Updated October 2026: AV2 1.0 is final (28 May 2026) but FFmpeg still has no AV2 support; JVET opened the H.267 Call for Proposals (July 2026); Microsoft open-sourced MLVC, the first learned codec in a shipping calling app. New status table and FAQ.

Why this matters

Every codec generation reshapes who pays for bandwidth, who controls the toll booth on patent licensing, and which hardware vendor sells the next decade of chips. If you build, operate, or fund a service that ships pixels — streaming, video conferencing, surveillance, telemedicine, OTT, e-learning, AR/VR — each codec generation has cut 30–50% of the bitrate at the same quality (Figure 2), and the next cycle will move that saving in a direction the past few cycles did not. AV2 is royalty-free under the AOMedia Patent License. H.267 will be patent-encumbered and excellent. Neural codecs will be neither — they will ship as machine-learning model weights instead of a fixed bitstream specification, and the supply chain for them looks nothing like the supply chain for H.264.

This article is for product managers, founders, video operations leads, and engineers who need to read the next five years of the codec roadmap, understand the three competing tracks well enough to talk to a CFO and a chip vendor in the same meeting, and know where the genuinely surprising risks sit. We walk through each of the three tracks, the technical claims, the practical state of decoders and tooling in 2026, the licensing and hardware story, and a planning frame you can take back to your team.

What is the next-generation video codec? Status in October 2026

The next-generation video codec for open-web streaming is AV2, the royalty-free successor to AV1 from the Alliance for Open Media. Its 1.0.0 specification was finalised on 28 May 2026 and it needs about 30% less bitrate than AV1, but as of October 2026 no consumer device, browser or FFmpeg build decodes it. 28 32 Behind it come H.267, the patent-pooled successor to H.266/VVC that JVET targets for October 2029, and neural video codecs, which reached a shipping video-calling product for the first time in 2026. 33 37

A neural video codec is a codec in which trained neural networks, not a hand-written specification, do the compression: an encoder network turns frames into a compact latent, and a decoder network rebuilds the frames from it. The table gives the state of each track today.

Track Where it stands (checked 14 October 2026) Next milestone Source
AV2 (AOMedia)Specification 1.0.0 final, 28 May 2026; AVM 1.0.0 reference encoder and decoder; dav2d software decoder from VideoLAN (May 2026); decoder IP from VeriSilicon, Chips&Media and Allegro DVT; no shipping devicesFirst AV2 chips expected in 2027; Intel SVT-AV2 encoder shown at IBC 2026AOMedia, VideoLAN, VeriSilicon
AV2 in FFmpegNo AV2 decoder or encoder in FFmpeg 8.1 (March 2026) or in the codec list on ffmpeg.orgFirst support expected through an external library (libavm or dav2d)ffmpeg.org
H.267 (JVET)ECM 20.0 (23 June 2026): about 28% lower luma bitrate than VVC, about 43% on screen content; joint Call for Proposals issued 17 July 2026Proposals evaluated January 2027; first version targeted for October 2029ITU-T JVET, Ofinno
Neural codecsDCVC-RT (CVPR 2025): 21% fewer bits than H.266 at 125 fps 1080p on one A100; Microsoft MLVC open-sourced in July 2026, 100 fps average on Apple, Intel and Qualcomm NPUsNo video standard yet; JPEG AI (ISO/IEC 6048-1:2025) covers images; JVET's neural-network video coding exploration continuesMicrosoft Research, ISO

Three tracks, one question

To understand the next five years of video codecs, it helps to draw three parallel tracks on a single road. Each track is a different answer to the same question — "given that we already have H.265 and AV1 and H.266, how do we compress video another 30% smaller without making the playback unwatchable on a phone?" — and each track has a different group of people working on it.

The first track is the Alliance for Open Media (AOMedia), the consortium of Google, Netflix, Amazon, Apple, Meta, Microsoft, Mozilla, Cisco, Tencent and others, that produced AV1 in 2018 and released AV2 in May 2026. AV2 keeps the same shape as every codec since H.264 — block-based prediction, transform, quantisation, entropy coding, loop filter — but pushes every box in that pipeline a little further. It is hand-engineered by humans, royalty-free under the AOMedia Patent License, and its final 1.0.0 specification was released on 28 May 2026 and announced on 9 June 2026. 27 1 2

The second track is the Joint Video Experts Team (JVET) of ITU-T and ISO/IEC, the same standards body that produced H.264, H.265, and H.266. JVET is working on the successor to H.266/VVC, currently labelled the Enhanced Compression Model (ECM) and expected to ship as H.267 around 2029. ECM is the same kind of hand-engineered hybrid codec as AV2, but it is being developed inside a patent-encumbered standards process, and it targets a more ambitious 40% bitrate reduction over H.266. 11 12

The third track is end-to-end learned video compression — neural networks trained to compress and decompress video directly. There is no fixed bitstream specification; the compressed file is whatever the trained model produces, and the decoder is the rest of the model. The Deep Contextual Video Coding (DCVC) family from Microsoft Research, the Deep Render commercial codec, JPEG AI, and JVET's NNVC exploration are all on this track. In 2026 the best learned codecs match or slightly beat H.266 on compression and run in real time on a single GPU, and Microsoft's open-source MLVC runs on laptop NPUs inside a shipping calling product. 36 17 18 19

The three tracks are not mutually exclusive. AV2 arrives first and is set to take over royalty-free streaming from AV1 as hardware ships in 2027–2029. H.267 will land at the high end of the patent-licensed ecosystem in 2029–2030. Learned codecs will keep climbing the quality curve in parallel and will, by the late 2020s, ship inside hybrid codecs as drop-in replacements for individual blocks (intra prediction, loop filter) before any pure learned codec replaces a generation outright.

