Updated October 2026: x265 4.3, SVT-AV1 4.2, libaom 3.14/3.15 and libvpx 1.16 release notes; a new rav1e vs SVT-AV1 vs libaom section; AV2 status after its June 2026 release.

Why this matters

If you operate any service that delivers video — OTT, e-learning, video surveillance, conferencing, telemedicine — the software encoder you choose decides three numbers on your bill: cloud transcoding cost per hour, bandwidth cost per stream, and customer retention via startup time and buffering. Marketing pages talk about codecs in the abstract ("AV1 is 30% better than H.265"), but in production you do not ship a codec; you ship an encoder binary with a specific version, preset, and tuning. This article is the operator's reference for the seven software encoders you will actually deploy in 2026, what each one is good and bad at, and how to choose without re-running every benchmark yourself.

What an encoder is (and why one codec needs many)

A video codec is a written specification. The H.264 specification, formally ITU-T Rec. H.264 and ISO/IEC 14496-10, is a 800+ page document that defines how a bitstream must look so that any decoder in the world can play it back. The specification says almost nothing about how to create that bitstream — it leaves every encoder author free to make speed-versus-quality tradeoffs.

That freedom is why the same codec can have many encoders, and why those encoders produce dramatically different results. The piece of software that turns raw pixels into a compliant bitstream is the encoder implementation. Two encoders for the same codec target the same final container and decoder, but everything between input pixels and output bits — motion search range, mode decision logic, rate control, perceptual heuristics, assembly optimization — is the implementation author's call.

A useful analogy is sheet music. Every orchestra plays the same Beethoven score (the codec), but the recording you actually listen to depends on the conductor, the players, and the hall (the encoder). You do not buy a "Beethoven recording"; you buy a specific recording by a specific orchestra. Same with video: you do not deploy "H.264"; you deploy x264 version X with preset Y.

The 2026 line-up at a glance

Seven open-source software encoders carry most of the world's professional video in 2026. Each has a primary codec, a primary maintainer, a license, and a clear niche.

Encoder Codec First release Primary maintainer License Sweet spot
x264 H.264 / AVC 2003 VideoLAN community GPLv2 + commercial Universal baseline; live and VOD where compatibility wins
x265 H.265 / HEVC 2013 MulticoreWare GPLv2 + commercial Premium VOD, 4K HDR catalogues, broadcast contribution
libvpx VP8 / VP9 2010 Google (WebM project) BSD WebRTC (VP8); legacy VP9 catalogues on YouTube
SVT-AV1 AV1 2018 Intel + Netflix + AOMedia BSD-3 Default production AV1 encoder, VOD and real-time
rav1e AV1 2018 Xiph.Org + Mozilla BSD-2 Memory-safe AV1 for embedded and tiling pipelines
libaom AV1 2018 AOMedia (Google-led) BSD-2 The reference; highest-quality AV1 if compute is free
vvenc H.266 / VVC 2021 Fraunhofer HHI BSD-3-Clause-Clear Cutting-edge VVC encoding; early-adopter 4K and 8K

Three observations follow from the table. First, every major encoder is open source — proprietary encoders exist (Beamr, Visionular, NVIDIA cloud encoders) but the open-source set is what most teams ship. Second, the same codec can have multiple competing encoders (three for AV1 alone). Third, the licenses are split: BSD-style permissive licenses for everything except x264 and x265, which use GPL plus a commercial-license escape hatch.

Vertical timeline of seven open-source video encoders, one card per row, with the year on the left. 2003: x264, H.264 / AVC, VideoLAN, GPLv2 or commercial, still the baseline tier. 2010: libvpx, VP8 and VP9, Google and AOMedia, BSD, VP8 is the WebRTC default. 2013: x265, H.265 / HEVC, MulticoreWare, GPLv2 or commercial, premium 4K HDR. 2018: SVT-AV1, AV1, Intel, Netflix and AOMedia, BSD-3, the production default. 2018: rav1e, AV1, Xiph.Org and Mozilla, BSD-2, Rust, memory-safe, tiled. 2018: libaom, AV1, AOMedia reference, BSD-2, best AV1 if compute is free. 2021: vvenc, H.266 / VVC, Fraunhofer HHI, BSD-3-Clause-Clear, early adopters. A legend marks dark blue as the ITU-T and MPEG line and green as the AOMedia open line. The closing line reads: one codec is one spec; the encoder is a choice, and AV1 alone shipped three in 2018. Figure 1. Two decades of open-source video encoders. Each release corresponds to a codec generation, with AV1 producing three competing encoders in the same year.

