{"id":829,"date":"2026-03-29T10:18:54","date_gmt":"2026-03-29T03:18:54","guid":{"rendered":"https:\/\/liveapi.com\/blog\/av1-codec\/"},"modified":"2026-04-17T10:21:43","modified_gmt":"2026-04-17T03:21:43","slug":"av1-codec","status":"publish","type":"post","link":"https:\/\/liveapi.com\/blog\/av1-codec\/","title":{"rendered":"AV1 Codec: What It Is, How It Works, and When to Use It"},"content":{"rendered":"Reading Time: <\/span> 12<\/span> minutes<\/span><\/span>

The AV1 codec delivers up to 50% better compression than H.264 at the same visual quality \u2014 without any licensing fees. For developers building video streaming applications, that means lower bandwidth costs, better image quality at the same bitrate, and no patent royalties to worry about.<\/p>\n

AV1 is now the codec of choice at YouTube, Netflix, and most major streaming platforms. YouTube encodes more than 75% of its video library in AV1. Netflix reported in December 2025 that 30% of their streams use AV1. The codec has crossed from “emerging technology” into mainstream production deployment.<\/p>\n

This guide covers what the AV1 codec is, how its compression actually works under the hood, how it stacks up against H.264, H.265, and VP9, what hardware and browser support looks like today, and how to decide when AV1 makes sense for your project. If you’re newer to codecs in general, our video codec overview<\/a> is a good starting point.<\/p>\n

What Is the AV1 Codec?<\/h2>\n

The AV1 codec is an open-source, royalty-free video coding format designed for efficient internet video transmission. AV1 stands for AOMedia Video 1 and was developed by the Alliance for Open Media<\/a> \u2014 a consortium that includes Google, Netflix, Amazon, Microsoft, Meta, Apple, Intel, and ARM.<\/p>\n

AV1 was released in 2018 as the successor to VP9. It was specifically designed to address two problems with existing codecs:<\/p>\n

    \n
  1. Licensing cost:<\/strong> H.264 and H.265 require royalty payments managed by patent pools. This adds real cost at scale and creates legal complexity for open-source developers and browser vendors.<\/li>\n
  2. Compression efficiency:<\/strong> The older royalty-free alternative, VP9, lagged behind H.265 in compression. AV1 was built to match or exceed H.265 quality while remaining completely free.<\/li>\n<\/ol>\n

    The result is a codec that compresses video more efficiently than any of its widely-adopted predecessors, while being freely implementable by anyone.<\/p>\n

    AV1 at a glance:<\/strong><\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n
    Property<\/th>\nDetails<\/th>\n<\/tr>\n<\/thead>\n
    Full name<\/td>\nAOMedia Video 1<\/td>\n<\/tr>\n
    Developed by<\/td>\nAlliance for Open Media<\/td>\n<\/tr>\n
    Released<\/td>\n2018<\/td>\n<\/tr>\n
    License<\/td>\nRoyalty-free, open-source<\/td>\n<\/tr>\n
    Compression vs. H.264<\/td>\n~50% better<\/td>\n<\/tr>\n
    Compression vs. H.265<\/td>\n~30% better<\/td>\n<\/tr>\n
    Compression vs. VP9<\/td>\n~20\u201330% better<\/td>\n<\/tr>\n
    Container formats<\/td>\nMP4, WebM, MKV<\/td>\n<\/tr>\n
    Primary use<\/td>\nInternet video streaming<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    AV1 uses the same foundational approach as H.264 and H.265 \u2014 block-based transform coding \u2014 but introduces advances in intra prediction, inter prediction, transform coding, and in-loop filtering that produce significantly better results at lower bitrates.<\/p>\n