The rest of this article is the operator's view of those three tracks: what each one buys you, what it costs, what the decoder story looks like, and how to plan your pipeline so you are not on the losing end of any of them.

Timeline of three tracks to the future of video codecs, 2024 to 2031. Green track, AOMedia AV2, royalty-free under the AOMedia Patent License: public draft v13 in January 2026, final 1.0 specification on 28 May 2026, first AV2 chips expected in 2027, and broad consumer rollout in 2028 to 2029. Dark blue track, JVET H.267 and ECM, licensed through patent pools: ECM 20 in 2026 at 28% below VVC, the H.267 first-version target of October 2029, and broad deployment in 2034 to 2036, shown as an arrow running off the chart. Purple track, end-to-end learned codecs with no fixed bitstream specification: DCVC-RT in 2025 at 21% below VTM, the JPEG AI standard in 2025, Microsoft MLVC on laptop NPUs in 2026, and end-to-end codecs in browsers in 2030 to 2032. A closing line says AV2 lands first, H.267 arrives at the end of the decade, and learned codecs ship first as blocks inside hybrid codecs, then as full codecs in closed apps. Figure 1. Three parallel tracks define the future of video codecs through 2030. AV2 is the closest: final since 28 May 2026, royalty-free, with first chips expected in 2027. H.267 follows the traditional ITU/ISO licensing model and lands at the back of the decade. End-to-end learned codecs run alongside both and will likely first ship as ML modules inside the hybrid codecs rather than as a full replacement.

Track one — AV2: the next royalty-free workhorse

AV2 is what most operators will switch their delivery pipeline to between 2027 and 2029, because it is the cheapest codec in licensing, the easiest to procure decoders for from the AOMedia membership, and the one with the closest match to current pipelines. It is also the smallest jump in technical risk from where most teams are today on AV1.

The Alliance for Open Media spent five years and more than 2,700 commits against its reference software building AV2. The public working draft followed in January 2026, and the final AV2 1.0.0 specification was released on 28 May 2026; for the release, encoder and hardware status in detail, see our AV2 codec guide. 3 1 4 AV2 will follow the same patent strategy as AV1: AOMedia members commit not to assert patents against implementers; non-members may still attempt to form a patent pool (Sisvel has said it is preparing one, and Access Advance opened an exploration phase for an AV1/AV2 pool on 25 August 2026 42), but the marketing positioning and contract path remains "royalty-free for the open web". 5

What AV2 actually changes versus AV1

AV2 is a hand-engineered hybrid codec — the same block-prediction-transform-quantise-entropy-filter pipeline that every codec since H.264 has shared. Where it differs from AV1 is mostly in how much detail each box of that pipeline now contains. The high-level shifts that matter for an operator are:

  • Larger and smarter blocks. The maximum coding unit grows from 128×128 in AV1 to 256×256 in AV2, with fully recursive partitioning into many more shapes. Large smooth regions (sky, walls, animated backgrounds) are coded with fewer overhead bits; fine detail (text, faces, sports motion) can drop down to small partitions when needed. 6
  • Better intra prediction. AV2 adds Multiple Reference Line Selection (MRLS) — the encoder can pick from four candidate top and left reference lines for every directional intra prediction mode, instead of always copying from the immediately neighbouring row and column. Combined with new data-driven intra modes and improved chroma-from-luma modelling, intra-only frames shrink noticeably. 6
  • Longer memory for motion. AV2 keeps up to seven active reference frames, as AV1 does, but reorganises how they are chosen and signalled through a ranked reference list. Slow scenes with periodic motion (sports loops, animated backgrounds, screen content) benefit most. 7
  • A redesigned loop filter. The in-loop filter chain is rebuilt: a redesigned deblocking filter, CDEF, loop restoration and two new filters — a guided detail filter (GDF) and a cross-component sample offset (CCSO) — clean up compression noise while preserving more detail. 7

The headline coding gain is around 30% lower bitrate than AV1 at equivalent perceptual quality. Netflix's Andrey Norkin reported approximately 28.7% BD-rate improvement on YUV-PSNR and 32.6% on VMAF using AOMedia's Common Test Conditions corpus; independent 4K benchmarks see 18–25% additional savings on top of AV1, with the biggest gains on high-motion and screen content. 8 9 6 In headline terms, an HD stream that needs 4 Mbps on H.264 and about 1.4 Mbps on AV1 will need roughly 1.0 Mbps on AV2 once encoders are tuned, as the worked example below shows.

Decoder and hardware story

AV2 also ships into a more favourable decoder landscape than AV1 did. 10 AOMedia's 2025 member survey shows 53% of members plan to deploy AV2 within 12 months of finalisation, and 88% within two years. 7 5 As of October 2026, though, no phone, TV or browser decodes AV2 in hardware. Decoder IP is already licensed to chip makers, among them VeriSilicon's VC9800D (June 2026), Chips&Media and Allegro DVT, so the first AV2-capable chips are expected in 2027. 43 At IBC in September 2026 Netflix said it looks forward to adopting AV2 as its next-generation streaming codec, and Prime Video committed to deployment. 32

Software decoding arrived first. AOMedia's AVM reference decoder ships with the 1.0.0 release, and on 31 May 2026 VideoLAN published dav2d, a fast open-source AV2 decoder from the dav1d team, feature-complete for 8- and 10-bit video but still being optimised. 29 Its authors note that AV2 decoding is roughly five times more complex than AV1, so real-time software playback depends on hand-tuned assembly. At CES 2026 Google played 1080p, 24 fps AV2 from YouTube in a custom Chrome build with the libavm decoder on a gaming laptop. 30 For a streaming operator, this means AV2 is a lab codec in 2026, hardware decode arrives through 2027 and 2028, and the long tail of older devices (TVs from 2022 and before, low-end Android) stays on AV1 or H.264 for another four to five years.

