Walking away to make a sandwich while a video buffers, only to return to 30 seconds of pixelated mush, feels like an ancient internet memory. Video codecs still have to compress raw video files so your phone, computer, or TV can decode the picture in real time. Choosing between AV1 and AV2 right now isn’t a debate over which codec compresses better, because the answer depends entirely on when your project ships.
Both formats are open-source, royalty-free codecs built by the Alliance for Open Media (AOMedia). As breakdowns from the Muvi YouTube channel and ToolWise AI, hosted by Dario, lay out, AV2 wins every efficiency metric on paper, while AV1 wins on actual hardware and ecosystem availability until 2027 to 2028. If your streaming pipeline or encoding project needs to run on devices people already own, you build on AV1 today and treat AV2 as a multi-year architecture target.
Key Takeaways
AV1 (released in 2018) is the practical, production-ready codec for streaming and encoding today, backed by widespread hardware support across Intel Arc, AMD Radeon 7000, NVIDIA GeForce RTX 40 series, and Apple M3 chips.
AV2 cuts bitrates by roughly 30% over AV1 at equal visual quality in early tests, but software encoding requires 2 to 3 times the computational power and pushes even a GeForce RTX 5090 to its limits.
Because hardware acceleration requires physical silicon rather than a driver update, 18 to 24 months of chip design plus 6 to 12 months of device integration places widespread consumer AV2 hardware support in the 2027 to 2028 window.
Table of Contents
AV1 vs AV2 at a glance: comparison table and verdict
AV1 and AV2 are two generations of open-source, royalty-free video codecs from AOMedia, where AV1 is the production-ready choice for shipping today and AV2 is a next-generation successor suited for roadmaps targeting 2027 to 2028 hardware. Both use AOMedia’s royalty-free baseline: AV1 shipped in 2018, while AV2 entered development in 2020 and targets specification finalization for the end of 2025. Early testing shows AV2 delivers roughly a 30% bitrate reduction at equal visual quality compared to AV1 (against a 30% to 40% development target) and 50% to 60% total data savings over HEVC (H.265), at the cost of 2 to 3 times more computational power during encoding.
| Feature or Metric | AV1 | AV2 |
|---|---|---|
| Release / Spec Timeline | 2018 release | 2020 dev start; end of 2025 spec target |
| Compression Efficiency | ~30% smaller than HEVC; ~50% smaller than VP9/H.264 | ~30% smaller than AV1 in early tests (30%-40% target); ~50%-60% smaller than HEVC |
| CPU Encoding Compute Cost | 1x baseline | 2x-3x AV1 compute (2x-3x longer on standard CPUs; 50x-100x slower than VVC) |
| Decoding Speed vs VVC (H.266) | Partial (hardware-accelerated on modern GPUs) | 8x-10x slower than VVC in software |
| Consumer Hardware Acceleration | Yes (widespread GPU, CPU, and mobile silicon today) | No (widespread hardware expected 2027-2028 or later) |
| Maximum Superblock Size | 128×128 pixels | 256×256 pixels |
| HDR Metadata Support | Basic HDR10 | HDR10+ with scene-by-scene dynamic metadata |
| Split-Screen & Dynamic Bitrate Switching | No | Yes (native split-screen delivery and dynamic bitrate switching) |
| Enhanced AR/VR & 360-Degree Video | Partial (standard video stream support) | Yes (dedicated low-latency and separate-eye VR optimizations) |
AV2 isn’t broadly available as a consumer playback format yet, and its largest compression gains concentrate in 4K and 8K content while demanding high-end hardware just to survive software encoding. Although one optimistic industry projection suggested AV2 could reshape streaming by 2026 during its early software era, real-world consumer playback depends on dedicated decoder chips that won’t reach volume devices until 2027 to 2028.
What is AV1 and why it runs the streaming world today
AOMedia founded the AV1 project in 2015 to break free from expensive HEVC patent pools,
Origins and licensing
In 2015, rival tech giants Google, Apple, Microsoft, Amazon, Netflix, and Samsung formed AOMedia to escape the costly, fragmented patent-pool licensing surrounding HEVC. They released AV1 in 2018 as an open-source, royalty-free video codec. At launch, AV1 delivered roughly 30% better compression than HEVC and about 50% better compression than older VP9 and H.264 streams. It operates primarily as a lossy codec with lossless options for retaining critical detail, packing specialized tools like film grain synthesis and constrained directional enhancement filters.
