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/tothepowerof2.com\/blog\/wp-json\/wp\/v2\/media\/59"}],"wp:attachment":[{"href":"https:\/\/tothepowerof2.com\/blog\/wp-json\/wp\/v2\/media?parent=150"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/tothepowerof2.com\/blog\/wp-json\/wp\/v2\/categories?post=150"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/tothepowerof2.com\/blog\/wp-json\/wp\/v2\/tags?post=150"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}