    AV1 vs. H.264 vs. H.265 vs. VP9<\/h2>\n

    Understanding where AV1 fits in the codec landscape helps you decide when to use it. Here’s how the four major codecs compare:<\/p>\n\n\n\n\n\n\n\n\n
    Codec<\/th>\nReleased<\/th>\nCompression vs. H.264<\/th>\nLicense<\/th>\nBrowser Support<\/th>\nBest For<\/th>\n<\/tr>\n<\/thead>\n
    H.264 (AVC)<\/td>\n2003<\/td>\nBaseline<\/td>\nLicensed (royalties)<\/td>\nUniversal<\/td>\nMaximum device compatibility<\/td>\n<\/tr>\n
    VP9<\/td>\n2013<\/td>\n~30% better<\/td>\nRoyalty-free<\/td>\nAll major browsers<\/td>\nYouTube, Android, web<\/td>\n<\/tr>\n
    H.265 (HEVC)<\/td>\n2013<\/td>\n~50% better<\/td>\nLicensed (multiple patent pools)<\/td>\nLimited (Safari, Edge)<\/td>\n4K broadcast, HDR, high-end OTT<\/td>\n<\/tr>\n
    AV1<\/td>\n2018<\/td>\n~50% better<\/td>\nRoyalty-free<\/td>\nAll major browsers<\/td>\nStreaming, OTT, web VOD<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    For a detailed breakdown of AV1 and H.264 performance differences, see our AV1 vs. H.264 comparison<\/a>. For how H.265 measures up against H.264, see our H.265 vs. H.264 guide<\/a>.<\/p>\n

    The licensing difference matters in practice.<\/strong> H.265 has notoriously fragmented licensing \u2014 multiple patent pools exist, which has blocked browser adoption and slowed open-source tooling. Firefox and Chrome still don’t have broad H.265 playback support for exactly this reason. AV1’s royalty-free status means it can be freely built into browsers, operating systems, and open-source tools without legal friction.<\/p>\n

    VP9 vs. AV1:<\/strong> VP9 was Google’s royalty-free codec before AV1. AV1 delivers roughly 20\u201330% better compression than VP9 at the same quality. Most platforms that adopted VP9 \u2014 YouTube, Google Meet, Chrome \u2014 are now migrating to AV1.<\/p>\n

    The practical summary:<\/strong> If you’re choosing a codec for web-based VOD delivery today, AV1 is the best combination of compression efficiency and licensing freedom. H.264 remains the default for live streaming and maximum compatibility. H.265 is strong for broadcast and closed OTT ecosystems where hardware support is controlled.<\/p>\n

    How Does AV1 Compression Work?<\/h2>\n

    AV1 uses a block-based transform coding architecture \u2014 the same fundamental approach as H.264 and H.265. What sets AV1 apart is that it improves nearly every step of the encoding pipeline.<\/p>\n

    Here’s how AV1 encodes a video frame:<\/p>\n

    1. Block partitioning<\/strong><\/p>\n

    AV1 divides each frame into blocks. Unlike H.264’s fixed 16\u00d716 macroblock structure, AV1 supports flexible block sizes from 4\u00d74 up to 128\u00d7128 pixels. Larger blocks work well for smooth regions like sky or backgrounds; smaller blocks handle fine detail and sharp edges. This flexibility means fewer bits are needed to represent each region accurately.<\/p>\n

    2. Intra prediction<\/strong><\/p>\n

    For blocks that don’t reference other frames (within-frame prediction), AV1 predicts pixel values from neighboring blocks. AV1 supports 56 directional prediction modes, compared to H.264’s 9. More modes means better guesses, which means fewer residual bits to encode.<\/p>\n

    3. Inter prediction (motion compensation)<\/strong><\/p>\n

    For blocks that reference previous or future frames, AV1 uses motion compensation to predict content across time. AV1 introduces several advances over older codecs:
    \n– Compound prediction:<\/strong> Blending two reference frames instead of one
    \n– Warped motion:<\/strong> Handling global camera motion (pan, zoom, rotation) with affine transformations
    \n– Overlapped block motion compensation (OBMC):<\/strong> Smoothing block boundaries during motion estimation<\/p>\n

    These techniques remove temporal redundancy more effectively than H.264 or H.265.<\/p>\n