Where AV1 runs today
In an AOMedia member survey, 88% of member organizations rated AV1 as critical or important to their operations, As of 2024, YouTube, Netflix, and Vimeo actively serve AV1 streams, backed by hardware encode and decode acceleration in Intel Arc graphics (Arc Alchemist), AMD Radeon 7000 series GPUs, Google Pixel 8 phones, NVIDIA GeForce RTX 40 series cards, and Intel Meteor Lake processors. NVIDIA Kepler GPUs also speed up AV1 decoding, while Apple M3 and newer chips support AV1 hardware decoding only, with no hardware encoding.
What is AV2: the successor codec and its 30-40% goal
AOMedia kicked off AV2 development in 2020 as a full generational leap over AV1, setting an explicit engineering target of 30% to 40% better compression efficiency and targeting the end of 2025 for specification finalization. A finalized specification means the mathematical rules are locked so compiler writers and chip architects can get to work, not that consumer televisions or browsers can decode the stream. No broad consumer playback ecosystem exists for AV2 yet.
Like its predecessor, AV2 carries AOMedia’s open-source, royalty-free licensing policy alongside free reference encoders to keep adoption friction low for developers. Open licensing doesn’t erase every legal headache, though. Enterprise teams still have to weigh lingering third-party patent claims against both AV1 and AV2 before rolling out new pipelines.
Compression efficiency: how much smaller are AV2 files really
AV2 achieves roughly a 30% bitrate reduction at equal visual quality compared to AV1 in early testing, sitting at the lower bound of AOMedia’s 30% to 40% development target while delivering 50% to 60% cumulative data savings compared to HEVC. AV1 already cuts file sizes by roughly 30% over HEVC and 50% over VP9 and H.264, and that’s before AV2’s next-generation prediction tools even touch the stream.
AOMedia set the 30% to 40% engineering target in 2020, while early test encodes show a ~30% bitrate drop against AV1. The 50% to 60% figure measures the multi-generational gap all the way back to HEVC. Actual file-size reductions shift heavily based on encoder settings, source content complexity, and playback requirements, with AV2’s biggest efficiency wins showing up on high-resolution 4K and 8K video.
Inside the AV2 specification: the tools behind the gains
The AV2 specification is AOMedia’s next-generation video compression architecture, distinguished by 256×256 pixel superblocks, seven-frame temporal interpolation prediction, and edge-aware in-loop filtering paired with upgraded entropy coding.
Superblocks and recursive partitioning
AV2 bumps the maximum block size from AV1’s 128×128 pixels up to 256×256 superblocks, using recursive partitioning to slice them up on the fly. When an encoder hits a flat blue sky or a dark horror-film wall, it can describe that entire chunk of screen with one large block instead of logging thousands of tiny entries. When the camera pans across complex textures like an actor’s eyes, recursive partitioning subdivides the frame right down into tiny detail blocks.
Temporal interpolation prediction
For motion across frames, temporal interpolation prediction builds on multi-reference frame prediction and sub-pixel motion compensation. Instead of comparing a frame against just one past frame, it references up to seven past frames to mathematically predict pixel trajectories, such as tracking a speeding car across seven frames and storing only the difference from that prediction, keeping fast-action scenes much sharper.
Filter tools
To clean up compression artifacts without turning the image into soup, AV2 combines enhanced intra-frame prediction and context-adaptive binary arithmetic coding (CABAC) with adaptive quantization alongside new in-loop filters. The guided detail filter acts like an automatic cleaning crew that uses the image itself as a map to polish flat areas while protecting sharp edges, working with cross component sample offset to cut color banding and noise.