The headline practical risk for AV2 is the same risk AV1 had: encoder maturity. The reference encoder, like libaom for AV1, is honest but slow. Intel showed an open-source SVT-AV2 encoder at IBC 2026, but production-quality AV2 encoders — the eventual equivalents of today's SVT-AV1 — will need 18–24 months after the bitstream freeze to reach the speed and tuning of today's best AV1 encoders. Until then, AV2 will be a premium-tier choice for VOD and a poor choice for live, mirroring the early-life trajectory of every codec generation in the last 25 years.

Bar chart of bitrate saved over the previous codec generation at the same perceptual quality, on a 0 to 60% axis with a grey band from 30 to 50% marking the historical envelope. H.264 to H.265: 50%, broad use from 2014. H.265 to AV1: 30%, broad use from 2020. AV1 to AV2: 30%, broad use from 2028. H.266 to H.267: 40%, drawn as a dashed outline because it is a target, not yet measured, with broad use in 2034 to 2036. A closing line says each step saves a similar share, and decoder rollout decides when you bank it. Figure 2. Each codec generation has delivered roughly a 30–50% bitrate reduction over its predecessor at the same perceptual quality. AV2's ~30% over AV1, and ECM/H.267's targeted 40% over H.266, both sit within this historical envelope.

Does FFmpeg support AV2?

No. As of 14 October 2026, FFmpeg can neither decode nor encode AV2. Neither FFmpeg 8.1 "Hoare" (March 2026) nor the codec list on ffmpeg.org has an AV2 entry; the AV1 row still names only libaom, libdav1d, librav1e and libsvtav1. 31 On your own build, ffmpeg -hide_banner -codecs | grep -i av2 finds nothing.

Until support lands, the working setup is FFmpeg around AOMedia's reference tools: FFmpeg prepares the input and measures the output, and the AVM tools do the AV2 part. A test run takes three steps:

  1. Export the source as Y4M with FFmpeg: ffmpeg -i input.mp4 -pix_fmt yuv420p -f yuv4mpegpipe source.y4m (for 10-bit, use yuv420p10le and add -strict -1).
  2. Build AVM v1.0.0 from the AOMedia repository and encode with its command-line encoder. AVM is a fork of libaom, so the tool and its options follow aomenc; check the binary names in your build directory and run them with --help for the rate-control flags. Expect encoding far below real time, because AVM is a reference encoder.
  3. Decode with the AVM decoder or dav2d back to Y4M, then score it in FFmpeg against the source, for example ffmpeg -i decoded.y4m -i source.y4m -lavfi libvmaf -f null - (distorted input first, reference second).

FFmpeg support will most likely arrive the way AV1 support did: first as a wrapper around an external decoder library, then as a native decoder with hardware-acceleration hooks once chips ship. For the full list of AV2 tools and chips, see our guide to AV2 encoder, decoder and hardware status.

Track two — H.267: the heir to VVC

While AV2 was finalised, JVET finished its exploration model. The final Enhanced Compression Model, ECM 20.0 (23 June 2026), cuts luma bitrate by about 28% versus H.266/VVC in random-access tests on natural content, and by about 43% on screen content. On 17 July 2026 ITU-T and ISO/IEC issued the joint Call for Proposals for video coding beyond VVC: coded submissions are due on 26 October 2026, proposals are evaluated at the JVET meeting in Brisbane on 13–22 January 2027, and the first version of the new standard, expected to become H.267, is tentatively targeted for October 2029. 33 34 Broad deployment would follow in roughly 2034–2036 if the historical eight-to-ten-year lag holds. 11 12

H.267 will be a hand-engineered hybrid codec, like AV2 and every codec before it. The technical agenda of ECM is more ambitious than AV2's because JVET is willing to absorb more decoder complexity in exchange for compression: deeper block partition trees, more transform variants, more intra prediction modes, more sophisticated motion compensation, and, as candidates, neural-network blocks for specific stages (intra prediction, in-loop filtering). ECM itself contains no neural tools; those are developed in JVET's parallel Neural Network-based Video Coding (NNVC) exploration. 14

The headline difference between H.267 and AV2 is licensing and timing. H.267 will almost certainly be licensed through patent pools, under the same model as H.264, H.265, and H.266. Today that means Access Advance, which took over Via LA's HEVC and VVC programs in December 2025, plus Avanci Video and Sisvel. 41 That model is excellent technology, painful licensing: it took H.265 years longer than expected to reach broad deployment because pools disagreed on royalty terms, and H.266 today has no browser support six years after spec finalisation because the same fight is repeating. 13

For a streaming operator in 2026 the practical takeaway is:

  • You will probably never adopt H.267 for open-web delivery. It will live in broadcast contribution, premium VOD with paid client apps, and broadcast distribution to capable smart TVs.
  • You will adopt H.265 and H.266 only where AV1 / AV2 cannot reach. That mostly means HEVC for Apple devices without AV1 decode (Apple ships HEVC hardware decode universally), the existing premium VOD pipeline, and VVC in broadcast.
  • You will watch H.267 for screen-content workloads — remote desktop, cloud gaming, screen-share in video conferencing — because that is where ECM's gain over H.266 is largest: about 43% in ECM 20.

