Per-title encoding is the technique of customising the bitrate ladder for each video individually instead of using one fixed ladder for all content. The insight is simple: a static lecture and a snow-storm action sequence have totally different compression needs. The lecture compresses to clean quality at 2 Mbps; the action scene needs 8 Mbps to look as good. A one-size-fits-all ladder gives the lecture more bits than it needs (wasting CDN bandwidth) and gives the action too few (giving the viewer a worse picture). Per-title encoding analyses each video's complexity and picks the right rungs for that video.
The technique was pioneered by Netflix in 2015 and shocked the streaming industry: across Netflix's catalogue, per-title encoding delivered roughly 20 % bitrate savings at the same perceived quality, meaning 20 % off their CDN bills with no viewer-side downside. The savings vary by content type — simple talking-head content saves more (40–50 %), complex action saves less (5–10 %), but the average across a real catalogue lands around 20 %.
For a product team in 2026, per-title encoding is table stakes for any serious VOD service. The pipeline: encode the video at several quality levels, measure the VMAF score at each, fit a quality curve, then pick the smallest bitrate at each resolution that hits the target VMAF. Modern services from Netflix, YouTube and Disney+ extend this with per-scene-encoding (optimisation per scene, not just per title) and context-aware-encoding (different ladders for different device classes). Off-the-shelf services that handle per-title encoding for you: Mux, Bitmovin Encoding, AWS MediaConvert QVBR, NETINT VPUs with built-in CAE.