The seven encoders in detail

x264 — the universal baseline

x264, started by Laurent Aimar in 2003 and shepherded since by VideoLAN, is the encoder against which every later encoder is measured. It produces compliant H.264 / AVC bitstreams and remains the highest-quality H.264 encoder publicly available. Its assembly-optimized hot paths run on x86, ARM NEON, ARM64, and ARM64EC; low-level work continues in the git tree — recent activity includes AArch64 NEON optimizations and ARM64EC support for Windows-on-ARM. x264 has no numbered releases: distributions ship git snapshots (core 165 in 2026 packages), so pin a commit in your build. The project is maintained at code.videolan.org/videolan/x264 and is dual-licensed: GPLv2 by default, with a commercial license available from x264 LLC for products that cannot comply with GPL.

x264's strengths are speed, predictability, and ten presets (ultrafast through placebo) that cover everything from 4K live ingest to archival-grade VOD. On a modern multi-core CPU, x264 encodes 1080p several times faster than real time even at slow, and that headroom is why every encoding farm on Earth still has x264 in production: H.264 still plays on every browser, every smart TV, every legacy STB, and x264 ships it with no royalty risk for the bitstream itself (decoder royalties are a separate matter).

Use x264 when compatibility outranks compression. Live streaming to a long tail of devices, surveillance, and any pipeline where the receiving device might be ten years old.

x265 — the HEVC workhorse

x265, started in 2013 by MulticoreWare, is the dominant open-source H.265 / HEVC encoder. Version 4.1 shipped on 22 November 2024 (AOM film-grain characteristics carried as SEI), 4.2 on 19 April 2026 (experimental threaded motion estimation, up to about 1.5× faster at 1080p on suitable machines), and 4.3 on 31 July 2026 with foveated encoding, noise-aware selective MCSTF filtering and LoongArch64 support, according to the x265 release notes. MulticoreWare also worked with Arm on Graviton4 optimizations. Like x264, it offers the same ten-preset ladder, the same dual licensing (GPLv2 plus commercial), and the same assembly-heavy hot paths.

The catch is speed. At comparable presets, x265 runs noticeably slower than x264 in exchange for a large bitrate saving at matched VMAF; how large depends on content and preset, which is why our measured BD-rate and fps comparison of x264, x265 and SVT-AV1 exists. That trade is worth it for premium VOD libraries, 4K and 8K archives, broadcast contribution links, and HDR catalogues — anywhere the storage and bandwidth savings amortize the higher CPU bill. It is not always worth it for live streaming at scale, where every channel pays the extra CPU in real time and nothing amortizes it.

libvpx — VP9 and the WebRTC default

libvpx is Google's reference encoder for VP8 and VP9, maintained by Google's WebM project. It is still alive: v1.16.0 ("Xenonetta Duck") was finalized on 27 January 2026 with AVX-512 SAD kernels, Arm SVE2 and Neon convolution work, per-frame and per-layer PSNR, and several memory-safety fixes. It is ABI-incompatible with 1.15, so anything linking libvpx must be rebuilt. Its current weight in production is two-fold. VP9 powered most of YouTube's catalogue throughout the 2016–2024 era, and the encoder is still live for older catalogue maintenance. VP8 remains the default real-time codec in WebRTC, which means every browser and every WebRTC SFU still depends on libvpx for VP8 encoding and decoding.

In the H.264-or-VP9 era, libvpx was the open alternative without the H.264 patent thicket. In the AV1 era, it has been overtaken by SVT-AV1 and libaom, but it survives because the deployed base of VP9 and VP8 streams is enormous.

SVT-AV1 — the production AV1 encoder

SVT-AV1 is the encoder that made AV1 a production-ready codec rather than a research curio. It started inside Intel as the Scalable Video Technology for AV1 in 2018, picked up a strategic partnership with Netflix in 2020, and was adopted by AOMedia in 2020 as the basis for future AV1 development. Version 4.0 shipped on 23 January 2026, with extended CRF range up to 70, finer quarter-step CRF granularity, ~5–15% RTC-preset speedups, Arm Neon and SVE2 optimizations, and the completion of porting all SVT-AV1-PSY perceptual features into the mainline tree under --tune vq (video) and --tune iq (AVIF). Version 4.2.0 followed on 14 July 2026: a VMAF-targeted tune (the changelog cites about 15% VMAF BD-rate gain), CBR rate control for real-time with Kalman-filter QP estimation and cyclic refresh, long-term reference frames, and more Arm Neon and SVE2 kernels.