Encoding and decoding cost: the compute ladder
AV2 encoding requires 2 to 3 times more computing power than AV1 on standard CPUs, taking 2 to 3 times longer to finish an encode and practically requiring high-end hardware like the GeForce RTX 5090. While an RTX 4080 handles AV1 decoding comfortably, heavy AV2 encoding pushes even the RTX 5090 to its limits, and in software benchmarks against VVC (H.266), AV2 encodes 50 to 100 times slower and decodes 8 to 10 times slower. Across the next-generation codec landscape, AV2’s advanced intra-frame and inter-frame prediction tools trade raw compute speed for royalty-free compression gains. Still, to keep custom silicon practical despite that software tax, AV2 was designed to be hardware-friendly from day one with chip designers like Apple, Qualcomm, and Intel in mind, targeting a hardware complexity goal of less than double that of AV1.
| Compute Metric | AV1 | AV2 | VVC (H.266) |
|---|---|---|---|
| CPU Encoding Compute vs AV1 | 1x baseline | 2x-3x compute (2x-3x longer) | 50x-100x faster than AV2 |
| Decoding Speed vs VVC | Hardware-accelerated on RTX 4080 and modern GPUs | 8x-10x slower than VVC | 1x VVC baseline |
| Silicon Complexity Target vs AV1 | 1x baseline | Under 2x AV1 complexity | Patent-pool silicon implementation |
At GeekExtreme, we know how early software-codec tests usually end on a workbench: you fire up a next-generation encode on the machine you already own, watch the speed drop to a crawl 2 to 3 times slower than AV1, and immediately decide to wait for dedicated hardware encoders. To keep that compute wall from becoming permanent in silicon, AOMedia set a strict hardware complexity target for AV2 of less than twice that of AV1, designing the spec from day one with chip architects from Apple, Qualcomm, and Intel in the room.
AV2 hardware support: new silicon required, not a driver update
No shipping consumer GPU or smartphone has dedicated AV2 hardware decoding today, which locks mass-market hardware rollouts into the 2027 to 2028 silicon window.
Why existing GPUs can’t gain support
No, AV2 hardware support doesn’t exist yet on current GPUs or phones, making AV1 the practical recommendation for targeting today’s devices. Codec hardware acceleration requires new physical silicon rather than a software patch, so existing hardware, even high-end cards like the NVIDIA GeForce RTX 3080 or RTX 5070 Ti, can’t gain hardware AV2 decoding through a driver update. While AV1 hardware acceleration is already baked into Intel Arc, AMD Radeon 7000, RTX 40 series, Google Pixel 8, Intel Meteor Lake, AMD Ryzen AI, and Apple M3 chips, low-power devices without dedicated AV2 silicon will drop frames and drain batteries faster when decoding AV2 streams.
The 2027-2028 hardware timeline
AV2 targeted specification finalization for the end of 2025, putting widespread consumer hardware support in the 2027 to 2028 window or later. Once a codec spec locks, semiconductor fabs still have to run the physical chip design cycle, which takes 18 to 24 months of new silicon design followed by 6 to 12 months of device integration.
One optimistic source projected AV2 would reshape streaming by 2026 during its initial software phase as a bridge toward H.267 in 2028. In practice, forcing low-power phones to decode AV2 in software causes dropped frames and drains batteries fast, much like heavy TikTok browsing, whereas dedicated hardware decoding barely wakes the processor. Apple’s slow AV1 rollout is a sobering preview: Apple only added AV1 hardware decode starting with the M3 chip, and Safari’s limited AV1 support still affects up to 1.6 billion devices.
Bandwidth and quality: what the savings feel like
Raw, uncompressed 4K video at 60fps packs over 8 million pixels per frame, chewing through hundreds of gigabytes in a few minutes and destroying a monthly mobile data plan. AV2 drops a compressed 4K stream from roughly 25 Mbps down to about 17.5 Mbps. That makes 4K reliable on rural lines or shaky Wi-Fi, and it shrinks Blu-ray-quality files while cleaning up banding, gradients, and dark scenes. With 8K demanding four times the data of 4K, those savings are critical, and one speculative projection estimates Netflix, where bandwidth is the biggest expense after content production, could cut its bandwidth bills by 30% to 50%, saving billions of dollars.
New AV2-only streaming features AV1 never had
While compression ratios grab the headlines, AOMedia also packed AV2 with native multi-stream, adaptive delivery, and dynamic metadata tools that AV1 never supported. AV2 upgrades streaming pipelines over AV1 in four specific areas:
- Split-screen delivery: Unlike AV1, AV2 natively streams multiple programs onto a single display using dynamic bitrate allocation between the segments, purpose-built for multi-angle live sports, gaming tournaments, and virtual events.