The thing not to do is to schedule a pipeline migration around H.267. Its timeline is too far out, its licensing too uncertain, and AV2 will be doing most of the same job about three years earlier, for free.

Track three — End-to-end learned compression

This is the track that will surprise the most teams in the late 2020s. End-to-end learned video compression replaces the entire hand-engineered hybrid codec pipeline with a single neural network. The network is trained on millions of hours of video to produce a compact internal representation (called a latent, the compressed-bits side of the network) and reconstruct the original frames from it. The compressed file is whatever the network produces; the decoder is the rest of the network.

The earliest end-to-end learned image codec that beat JPEG2000 was published by Ballé et al. in 2017. From 2021 (NeurIPS) the Microsoft Research DCVC family extended the approach from images to video. By 2024 the DCVC-FM (feature modulation) variant reached the quality of the H.266 reference encoder. DCVC-RT (real-time) was posted on arXiv in February 2025 and presented at CVPR in June 2025, and it is the milestone that matters most for any practical conversation about neural codecs in 2026. 15 16 17

What DCVC-RT actually delivers

DCVC-RT is the first end-to-end learned video codec that is both competitive with H.266 on compression and fast enough to run in real time on a single GPU. The specific numbers from the CVPR 2025 paper:

  • Average encoding speed: 125.2 fps for 1080p video on an NVIDIA A100.
  • Average decoding speed: 112.8 fps for 1080p video on an NVIDIA A100.
  • BD-rate of −21.0% versus the H.266/VTM reference encoder — that is, DCVC-RT needs 21% fewer bits than H.266 to reach the same quality, measured in PSNR. 15 18

A higher-capacity variant, DCVC-RT-Large, pushes BD-rate to −30.8% while staying near real-time. The same paper reports approximately 40 fps encoding and 34 fps decoding for 1080p on a consumer NVIDIA RTX 2080 Ti — three generations old at the time of publication. On the latest consumer GPUs, real-time HD playback is comfortable. 18

In other words: the gap between "interesting research demo" and "you could ship this in a desktop application" closed in 2025 for high-end consumers, and in 2026 it closed for laptops with NPUs; on phones it is still open.

2026: the first learned codec in a shipping product — MLVC

In July 2026 Microsoft open-sourced MLVC (Multi-platform Learned Video Codec) under the MIT license, with code and model weights on GitHub. The paper, posted on 26 June 2026, reports a BD-rate improvement of more than 70%, measured on mean opinion score, over hardware HEVC on a video-conferencing benchmark, with encoder and decoder both averaging 100 fps on commodity NPUs from Apple, Intel and Qualcomm. 35 Microsoft uses it in Teams peer-to-peer calls, according to Streaming Learning Center's August 2026 analysis. 37

Read that 70% with its caveats. The baseline is a hardware HEVC encoder at conferencing bitrates, not a tuned software encoder. Full 1080p real-time playback is reached on Apple hardware only, Microsoft publishes no power figures, and a participant without a compatible NPU falls back to H.264, simulcast or cloud transcoding. The MIT license covers copyright only, and neural video coding has no patent pool yet. Speed keeps improving: Microsoft Research's CVPR 2026 paper on DCVC-UF codes a chunk of several frames into one latent to push encoding and decoding faster. 38

Scatter chart of BD-rate against the H.266 reference encoder VTM, lower is better, from plus 10% to minus 30%, against encoding speed at 1080p on a log scale from 1 to 200 frames per second, with a dashed line at 30 fps marking real time. Traditional hybrid encoders are navy dots: vvenc medium at about 2 fps and 0%, libaom cpu-used 4 at 3 fps and minus 4.5%, SVT-AV1 preset 4 at 10 fps and minus 4.5%, SVT-AV1 preset 8 at 25 fps and plus 5.5%, x265 medium at 50 fps and plus 10.5%. Learned encoders are purple squares: DCVC-FM at 5 fps and minus 14.5%, DCVC-RT at 125 fps and minus 21%, DCVC-RT-Large at about 60 fps and minus 31%. The two DCVC-RT points sit alone in the fast, efficient lower-right corner. A closing line says DCVC-RT needs 21% fewer bits than H.266 at 125 fps on one NVIDIA A100. Figure 3. The compression-versus-speed map of leading video encoders in 2026. Traditional hybrid encoders (x265, SVT-AV1, libaom, vvenc) cover the wide left-to-right speed-quality trade-off. The DCVC family of learned encoders has reached the same Pareto frontier as H.266 and is climbing fast.

Why learned codecs are not a drop-in replacement

The headline numbers above are encouraging, but four properties of learned codecs make them genuinely different from traditional codecs, and these properties define the practical timeline for adoption.