What sets SVT-AV1 apart is parallelism. It scales horizontally across tens of CPU cores with frame-level and tile-level parallelism, which is why it fills a 64-core server where libaom cannot. Streaming Media's 2022 test of SVT-AV1 1.0 already found it matched libaom-AV1 quality (about 1.4% BD-rate difference) with a much wider speed range, and it has gained since. Presets 4–6 are the production sweet spot for VOD; presets 10 and above are used for real-time AV1.

Use SVT-AV1 as your default AV1 encoder unless you have a specific reason to switch.

rav1e — the safety-first AV1 encoder

rav1e is the AV1 encoder written entirely in Rust, started by Xiph.Org and Mozilla in 2018 and maintained under the Xiph.Org GitHub organisation. The 0.8 line ended a long gap after 0.7 in late 2023; 0.8.1 is the version Linux distributions ship in 2026. Its design priorities are memory safety, simple integration into Rust-native pipelines, and a tile-based architecture that maps cleanly to embedded multi-core SoCs.

In practice, rav1e is competitive with libaom at high speed settings but does not match SVT-AV1's quality at equal encoding time. The open mendevi dataset (2024–2026) shows the same pattern: SVT-AV1 is much faster, while rav1e reaches lower bitrates in constant-quality mode and lower distortion at high bitrates. It earns its place in three niches: any pipeline already written in Rust that wants AV1 without an FFI hop, hardened deployments where memory-safety guarantees matter (a class of CVEs in C/C++ video code is structurally impossible in safe Rust), and tightly tiled embedded encoders where the architectural match pays off.

libaom — the reference

libaom is the AV1 reference encoder, the encoder that defines compliant AV1 output for the spec, maintained by AOMedia primarily by Google engineers. It is the slowest AV1 encoder by a wide margin: at cpu-used 0–2 it can be orders of magnitude slower than x265 on the same clip. The 0–8 cpu-used ladder trades quality for speed, and even at cpu-used 8 it lags SVT-AV1's mid-presets on throughput.

It also produces the highest-quality AV1 bitstreams at the lowest cpu-used values. Streaming Media reported in 2022 that YouTube had produced AV1 with FFmpeg and libaom for years, which fits a catalogue where compute amortizes across billions of views. Development is active: v3.14.0 shipped on 12 May 2026 (multithreading and compound-prediction crash fixes), and the v3.15.0 changelog, dated 24 August 2026, adds faster speeds 0–3, VMAF-oriented tuning that keeps PSNR, Arm convolve and variance kernels, and a low-complexity decode mode for VOD. Use libaom where every saved byte is multiplied by a huge view count; elsewhere SVT-AV1 is the better trade.

vvenc — the cutting edge

vvenc is the Fraunhofer Versatile Video Encoder, the open-source H.266 / VVC encoder from the same lab that gave us H.264 and H.265 reference work. Version 1.14, released in January 2026 with more Arm Neon and SVE optimizations (SVE on by default), is still the current release in October 2026. Fraunhofer reports large speedups over the VTM reference depending on which of its five presets you pick (faster, fast, medium, slow, slower). It scales to 32 CPU threads with frame-level parallelization.

VVC was designed to cut bitrate by about 50% versus HEVC at the same perceived quality, the highest target of any standardized codec in production use. The catch in 2026 is decode: Intel's Lunar Lake laptops added initial VVC hardware decode in late 2024, but no mainstream browser plays VVC and most phones, TVs and streaming sticks still lack it. vvenc is therefore an early-adopter encoder: invest now if you want to be ready when decoders become mainstream in 2027–2028; ignore it if your audience cannot yet play H.266 at all.

And AV2? The Alliance for Open Media released AV2 in June 2026 with the libavm reference software (tag v1.0.0 in May 2026). Like libaom in 2018, it is a reference implementation: far too slow for a production ladder, and no AV2 decoders ship in devices yet. Track it; do not plan 2026–2027 capacity around it.