- Adaptive and screen-content streaming: AV2 adds dynamic bitrate switching for multi-resolution streaming alongside sharp text and UI optimizations for enterprise conferencing apps like Zoom, education platforms, and software demos.
- HDR10+ dynamic metadata: Where AV1 supports basic HDR10 for 10-bit video, AV2 upgrades 10-bit video pipelines to HDR10+ with scene-by-scene dynamic metadata.
- AR/VR, 360-degree, and workflow hooks: With dedicated AR/VR support, AV2 lowers latency and delivers separate eye views for VR headsets (which have to push two views or full 360-degree worlds for spatial computing, the metaverse, and speculative holographic meetings requiring real-time depth, texture, and lighting data). Early workflow specs also add direct HLS output, PCIe 5.0 and AMD ROCm affinity, and frame-level multi-threading that demands careful thread management to avoid freezing the pipeline.
Where AV2 fits alongside VVC, H.264, and the road to H.267
Choosing between AOMedia’s open-source codecs and MPEG’s standards comes down to a split between royalty-free licensing and patent-pool encoding speed.
AV1 vs AV2 vs VVC/H.266
Neither codec wins outright: VVC wins if you prioritize raw encoding and decoding speed, while AV2 wins if you need royalty-free licensing and broader open-source adoption. Approved as an international standard in 2020 through ITU-T and ISO/IEC MPEG cooperation, VVC encodes 50 to 100 times faster and decodes 8 to 10 times faster than AV2, and developers can already tap software decoders from Fraunhofer HHI and VVdec. However, VVC relies on traditional patent-pool licensing, which has noticeably restricted its hardware adoption compared to AOMedia’s royalty-free formats.
Coexistence and the H.267 roadmap
Just as codec coexistence keeps H.264 and even legacy MPEG-2 running across millions of devices alongside AV1 today, streaming servers hand each client the best stream its silicon can decode. With CDN rollouts expected post-2025, AV2 is an open-source bridge toward H.267 in 2028. Looking further out, speculative research into neural codecs, echoing AI video models like OpenAI’s Sora, points toward AI decoders reconstructing video from a few prompts,
Adoption reality: survey data and the chicken-and-egg trap
An AOMedia survey found that 53% of member organizations plan to adopt AV2 within 12 months of its final release and 88% expect implementation within two years, even as 88% simultaneously rate AV1 as critical or important. Those numbers expose the classic chicken-and-egg adoption deadlock. Content creators wait for widespread hardware decoding before encoding in AV2, while device makers wait for content demand before spending silicon area on decoders. Practical teams are shipping on AV1 today and keeping an upgrade path ready for AV2.
Which codec should you choose, and how to verify your devices
Switching from AV1 to AV2 isn’t worth it for most workflows right now unless you’re planning post-2027 8K, VR, or CDN infrastructure. Because video deployment is bound by the chicken-and-egg adoption cycle and physical chip manufacturing, content can’t reach mass audiences until decoders ship in consumer devices. Even though 53% of AOMedia members plan to adopt AV2 within 12 months of final release and 88% within two years, consumer hardware decoding won’t hit volume until the 2027 to 2028 silicon window.
Choose AV1 for
AV1 is the better codec for practical deployment today, while AV2 is the better architecture for high-resolution and multi-stream roadmaps shipping after 2027 to 2028 hardware arrives. Between the two royalty-free options, AV1 already has wide hardware and software support that can shrink a 50 GB Blu-ray master down to a 3 to 5 GB near-lossless file on machines you already own. Use AV1 right now for:
- Mainstream streaming platforms (following YouTube, Netflix, and Vimeo)
- Archiving and shrinking 50 GB Blu-ray and large HD masters to 3 to 5 GB files
- Mobile video streaming where hardware decode saves battery life
- Real-time video conferencing, gaming streams, and cloud video storage
Choose AV2 later for
Build AV2 into your multi-year engineering roadmap for post-2025 CDN backend upgrades and post-2027 consumer delivery where its specialized tools pay off:
- Live sports broadcasts that use native multi-angle and split-screen delivery
- Interactive, Twitch-style adaptive streaming with dynamic bitrate switching
- Mobile apps aiming to slash cellular data consumption once phones have AV2 silicon
- Enterprise conferencing, education platforms, virtual events, and 8K or VR/AR streams
Verify support per device, not per marketing
Skip AV2 for any project shipping before roughly 2027, when streaming to battery-powered devices that lack hardware decode (which triggers dropped frames and heavy battery drain), when targeting Apple and Safari users (where limited AV1 support already impacts up to 1.6 billion devices), or when legal teams won’t tolerate residual patent ambiguity. Here at GeekExtreme, we verify codec support against the actual silicon block inside a target device rather than reading the retail box.
