Per-scene encoding is the next step beyond per-title encoding: instead of optimising bitrate once per video, the encoder optimises per scene — sometimes per shot, sometimes per group of similar shots. A dialogue scene gets one set of encoding parameters, the action sequence that follows gets another, the slow drone-shot landscape after that gets a third. The result: bits go exactly where they matter, scene by scene, instead of being averaged across the whole movie.
The savings are substantial and content-dependent. Sports with mostly-uniform action see only ~3 % gain over per-title. Animation and motion-heavy content can save up to ~14 %. Nature documentaries and films with very mixed scene complexity can save up to ~30 % — the more variable the content, the more per-scene encoding helps. Combined with per-title and content-aware encoding, total savings of 25–40 % over fixed-bitrate ladders are routine on premium catalogues.
For a product team in 2026, per-scene encoding has become the default for premium streaming. Netflix uses per-shot encoding on UHD content. YouTube, Disney+, Apple TV+ all do scene-aware optimisation at varying levels of sophistication. The computational cost is higher — the encoder needs scene-cut detection, optimal-parameter search per scene, and often multiple encode passes — but the bitrate savings at the same VMAF pay for themselves quickly at scale. Commercial encoder services (Bitmovin, Mux, AWS MediaConvert with QVBR, NETINT) ship per-scene optimisation as a standard feature in 2026.