One raw input, 1080p at 24 fps, 60 seconds, YUV420, splits into three encoder cards. x264 at preset slow: encode time 30 s, bitrate at VMAF 93 is 3.8 Mbps, encode rate 120 fps, bitstream H.264 / AVC, best for live and baseline; cheapest CPU time, biggest files. SVT-AV1 at preset 5: encode time 2.5 min, 1.5 Mbps, 25 fps, AV1 bitstream, best for AV1 VOD at scale; best VOD tradeoff, needs AV1 decoders. vvenc at preset medium: encode time 10 min, 1.3 Mbps, 4 fps, H.266 / VVC bitstream, ready for 2027; smallest files, few decoders yet. The closing lines read: illustrative numbers for one 16-core CPU at the production preset; SVT-AV1 and vvenc keep scaling past 16 cores, x264 gains little. Figure 2. The same pixels, three different cooks. CPU time and output bitrate at matched VMAF differ by an order of magnitude across the encoder choices.

rav1e vs SVT-AV1 vs libaom: which AV1 encoder?

Short answer: use SVT-AV1 by default; it gives the best quality at equal encoding time and is the only one of the three built to fill a many-core server. Pick rav1e when your pipeline is in Rust or memory safety outranks speed, and pick libaom when one encode is watched millions of times and the slowest presets pay for themselves.

Question SVT-AV1 rav1e libaom
Current version (Oct 2026) 4.2.0 (14 Jul 2026) 0.8.1 3.14.0 (12 May 2026); 3.15.0 changelog dated 24 Aug 2026
Language C, with assembly Rust, with assembly C, with assembly
Speed at equal quality Fastest; scales across many cores Slower than SVT-AV1 Slowest at low cpu-used
Where it wins on quality Best at equal encoding time Lower bitrate in constant-quality mode; less distortion at high bitrates Top quality at cpu-used 0–2
Real-time mode Yes; 4.2 adds RTC CBR and long-term references Low-latency mode, rarely used in production Yes (realtime speeds, used in WebRTC AV1)
Best fit Default for VOD and live AV1 Rust pipelines, hardened or embedded encoders Premium VOD with huge view counts; spec reference

Sources: SVT-AV1 and libaom changelogs, the mendevi encoder dataset, Streaming Media's 2022 SVT-AV1 1.0 test. For bitrate and fps on real footage, see the guide to benchmarking encoders on your own content; versions move every few months, so re-run before you switch.

Worked example: encoding the same clip seven times

To make the trade-offs concrete, here is the same 1080p 24 fps clip, 60 seconds long, encoded with each of the seven encoders at their respective "production VOD" preset on a 16-core CPU. The numbers below are representative ranges synthesised from Streaming Learning Center and Phoronix Test Suite benchmarks; treat them as orders of magnitude for encoder versions current in May 2026, not as a final benchmark.

Encoder Preset Encode time (60-s 1080p clip) Bitrate @ VMAF 93 Speed vs x264
x264 slow 30 s ~3.8 Mbps 1×
x265 slow 60–90 s ~1.9 Mbps 0.3–0.5×
libvpx VP9 cpu-used 1 4–5 min ~2.1 Mbps ~0.10–0.12×
SVT-AV1 preset 5 90–180 s ~1.5 Mbps 0.2–0.3×
rav1e speed 6 5–8 min ~1.7 Mbps ~0.06–0.10×
libaom cpu-used 4 30–50 min ~1.4 Mbps ~0.01–0.02×
vvenc medium 8–15 min ~1.3 Mbps ~0.03–0.06×

Reading the table: a 60-second clip costs ~30 seconds with x264 and ~30–50 minutes with libaom, while the output is ~2.7× smaller for libaom. That 60× cost gap is exactly why "best" depends entirely on whether you are paying for CPU once and bandwidth a million times, or CPU once and bandwidth once.

The math you should run before picking is the same in every case:

total cost per stream = (encode CPU cost / number of views) + (bandwidth cost per view × number of views)

If number of views is in the millions (Netflix premium VOD, YouTube), the bandwidth term dominates and libaom or vvenc is worth the encode. If number of views is in the dozens (corporate training, e-learning, internal video), the encode term dominates and x264 is correct.

Speed and quality, visualised

A two-axis view makes the relationships easier to hold in one head. The horizontal axis is encode speed at a 1080p reference workload; the vertical axis is bitrate at matched VMAF.