In the wild, a laptop’s dedicated GPU will routinely decode a codec that its older integrated CPU can’t handle, and smart TVs often advertise codec compatibility that works only for local USB stick playback while failing inside streaming apps. If you want to skip manual codec wrangling altogether, managed video platforms like Movie One let you upload a single master file and automatically generate AV1, H.264, and eventual AV2 renditions, detect device hardware support at playback, and handle DRM and security.
Encoding AV2 today: tooling reality and workflow limits
If you’re attempting an AV2 test encode today, you’ll need to rely on FFmpeg and reference command-line software rather than consumer GUI tools, since HandBrake support for AV2 isn’t established in current documentation. Practical AV2 encoder choices right now are limited to AOMedia’s free reference encoders and early FFmpeg integrations, which support direct HLS output for streaming pipelines but demand 2 to 3 times the CPU compute of AV1.
If you’re building an experimental encoding rig to test those workflows, AV2 performs best on systems outfitted with PCIe 5.0 bandwidth and AMD ROCm acceleration. While the AV2 specification includes frame-level multi-threading to parallelize heavy encodes across many CPU cores, it requires careful, manual thread management. Sloppy thread allocation will create resource bottlenecks that freeze the entire encoding pipeline, which is why most early testers run one software encode, see the wall, and wait for dedicated hardware encoders.
AV1 vs AV2 questions readers ask
Four recurring questions come up whenever streaming engineers and home-lab builders compare AOMedia’s two codecs.
AV2’s current existence and playback status
AV2 exists as an AOMedia codec specification that began development in 2020 with finalization targeted for the end of 2025, not as a broadly available consumer playback format with shipping hardware decoders.
YouTube and Netflix support for AV1 and AV2
YouTube, Netflix, and Vimeo stream AV1 in production today, making AV1 the right target for current platform delivery while CDN rollouts for AV2 sit on post-2025 roadmaps. Industry projections suggesting Netflix could cut its bandwidth bills by 30% to 50% (saving billions of dollars) with AV2 remain speculative.
Codecs vs containers like MP4 and WebM
AV1 and AV2 are video stream compression codecs, not file containers like MP4 or WebM, static image formats like JPG or JPEG XS, or audio codecs such as Ogg Opus and Ogg Vorbis. LCEVC (Low Complexity Enhancement Video Coding) works differently: it’s an enhancement layer that stacks on top of base codecs like AV1 or AV2, acting like a turbocharger on an already fast engine.
HEVC vs AV1 file sizes
AV1 compresses video files, including 10-bit video, roughly 30% smaller than HEVC and about 50% smaller than VP9 and H.264 at equivalent visual quality, letting you compress a 50 GB Blu-ray down to 3 to 5 GB near-lossless today without waiting for AV2.
Frequently Asked Questions
AV2 vs AV1 which codec is better
AV2 is the better codec on paper, cutting bitrates by roughly 30% over AV1 at the same visual quality while adding native split-screen delivery, HDR10+ dynamic metadata, and dedicated AR/VR optimizations. In practice, AV1 is the better choice for anything shipping right now because it already has widespread hardware acceleration across modern GPUs, CPUs, and phones. Unless you are building a long-term roadmap for 2027 to 2028 hardware, stick with AV1.
does AV2 hardware support exist yet on GPUs and phones
No consumer GPU or smartphone ships with dedicated AV2 hardware decoding right now, and existing cards like the GeForce RTX 3080 or RTX 5070 Ti cannot get it through a driver update. Hardware acceleration requires physical transistors baked directly into the silicon. Because chipmakers need 18 to 24 months for silicon design plus another 6 to 12 months for device integration, mass-market AV2 hardware won’t land until the 2027 to 2028 window.