The compressed bitstream is not a fixed specification. A learned codec is a trained model. Two different versions of "DCVC" are two different codecs. There is no equivalent of "the H.264 spec" against which any decoder can validate; there is the specific model weights that produced the file. This is the single biggest cultural change. The industry has built thirty years of interoperability on top of fixed bitstream specifications. Learned codecs throw that frame away. Standards bodies (JPEG, JVET) are now writing specifications for how to ship model weights alongside the bitstream so that interoperability becomes possible at all. JPEG AI, the first learned image-coding standard, was published as ISO/IEC 6048-1 in September 2025. 39 20

Decoding is compute-heavy and floating-point. Traditional codecs decode in integer arithmetic on small, well-understood DSP blocks. Neural codecs decode by running a deep network forward, which historically needed floating-point arithmetic on a GPU. Quantising the model to integer precision so it can run on mobile NPUs is an active research area; floating-point cross-platform reproducibility is famously hard. 21 22

Mobile support is still early. MLVC runs on laptop and desktop NPUs, but on phones real-time 1080p neural decoding is still shown only by research prototypes such as Qualcomm's MobileNVC and MobileCodec, and battery cost is the open question. 23 24

The compressed-file size is not the only metric anymore. Some learned codecs (perceptual / generative variants such as HiFiC for images, and emerging generative video codecs) deliberately synthesise reconstructed content that looks like the original but is not pixel-accurate. This is a useful trade-off for some applications (entertainment streaming) and unacceptable for others (medical imaging, surveillance forensics, broadcast contribution). Procurement of a learned codec has to ask "what kind of reconstruction does it produce" alongside "what BD-rate does it achieve". Standards are already making room for generative coding: JVET has added Generative Face Video SEI messages, which let a decoder rebuild a talking face from a compact set of facial features, to the working draft of VSEI version 4 and to the next versions of H.264 and H.265. 40

Where learned codecs will land first

The way neural codecs will actually enter the streaming stack is not as a full replacement for AV1 or H.265 in 2027. It is in three more practical patterns:

  1. Neural blocks inside hybrid codecs. JVET's NNVC test model already adds neural-network in-loop filters and intra prediction on top of VVC, and these tools are candidates for H.267. AV2 has data-driven intra modes. By H.268 / AV3, almost every box in the hybrid pipeline will have a neural alternative the encoder can choose. The bitstream stays interoperable; the gains accrue gradually.
  2. Pre- and post-processing wrappers. Tools like Deep Render's neural pre-encoder and the SimaBit pre-processing pipeline use neural networks to reshape pixels before a traditional codec runs, recovering 20–30% bits without changing the downstream decoder. These ship today and will be the dominant production pattern through 2027–2028. 9 26
  3. Niche end-to-end deployment. Closed-ecosystem applications where one team controls both encoder and decoder — cloud gaming with a custom client, video conferencing with a managed app, machine-to-machine video for ADAS and surveillance — will be the first to ship pure end-to-end learned codecs. Video conferencing got there first: MLVC in Teams calls is exactly this pattern. Open-web streaming will be the last.

The realistic projection for the open-web: pure end-to-end learned codecs ship in production browsers some time around 2030–2032, and they will arrive labelled as a standard (a JPEG AI-based video standard, an MPEG-AI, or a successor JVET project) with shipped model weights rather than as a per-vendor proprietary format.

Decision tree that starts from the question, planning horizon, and walks down three horizons, each pointing right to four actions. Now to 2027, plan the AV2 move: keep AV1 as the default delivery codec, benchmark AV2 on your catalogue with AVM, test neural pre-processing such as SimaBit and Deep Render, and audit your bitrate ladder against AV2 numbers. 2028 to 2029, AV2 mainstream: ship AV2 to all hardware-capable devices, keep AV1 as the fallback for the long tail, add neural pre-processing for premium tiers, and evaluate H.266 for broadcast delivery. 2030 and beyond, watch and evaluate: evaluate H.267 or ECM for premium VOD only, watch end-to-end neural codecs in closed apps, plan AV2 hardware encoders for transcoding, and reassess neural decode coverage on mobile NPUs. Figure 4. A planning decision tree for codec roadmaps through 2030. The recommendation in 2026 is to plan an AV2 transition, evaluate neural pre-processing on top of your existing pipeline, and treat H.267 and pure end-to-end neural codecs as horizon items rather than near-term commitments.

Comparative landscape — the four codecs that will matter through 2030

If you are reading this in 2026 and trying to plan capital allocation for the next four years of video infrastructure, four codecs sit on the table. We summarise the situation as you should walk into a budget meeting with it.

Codec Available Coding gain Licensing Decoder support 2026 When to use
AV1 2018 reference for this row royalty-free, AOMedia Patent License broad — phones, TVs, browsers Default open-web delivery today
H.266 / VVC 2020 ~40–50% vs H.265 patent pools (Access Advance and others) sparse — some TV and set-top chips, no browsers Broadcast (DVB, Brazil TV 3.0), contribution
AV2 2026 (spec 1.0, 28 May) ~30% vs AV1, ~75% vs H.264 royalty-free, AOMedia Patent License software only (AVM, dav2d); first chips 2027 Next-gen open-web delivery, plan the move now
ECM / H.267 Call for Proposals July 2026; v1 target Oct 2029 ~28% vs H.266 today (ECM 20), 40% target patent pools none in 2026 Plan no further than evaluation for 2029+ premium VOD and screen content
Learned codecs (DCVC-RT, MLVC) 2025–2026, no video standard ~−21% to −30% vs H.266 open-source code (MLVC, MIT), no patent pool; JPEG AI covers images only GPU; laptop NPUs for MLVC Pre-processing wrapper, closed-ecosystem apps

The lines that read "available" describe when the bitstream is locked, not when broad deployment is realistic. For royalty-free codecs, add about three years to "available" for "broad consumer hardware decoder coverage", and another two for "you can drop your AV1 fallback"; patent-pooled codecs have taken much longer. For head-to-head numbers between individual codecs, such as H.266 vs AV2, see our video codec comparison table.