Scatter chart of seven encoders, encode speed in fps on a logarithmic horizontal axis against bitrate at VMAF 93 in Mbps on the vertical axis. libaom sits at 0.6 fps and 1.4 Mbps, vvenc at 2 fps and 1.3 Mbps, rav1e at 3.7 fps and 1.7 Mbps, libvpx VP9 at 4 fps and 2.1 Mbps, SVT-AV1 at 10 fps and 1.5 Mbps, x265 at 14 fps and 1.9 Mbps, and x264 at 120 fps and 3.8 Mbps. The empty upper area carries the note that nothing lives up there, because slower encoders buy smaller files, not bigger ones. The closing lines read: down is cheaper bandwidth, right is cheaper CPU, no encoder wins both; AV1 spans the widest range, with SVT-AV1 about 15 times faster than libaom. Figure 3. The encoder landscape on two axes. The frontier moves down-and-right with each codec generation; the same codec produces multiple points depending on the encoder.

Three things become clear from the chart. First, each codec generation moves the frontier down-and-right — better compression at the same speed, or the same compression faster. Second, AV1's three encoders span a wide range: SVT-AV1 is roughly an order of magnitude faster than libaom at comparable settings. Third, vvenc is currently slow, but it has more headroom than any other encoder on the chart, because VVC tooling is at version 1.x and will be optimised heavily in 2026–2027.

Picking the right encoder — a decision tree

The choice question is rarely "which codec" — it is "which encoder of which codec for which workload." Three workload classes account for almost all decisions.

Decision tree starting from the question what is the workload, splitting into three branches. Real-time, for WebRTC, surveillance and cloud gaming with latency under 300 ms: use x264 superfast, libvpx VP8 for interop, SVT-AV1 preset 11. Live at scale, for sports, news and live classes with latency of 2 to 10 seconds: use x264 fast to medium, and x265 only if HEVC runs end to end. VOD library, for OTT, training, podcasts and archives where you encode once and replay: use x264 plus x265 plus an SVT-AV1 ladder, and libaom past a million views. Four rules override the default: a closed-source product prefers BSD encoders over GPL x264 and x265; a long tail of old devices keeps x264 in the ladder whatever else you add; HDR 4K or 8K makes x265 or vvenc earn their CPU while SVT-AV1 carries HDR10+; vvenc is only for early adopters, because decoders are still rare in 2026. Figure 4. Decision tree for picking the right encoder in 2026. The dominant inputs are workload type, audience device base, and the number of views each encode amortizes across.

For real-time / sub-second-latency workloads (WebRTC conferencing, surveillance live, cloud gaming), x264 at superfast–veryfast is the safe default; VP8 via libvpx is the WebRTC interop fallback; SVT-AV1 at preset 10 or faster is the AV1 option, with 4.2's real-time CBR rate control, now that real-time AV1 hardware exists.

For live streaming at scale (sports, news, e-learning live), x264 at fast–medium is still the default; x265 only if your contribution and distribution chain plays HEVC end-to-end.

For VOD at scale (OTT, training catalogues, podcasts), the order of operations is: ladder using x264 for compatibility, add an HEVC tier with x265 for HDR or 4K, add an AV1 tier with SVT-AV1 for the top of the bandwidth curve, and consider VVC via vvenc only if your audience demonstrably owns 2024+ flagship phones or recent OLED TVs.

Common mistakes when choosing an encoder

Pitfall: picking by codec, deploying by encoder. A "use AV1" decision made by a CTO with a Streaming Media headline in hand becomes "use libaom" in the engineer's hands, and the team discovers the encoder farm cost jumped 30× overnight. Always specify codec and encoder and preset in your design docs.

Pitfall: trusting old benchmarks. SVT-AV1 went from 3.0 (February 2025) to 4.0 (January 2026) and 4.2 (July 2026); x265 shipped 4.2 and 4.3 three months apart in 2026. Any benchmark older than twelve months is misleading; quality and speed have moved measurably in the last two release cycles alone. Re-run on the encoder version you will actually ship.

Pitfall: ignoring licensing in CI/CD. x264 and x265 ship under GPLv2 by default. If your encoder runs inside a closed-source product you sell, you must either obtain a commercial license from x264 LLC / MulticoreWare or use BSD-licensed encoders (libvpx, SVT-AV1, rav1e, libaom, vvenc). Legal review of the encoder license is as important as the bitstream patent review.

