Netflix’s AV1 Journey: From Android to TVs and Beyond

Original Article ↗ Hacker News Discussion ↗

Open codecs, DRM, and business realities

  • Many welcome Netflix’s AV1 use as validation of non-proprietary codecs and a way to push hardware vendors toward open standards.
  • A recurring objection: if all streams are DRM-protected, how much does “open” really help?
  • Counterpoint: device makers care about codec support, not DRM internals; once AV1 decoders ship for Netflix, everyone else can target the same hardware.
  • Debate over DRM’s necessity: some argue studios would still license content without it; others with industry experience say serious anti-piracy and “non‑leaky” platforms are actively rewarded in licensing deals and measurably reduce some leakage, even though everything is pirated eventually.

Hardware support, decoding, and why only ~30%

  • Several are surprised AV1 is only 30% of viewing; others note the massive installed base of older smart TVs, sticks, and low-end SoCs that must use hardware decode and often can’t do AV1 at all.
  • Netflix appears to prefer hardware decode for power and thermal reasons; software AV1 (e.g., via dav1d) works but can murder battery or be too heavy for cheap TV CPUs.
  • Confusion around “30% of viewing” vs “30% of devices”; commenters stress this is viewing time/sessions, not hardware penetration.

Codec comparisons, patents, and the “next codec”

  • AV1 vs H.264/H.265: article claims same/higher quality with ~⅓ less bandwidth; some question details of the comparison and VMAF targeting.
  • H.265/HEVC and H.266/VVC are widely seen as poisoned by complex, expensive licensing; examples given of OEMs disabling HEVC hardware to avoid royalties.
  • Many expect AV2 (successor to AV1) to outcompete VVC in the web/consumer space; VVC is reported to see limited use in broadcast and some regions.
  • Discussion on AI/neural codecs: some see them as inevitable for better rate–distortion tradeoffs; others are wary of hallucinated details and argue any “invented” content is unacceptable for trustworthy video, though acceptable for entertainment.

Piracy and scene practices

  • Warez groups largely stick to H.264/H.265: better compatibility, mature tooling, and much faster encoding than AV1 at archival quality.
  • AV1 WEB-DLs are slowly appearing but are constrained by player support and lack of agreed “scene rules.”

HDR “brightness war” and UX problems

  • Long subthread compares HDR abuse to the “loudness war”: TikTok/Instagram clips and phone-shot HDR frequently appear retina-searing and inconsistent with UI brightness.
  • Complaints that mobile OSes let HDR video ignore user brightness, especially on some Apple devices; people want HDR capped or tone-mapped to the chosen brightness except in full-screen, immersive playback.
  • Proposed fixes: HDR “volume leveling” analogous to audio normalization, heuristics or AI to detect over-bright content, temporal-aware tone mapping, or simply disabling HDR for feed-style content.
  • Some note HDR ecosystems (Windows, phones, cheap monitors, social apps) are still a mess; many users turn HDR off entirely.

Netflix video quality and bitrate choices

  • Multiple commenters say Netflix has among the worst perceived quality: macroblocking, noise, crushed bitrates even on 4K plans, especially compared to Apple, Disney, or Blu-ray.
  • Others report decent quality, suggesting differences in edge node proximity, device settings, or UI flags (e.g., Netflix “Data usage per screen”, Apple TV “Match Content”).
  • Several argue AV1 isn’t the culprit; Netflix allegedly uses very low bitrates for partner content and reserves higher-bitrate HEVC for its own originals.

Film grain synthesis and mastering concerns

  • Netflix’s AV1 film-grain synthesis is seen as technically clever: encode a cleaner base image plus metadata, then synthesize grain client-side to save bandwidth.
  • Some worry this discards real (especially photochemical) grain that carries subtle information, replacing it with an approximation.
  • A few suggest a better workflow would allow creators to deliver a “clean” master plus an explicit “grain track” or parameters, instead of adding grain, having Netflix analyze and strip it, then re-add synthetic grain.

Meta: HN titles and moderation

  • Side discussion about the HN title change: users question why a descriptive original title was replaced by a vaguer one.
  • A moderator explains the general policy (prefer original titles, remove numbers/bait when needed) and admits this particular edit under-represented the content, later choosing a better subtitle.