Common mistake — extrapolating from BD-rate alone

Every benchmark on the table above reports a single BD-rate number versus a reference codec. The most common procurement mistake we see is to take those numbers literally and assume that the codec with the better BD-rate is the better business choice.

It is not. Codec decisions are joint decisions about compression, decoder availability, licensing cost, encoder cost, ecosystem maturity, and time-to-market. A 40% BD-rate gain on a codec whose decoders exist in 2030 is worth less than a 30% gain on a codec whose decoders arrive in 2027, because the bandwidth savings are integrated over the four years of difference. Walk into every benchmark conversation with the integrated cost in mind, and let the BD-rate number be one input among many.

A worked example — the bitrate ladder for a 1080p stream in 2030

Concrete numbers anchor the discussion. Take a 30-minute episode of a typical drama, encoded at 1080p / 24 fps for OTT delivery. Assume a top-rung target of VMAF 95 (broadcast quality) and a baseline of 4.0 Mbps on H.264 today. We can project the bitrate at the same quality through each generation:

H.264         baseline                 = 4.0 Mbps
H.265 / HEVC  −50% vs H.264            = 2.0 Mbps
AV1           −30% vs H.265 (~−65% vs H.264) = 1.4 Mbps
AV2           −30% vs AV1               = 1.0 Mbps
H.266 / VVC   −40% vs H.265 (~−70% vs H.264) = 1.2 Mbps
H.267 / ECM   −40% vs H.266            = 0.72 Mbps
DCVC-RT       −21% vs H.266            = 0.95 Mbps
DCVC-RT-Large −30.8% vs H.266          = 0.83 Mbps

For a service with 10 million MAU streaming 5 hours per month at 1080p, the per-user monthly egress at H.264 is 5 × 60 × 60 × 4 Mbps ÷ 8 ≈ 9 GB. At AV2 that drops to 9 × (1.0 / 4.0) = 2.25 GB. Across 10 million users that is 67.5 PB of monthly egress eliminated compared to a stuck-on-H.264 pipeline. At a typical CDN egress price of 0.5–1 cent per GB at scale, the AV2 transition is worth several million dollars per year for a service of that size. The further codec hops (DCVC-RT, H.267) keep adding margin, but the first hop — from a legacy H.264 / H.265 stack to AV1/AV2 — is the one with the dominant payoff. 25

This is the headline reason the AV1-then-AV2 path is the priority roadmap for any operator in 2026 — AV1 now, AV2 as hardware arrives — even though H.267 and end-to-end learned codecs are technically more interesting on paper. The bandwidth bill is paid in money today, not in slides about 2030.

Frequently asked questions

What is a neural video codec?

A neural video codec compresses video with trained neural networks instead of hand-designed coding tools: an encoder network maps frames to a compact latent, and a decoder network rebuilds them. The best ones, such as Microsoft Research's DCVC-RT (CVPR 2025), need about 21% fewer bits than H.266 for the same quality.

What is the next-generation video codec after AV1?

AV2. AOMedia finalised the AV2 1.0.0 specification on 28 May 2026; it needs about 30% less bitrate than AV1 for the same quality, and the first AV2 hardware decoders are expected in chips from 2027.

Does FFmpeg support AV2?

Not yet. As of October 2026 FFmpeg has no AV2 decoder or encoder, so you encode and decode with AOMedia's AVM reference tools or VideoLAN's dav2d decoder and pass Y4M files to FFmpeg for everything else.

When will H.267 be released?

JVET issued the Call for Proposals on 17 July 2026 and targets the first version of the standard for October 2029. Broad device support is unlikely before the mid-2030s.

Will video codecs use generative AI in the future?

Yes, and partly they already do. Neural tools are entering hybrid codecs, JVET is standardising generative face-video SEI messages, and generative codecs that synthesise plausible detail suit entertainment but not medical or forensic video, where pixels must stay exact.

Are neural video codecs used in production today?

Yes, in closed ecosystems. Microsoft's MLVC, open-sourced in July 2026, runs on Apple, Intel and Qualcomm NPUs in Teams calls; open-web streaming still relies on AV1, HEVC and H.264.

Should I wait for AV2 or H.267 before upgrading my pipeline?

No. Move the bulk of your traffic to AV1 now, benchmark AV2 on your own content with the reference tools, and treat H.267 and pure neural codecs as horizon items.

↓ Free download: Future Codecs Roadmap ChecklistOne page for your planning meeting: the AV2 migration steps, how to evaluate neural pre-processing on your own catalogue, and the H.267 and learned-codec watch-list through 2030.

Where Fora Soft fits in

We have built video pipelines for OTT, video conferencing, video surveillance, telemedicine, e-learning, and AR/VR customers since 2005, across 250+ projects. The pattern we see most often in customer planning conversations is the inverse of the pattern in the trade press: customers are over-rotated on the codec that will arrive in 2030 and under-rotated on the codec whose specification is already final (AV2) and the AV1 rollout that comes before it. The practical 2026 task for almost every operator is to run a same-VMAF AV1-versus-AV2 evaluation against their own catalogue, build a hardware decode roadmap by device family, and design a fallback ladder that ships AV2 first to capable devices and AV1 to the rest. We are happy to run that evaluation, design the ladder, or audit one a customer's team has built — none of those engagements need to start with a codec licence purchase. That evaluation is part of our video encoding and processing development work, and the resulting multi-codec ladder ships through the streaming software we build for OTT and live products. If you run a calling product, the same evaluation covers learned codecs such as MLVC: our video conferencing app development team can add an NPU-decoded path next to an H.264 fallback and measure where it pays off.