↓ Free download: Encoder Selection Cheat SheetOne page with all seven encoders side by side — use case, production preset, license, and the pitfall each one is known for.

Where Fora Soft fits in

Fora Soft has shipped video pipelines since 2005 for conferencing, OTT and IPTV, video surveillance, e-learning and telemedicine products. Across those projects we have deployed every encoder in this article in production: x264 for the universal baseline tier on conferencing and surveillance, x265 for premium OTT 4K HDR, libvpx for WebRTC VP8 fallback paths, SVT-AV1 for new AV1 OTT tiers, and vvenc in proof-of-concept work for clients planning H.266 rollouts in 2027. If you are building the transcoding side of a product, our video streaming software development team sets up the encoder, preset and ladder per tier; for catalogues and live channels, our OTT and IPTV platform development work adds the HEVC and AV1 renditions, DRM and device QA around them. If an existing farm is too slow or too expensive, transcoding performance troubleshooting and optimization starts with profiling the encoder settings you already run. The recurring lesson from those projects: the encoder choice is rarely the bottleneck; the bitrate ladder, the rate-control mode (CRF vs capped-CRF vs CBR), and the per-title tuning move quality far more than swapping one encoder for another at the same preset.

Video Encoding: A Complete Guide to Video Compression - book cover

The book

Video Encoding: A Complete Guide to Video Compression — From Pixels and Color to H.264, HEVC, AV1, Quality Metrics, and AI

Picking x264 over SVT-AV1 is only the first decision; the preset, the rate-control mode and the ladder behind it decide what the 3.8 Mbps on this page becomes in production. The book spends 1,522 pages on what sits under the preset name: how mode decision and RDO spend your CPU budget, why CRF and capped CRF behave differently on the same clip, and how to prove with VMAF that the cheaper encode does not look cheaper. It is volume one of the seven-volume Fora Soft Video Engineering Handbook, written by Nikolay Sapunov, CEO at Fora Soft.

Read it on Amazon →Kindle & paperback · Fora Soft Video Engineering Handbook

Not sure which encoder belongs on each rung of your ladder?

Send us your sources, your rendition ladder and the devices your audience actually plays on, and we will tell you which encoder and preset each tier should run, what the switch costs in CPU minutes, and whether the GPL encoders are a licensing problem for your product — 250+ video products shipped since 2005.

