H.264 Is Magic (2016)

Original Article ↗ Hacker News Discussion ↗

Modern video compression still feels “magical,” with H.264 as the long-lived baseline and newer codecs like AV1 and eventually AV2 promising major bandwidth savings at the cost of uneven hardware support. Commenters debate when a codec can truly be called mainstream, how patent issues shape adoption (especially for H.265), and where legacy compatibility still forces H.264 in 2026 despite expiring patents. The thread also branches into related topics such as lossy vs. lossless encoding, emerging neural-network-based codecs, and successor image formats like WebP and JPEG XL that aim to do for images what modern codecs do for video.

State of Codecs in 2026

  • H.264 is still seen as “magic,” but commenters note the landscape has shifted to AV1 and emerging AV2.
  • AV1 is praised for markedly better compression than H.264/H.265 at low bitrates, especially for noisy/handheld-style video.

Is AV1 Mainstream?

  • One side argues AV1 is already mainstream: widely used by Netflix and YouTube, supported in many newer TVs, streaming sticks, and CPUs/GPUs.
  • Others counter that lack of support in popular or older devices (e.g., many Roku models, older Firesticks, older Apple hardware) plus software-decoding issues mean it’s not “mainstream” yet.
  • Some report devices overheating or stuttering when AV1 is forced without hardware decode; others say even 4K AV1 software decoding is fine on modern chips.

“Final Codec” vs Ongoing Evolution

  • Some wish for a “final” codec to avoid fragmentation.
  • Others argue new codecs that cut bandwidth by ~30% each decade are worth it; practically there are only a couple of families (H.26x vs VPx/AVx) active at any time.

AI / Neural Compression

  • Multiple comments expect “neural” or “learned” video codecs to be the future.
  • Today they can be competitive in compression ratios but are considered too slow.
  • Comparisons are drawn to DLSS-style frame interpolation and to AI-based text compression and the Hutter Prize.

H.264, H.265, AV1: Which to Use?

  • H.264: still best for compatibility, especially on older or constrained devices and platforms that only accept it (e.g., some social networks).
  • H.265: technically more efficient than H.264, especially for >1080p, but criticized for messy patents and sometimes spotty ecosystem support.
  • AV1: recommended when possible for personal use due to openness and efficiency; main drawback is hardware support on older/cheap devices.

Patents and Royalty Issues

  • H.264 patents are expiring or nearly expired, raising hopes for broader free use.
  • AV1 is nominally royalty-free, but a recent lawsuit is cited as putting that promise into question.

Lossless vs Lossy & Image Formats

  • Discussion notes H.264, VP9, AV1 can all be used in lossless modes, though chroma formats and tooling limit real-world use.
  • Some argue the original article’s comparison to PNG is valid for teaching lossy vs lossless; others think it confuses beginners.
  • PNG is criticized as inefficient; WebP, AVIF, and JPEG XL are highlighted as more modern options, with JPEG XL praised for features and flexibility.

Technical Nuances & Corrections

  • Debate over the article’s explanation of “quantization”: some say it conflates value mapping with spatial masking; others think it’s acceptable pedagogy for non-experts.
  • Clarifications around what “lossless” means (compression vs color space/bit-depth changes) and around chroma subsampling not being truly lossless.

Performance, Hardware, and Storage

  • Hardware support for AV1 decoding is now common in recent Intel, AMD, and ARM chips, and many newer TVs/boxes, but not universal.
  • Some doubt that small/cheap boxes can handle high-res AV1 in software; others think small PCs/NUCs can, though with higher power use.
  • A decade-on reflection: raw data rates once called “impossible to handle” are now within reach of commodity NVMe and multi-gigabit networking, though sustained write hiccups can still be an issue.

Anecdotes and Use Cases

  • Space missions: strong internal pressure to keep images raw, but practitioners often find that lossy compression plus more images is scientifically better, given bandwidth constraints.
  • Medical imaging: proposals to use lossless H.264/AV1 for CT stacks to exploit inter-slice redundancy face practical hurdles with tooling, formats, and regulatory trust.
  • Historical anecdotes describe multi-CPU/chip parallel encoders and complex optimization work in x264, underscoring perceived “magic” in modern video codecs.