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

Volume 1 walks the codec family tree from MPEG-2 to AV1, VVC and AV2, and shows which block of the hybrid pipeline each generation improved to earn its 30–50%. It also teaches you to read a BD-rate chart and run a same-VMAF comparison on your own content, so you can judge an AV2, H.267 or neural-codec claim before it reaches your roadmap. Written by Nikolay Sapunov, CEO at Fora Soft.

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

When should your service move to AV2, and what will it save on your CDN bill?

Send us your current ladder, codec mix and device split — we will size an AV1-versus-AV2 evaluation on your own catalogue, sketch the hardware decode roadmap by device family, and tell you where neural pre-processing pays off first. 250+ video products shipped since 2005.

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

See our case studies

References


  1. Alliance for Open Media, "AOMedia Announces Year-End Launch of Next Generation Video Codec AV2 on 10th Anniversary", September 2025. https://aomedia.org/press%20releases/AOMedia-Announces-Year-End-Launch-of-Next-Generation-Video-Codec-AV2-on-10th-Anniversary/ ↩↩

  2. Alliance for Open Media, "AV2: What the Final Specification Means for Implementers", July 2026. https://aomedia.org/blog%20posts/AV2-What-the-Final-Specification-Means-for-Implementers/ ↩

  3. CNX Software, "AOMedia AV2 video codec draft specification release, and a quick try at the reference implementation", February 2026. https://www.cnx-software.com/2026/02/03/aomedia-av2-video-codec-draft-specification-release-and-a-quick-try-at-the-reference-implementation/ ↩

  4. Phoronix, "AOMedia To Release AV2 Video Codec At Year's End", 2025. https://www.phoronix.com/news/AOMedia-AV2-Coming-Soon ↩

  5. IAM Media, "AOMedia announces launch of 'royalty-free' AV2 codec", 2026. https://www.iam-media.com/article/aomedia-announces-launch-of-royalty-free-av2-codec ↩↩

  6. Alpha1Convert, "AV2 Codec: Architecture, Benchmarks, and the Road to Replacing AV1", 2026. https://www.alpha1convert.com/av2-codec-architecture-benchmarks-and-the-road-to-replacing-av1/ ↩↩↩

  7. Muvi, "AV1 vs AV2: The Next Generation Video Codec Battle Explained", 2026. https://www.muvi.com/blogs/av1-vs-av2-the-next-generation-video-codec-battle-explained/ ↩↩↩

  8. Andrey Norkin, "AV2 Video Codec Architecture", QoMEX 2025. https://norkin.org/pdf/QoMEX_2025_AV2_Architecture_slides.pdf ↩

  9. SimaLabs, "AV2 vs. AV1: 2025 4K-Streaming Benchmarks Using Netflix Open Content & SimaBit Pre-processing", 2025. https://www.simalabs.ai/resources/av2-vs-av1-2025-4k-streaming-benchmarks-netflix-simabit-preprocessing ↩↩

  10. Cloud News, "AV2: AOMedia's New Open Codec Aiming to Lead the Next Streaming Era", 2026. https://cloudnews.tech/av2-aomedias-new-open-codec-aiming-to-lead-the-next-streaming-era-against-h-264-hevc-vp9-and-av1/ ↩

  11. Streaming Media, "H.267: A Codec for (One Possible) Future", 2025. https://www.streamingmedia.com/Articles/News/Online-Video-News/H.267-A-Codec-for-(One-Possible)-Future-167889.aspx ↩↩

  12. Jan Ozer, "AI Video Compression Standards: Who's Doing What and When", Streaming Learning Center, 2025. https://streaminglearningcenter.com/codecs/ai-video-compression-standards-whos-doing-what-and-when.html ↩↩

  13. Streaming Media, "Independent Report Compares Codec Royalty Costs Across Two Major Licensing Pools", 2025. https://www.streamingmedia.com/Articles/News/Online-Video-News/Independent-Report-Compares-Codec-Royalty-Costs-Across-Two-Major-Licensing-Pools-174221.aspx ↩

  14. Kiho Choi, "Neural Network based Video Coding in JVET", 2023. https://www.kibme.org/resources/journal/20230118120317089.pdf ↩

  15. Zhaoyang Jia et al., "Towards Practical Real-Time Neural Video Compression", arXiv 2502.20762, 2025. https://arxiv.org/abs/2502.20762 ↩↩

  16. Microsoft, "DCVC: Deep Contextual Video Compression" repository (DCVC family, including DCVC-RT), GitHub. https://github.com/microsoft/DCVC ↩

  17. Jia et al., "Towards Practical Real-Time Neural Video Compression", CVPR 2025. https://openaccess.thecvf.com/content/CVPR2025/papers/Jia_Towards_Practical_Real-Time_Neural_Video_Compression_CVPR_2025_paper.pdf ↩↩

  18. Streaming Learning Center, "Evaluating DCVC-RT: A Real-Time Neural Video Codec That Delivers on Speed and Compression", 2025. https://streaminglearningcenter.com/articles/evaluating-dcvc-rt-a-real-time-neural-video-codec-that-delivers-on-speed-and-compression.html ↩↩↩