Book a 30-min call →WhatsApp →Email us →

See our case studies

References

  1. ITU-T Rec. H.264, Advanced video coding for generic audiovisual services. https://www.itu.int/rec/T-REC-H.264. Accessed 2026-05-17. Supports: codec specification authority.
  2. VideoLAN, "x264, the best H.264/AVC encoder." http://www.videolan.org/developers/x264.html. Accessed 2026-05-17. Supports: x264 licensing, repository, and history.
  3. VideoLAN x264 GitLab. https://code.videolan.org/videolan/x264. Accessed 2026-05-17. Supports: recent x264 development activity, ARM64EC support.
  4. x265 documentation, Release Notes. https://x265.readthedocs.io/en/master/releasenotes.html. Accessed 2026-05-17. Supports: x265 4.1, 4.2 release dates and feature list.
  5. MulticoreWare / Arm collaboration on x265 + AWS Graviton4. https://www.prnewswire.com/news-releases/multicoreware-collaborates-with-arm-to-optimize-and-advance-x265-video-encoding-on-aws-graviton4-302250055.html. Accessed 2026-05-17. Supports: x265 Arm optimisation context.
  6. Google webm libvpx repository. https://chromium.googlesource.com/webm/libvpx. Accessed 2026-05-17. Supports: libvpx ongoing maintenance.
  7. AOMedia, "SVT-AV1 4.0 Boasts Major API Updates." http://aomedia.org/blog%20posts/SVT-AV1-4_0-Boasts-Major-API-Updates/. Accessed 2026-05-17. Supports: SVT-AV1 4.0 release date and feature list.
  8. Phoronix, "SVT-AV1 4.0 Released With More Performance Optimizations." https://www.phoronix.com/news/SVT-AV1-4.0. Accessed 2026-05-17. Supports: SVT-AV1 4.0 performance claims.
  9. xiph/rav1e GitHub repository. https://github.com/xiph/rav1e. Accessed 2026-05-17. Supports: rav1e design goals, current status.
  10. Fraunhofer HHI vvenc project. https://github.com/fraunhoferhhi/vvenc. Accessed 2026-05-17. Supports: vvenc 1.14 release, preset list, threading model.
  11. Fraunhofer HHI, "VVenC — the Fraunhofer Versatile Video Encoder." https://www.hhi.fraunhofer.de/en/departments/vca/technologies-and-solutions/h266-vvc/fraunhofer-versatile-video-encoder-vvenc.html. Accessed 2026-05-17. Supports: VVC compression efficiency vs H.264 / AV1.
  12. Netflix Technology Blog, "AV1 — Now Powering 30% of Netflix Streaming" (December 2025), as reported by Streaming Media Blog. https://www.streamingmediablog.com/2025/12/netflix-av1.html. Accessed 2026-10-24. Supports: AV1 at about 30% of Netflix on-demand viewing.
  13. Streaming Learning Center, "SVT-AV1 vs. LibAOM." https://streaminglearningcenter.com/encoding/svt-av1-vs-libaom.html. Accessed 2026-05-17. Supports: SVT-AV1 vs libaom quality and speed trade-offs (2022 test).
  14. Streaming Learning Center, "Choosing a Preset for SVT-AV1 and libaom-AV1." https://streaminglearningcenter.com/encoding/choosing-a-preset-for-svt-av1-and-libaom-av1.html. Accessed 2026-05-17. Supports: libaom and SVT-AV1 preset behaviour.
  15. FFmpeg AV1 Encoding Wiki. https://trac.ffmpeg.org/wiki/Encode/AV1. Accessed 2026-05-17. Supports: AV1 encoder usage patterns.
  16. x265 documentation, Release Notes (master), version 4.3. https://x265.readthedocs.io/en/master/releasenotes.html. Accessed 2026-10-24. Supports: x265 4.3 date (31 July 2026) and features.
  17. SVT-AV1 v4.2.0 release notes (mirror of the GitLab changelog). https://newreleases.io/project/gitlab/AOMediaCodec/SVT-AV1/release/v4.2.0. Accessed 2026-10-24. Supports: SVT-AV1 4.2.0 date and features.
  18. libaom, "Update CHANGELOG for libaom v3.15.0." https://aomedia.googlesource.com/aom/+/refs/tags/v3.15.0-rc1%5E%21/. Accessed 2026-10-24. Supports: libaom 3.15.0 changelog.
  19. VideoLAN, "contrib: aom: update to 3.14.0." https://mailman.videolan.org/pipermail/vlc-commits/2026-May/073189.html. Accessed 2026-10-24. Supports: libaom 3.14.0 release (12 May 2026).
  20. WebM Project, "libvpx v1.16.0 Xenonetta Duck Finalized." https://groups.google.com/a/webmproject.org/g/webm-discuss/c/NsLvY2mRj7M. Accessed 2026-10-24. Supports: libvpx 1.16.0 date and changes.
  21. Arch Linux, rav1e package. https://archlinux.org/packages/extra/x86_64/rav1e/. Accessed 2026-10-24. Supports: rav1e 0.8.1 as the shipped version in 2026.
  22. mendevi, "libsvtav1 vs librav1e vs libaom-av1 Dataset." https://mendevi.readthedocs.io/1.3.3/dataset/explore/svtav1_vs_rav1e_vs_aom.html. Accessed 2026-10-24. Supports: SVT-AV1 vs rav1e speed and quality pattern.
  23. Streaming Media, "AOMedia Delivers on SVT-AV1's Promise" (2022). https://www.streamingmediaglobal.com/Articles/Editorial/Featured-Articles/AOMedia-Delivers-on-SVT-AV1s-Promise-154488.aspx. Accessed 2026-10-24. Supports: SVT-AV1 1.0 vs libaom quality; YouTube on libaom.
  24. Alliance for Open Media, "Alliance for Open Media Releases AV2 Codec." https://aomedia.org/press%20releases/Alliance-for-Open-Media-Releases-AV2-Codec/. Accessed 2026-10-24. Supports: AV2 release (June 2026), libavm reference software.
  25. Phoronix, "Intel Media Stack Updated With Full Support For Lunar Lake." https://www.phoronix.com/news/Intel-Media-Driver-2024-Q3. Accessed 2026-10-24. Supports: initial VVC hardware decode on Lunar Lake (October 2024).