  19. Streaming Media, "The State of Streaming Codecs 2026", 2026. https://www.streamingmedia.com/Articles/Editorial/Featured-Articles/The-State-of-Streaming-Codecs-2026-173838.aspx ↩

  20. JPEG, "106th Meeting – Online – JPEG AI becomes an International Standard", February 2025. https://jpeg.org/items/20250219_press.html ↩

  21. Tian et al., "Towards Real-Time Neural Video Codec for Cross-Platform Application Using Calibration Information", ACM MM 2023. https://arxiv.org/abs/2309.11276 ↩

  22. Sun et al., "On the Quantization of Neural Video Codecs", OpenReview, 2024. https://openreview.net/forum?id=dLqDqzlDxZ ↩

  23. van Rozendaal et al., "MobileNVC: Real-time 1080p Neural Video Compression on a Mobile Device", WACV 2024. https://arxiv.org/abs/2310.01258 ↩

  24. Le et al., "MobileCodec: Neural Inter-frame Video Compression on Mobile Devices", ACM MMSys 2022. https://arxiv.org/abs/2207.08338 ↩

  25. Transcodely, "AV1 in 2026: Adoption, Hardware Support, and What It Means for Your Pipeline", 2026. https://www.transcodely.com/blog/av1-in-2026 ↩

  26. Streaming Learning Center, "Deep Render AI Codec Running in FFmpeg and VLC", 2025. https://streaminglearningcenter.com/codecs/deep-render-ai-codec-running-in-ffmpeg-and-vlc.html ↩

  27. Alliance for Open Media, "Alliance for Open Media Releases AV2 Codec, Advancing Next-Generation Open Video Coding", June 2026. https://aomedia.org/press%20releases/Alliance-for-Open-Media-Releases-AV2-Codec/ ↩

  28. Alliance for Open Media, "AV2 Bitstream & Decoding Process Specification v1.0.0" and AVM v1.0.0 reference software, 28 May 2026. https://av2.aomedia.org/ ↩

  29. Jean-Baptiste Kempf, "Let dav2d be", VideoLAN, 31 May 2026. https://jbkempf.com/blog/2026/dav2d/ ↩

  30. Alliance for Open Media, "Demonstrating Real-Time AV2 Decoding on Consumer Laptops", March 2026. http://aomedia.org/blog%20posts/Demonstrating-Real-Time-AV2-Decoding-on-Consumer-Laptops/ ↩

  31. FFmpeg, "General Documentation: supported file formats, codecs or features" (Video Codecs table), checked 14 October 2026. https://www.ffmpeg.org/general.html ↩

  32. Alliance for Open Media, "AOMedia Members Showcase AV2 Ecosystem Developments at IBC 2026", September 2026. http://aomedia.org/press%20releases/AOMedia-Showcase-AV2-Ecosystem-Developments-at-IBC-2026/ ↩↩

  33. Ofinno, "JVET Issues the Joint Call for Proposals for Video Coding Beyond VVC", 2026. https://ofinno.com/standards-readout/jvet-issues-the-joint-call-for-proposals-for-video-coding-beyond-vvc/ ↩↩

  34. ITU-T, "Joint Video Experts Team (JVET)", study period 2025–2028. https://www.itu.int/en/ITU-T/studygroups/2025-2028/21/video/Pages/jvet.aspx ↩

  35. "MLVC: Multi-platform Learned Video Codec for Real-World Deployment", arXiv 2606.28027, June 2026. https://arxiv.org/abs/2606.28027 ↩

  36. Microsoft Tech Community, "Announcing the Open-Source Release of ML Video Codec (MLVC)", July 2026. https://techcommunity.microsoft.com/blog/linuxandopensourceblog/announcing-the-open-source-release-of-ml-video-codec-mlvc/4539875 ↩

  37. Jan Ozer, "Microsoft MLVC: Interoperability, Power Realities, and the Neural Compression Patent Landscape", Streaming Learning Center, August 2026. https://streaminglearningcenter.com/articles/microsoft_mlvc.html ↩↩

  38. Jiahao Li et al., "Ultra-Fast Neural Video Compression", CVPR 2026. https://openaccess.thecvf.com/content/CVPR2026/html/Li_Ultra-Fast_Neural_Video_Compression_CVPR_2026_paper.html ↩

  39. JPEG, "JPEG AI": Version 1 published as ITU-T T.840.1 | ISO/IEC 6048-1:2025 (September 2025). https://jpeg.org/jpegai/ ↩

  40. "Standardizing Generative Face Video Compression using Supplemental Enhancement Information", arXiv 2410.15105, v3 September 2025. https://arxiv.org/abs/2410.15105 ↩

  41. Access Advance, "Access Advance and Via Licensing Alliance Announce HEVC/VVC Program Acquisition", 15 December 2025. https://accessadvance.com/2025/12/15/access-advance-and-via-licensing-alliance-announce-hevc-vvc-program-acquisition/ ↩

  42. Access Advance, "Access Advance Launches Exploration Phase for an AV1/AV2 Device and Software Patent Pool", 25 August 2026. https://accessadvance.com/2026/08/25/access-advance-launches-exploration-phase-for-an-av1-av2-device-and-software-patent-pool-invites-participation-of-potential-licensors/ ↩

  43. VeriSilicon, "VC9800D AV2 hardware video decoder IP", press release, June 2026. https://www.verisilicon.com/en/PressRelease/VC9800DAV2 ↩