    4. Transform coding<\/strong><\/p>\n

    After prediction, the residual (the difference between the predicted block and the actual pixels) is transformed. H.264 uses only DCT (discrete cosine transform). AV1 supports multiple transform types \u2014 DCT, ADST (asymmetric DST), WHT (Walsh-Hadamard transform), and identity \u2014 and can choose the best transform type per block. The transformed coefficients are then quantized and entropy-coded.<\/p>\n

    5. In-loop filters<\/strong><\/p>\n

    AV1 applies three post-processing filters during decoding to reduce artifacts:
    \n– Deblocking filter:<\/strong> Reduces blocking at block boundaries
    \n– CDEF (Constrained Directional Enhancement Filter):<\/strong> Reduces ringing artifacts while preserving edges
    \n– Loop restoration filter:<\/strong> Applies Wiener filters and self-guided restoration to recover fine detail<\/p>\n

    These filters improve perceptual quality without storing extra data in the stream.<\/p>\n

    6. Internal bit depth<\/strong><\/p>\n

    AV1 processes video at 10-bit or 12-bit precision internally, even for 8-bit output. This reduces rounding errors during encoding operations and produces better output quality, particularly in areas with subtle gradients.<\/p>\n

    The compression gains from AV1 come from all of these improvements working together \u2014 not any single change to the pipeline. Each stage squeezes out a bit more redundancy, and the cumulative result is a 30\u201350% reduction in file size or bitrate at the same perceptual quality.<\/p>\n

    Advantages of the AV1 Codec<\/h2>\n

    Royalty-free licensing<\/h3>\n

    AV1 is free for anyone to use, implement, and distribute. No per-unit royalties, no patent pools, no licensing agreements. For open-source projects, browser vendors, and companies shipping at scale, this eliminates real legal and financial complexity. It’s why AV1 achieved universal browser support much faster than H.265 ever did.<\/p>\n

    Significantly better compression<\/h3>\n

    AV1 compresses video approximately 50% more efficiently than H.264 and 30% more efficiently than H.265. At the same perceptual quality, an AV1-encoded file needs roughly half the data of an H.264 file. That translates directly to lower streaming bitrates<\/a>, lower CDN bandwidth costs, and lower data consumption for your users.<\/p>\n

    For a platform delivering millions of video minutes per month, this compression advantage has real financial impact.<\/p>\n

    4K, 8K, and HDR support<\/h3>\n

    AV1 was designed for high-resolution delivery. It supports 4K and 8K video, HDR (High Dynamic Range), wide color gamut (WCG), and 10-bit\/12-bit color depth. YouTube has served 8K AV1 content since 2020. If you’re building a platform where visual quality at high resolutions matters, AV1 handles it well.<\/p>\n

    Universal browser support<\/h3>\n

    AV1 is supported across Chrome (since version 70, 2018), Firefox (since version 67, 2019), Edge (Chromium-based), Opera, and Safari (since version 16, 2022). As of 2025, AV1 playback is available in all major desktop browsers. Safari was the last major holdout \u2014 Apple added AV1 software decode in macOS Ventura and hardware decode in M-series Macs and iPhone 15+.<\/p>\n

    Growing hardware acceleration<\/h3>\n

    Early AV1 adoption was slowed by the lack of hardware decoders. Software AV1 decoding is CPU-intensive, which made high-resolution playback impractical on older or lower-power devices.<\/p>\n

    That has changed significantly. Intel Ice Lake (11th gen) and newer, AMD RDNA 2 (RX 6000 series) and newer, NVIDIA Ampere (RTX 30 series) and newer, Apple M1 and newer, and Qualcomm Snapdragon 888 and newer all include dedicated AV1 hardware decode support. Hardware encode support is also spreading, though it remains more limited than decode.<\/p>\n

    Strong industry momentum<\/h3>\n

    AV1 adoption has reached a point where it’s no longer experimental. YouTube encodes over 75% of its library in AV1. Netflix uses AV1 for 30% of streams. Discord, Meta, Twitch, and Google Meet all deploy AV1. The codec has proven itself in production at the largest scale in the industry.<\/p>\n

    Disadvantages and Limitations of AV1<\/h2>\n

    Slow software encoding<\/h3>\n

    AV1’s biggest limitation is encoding complexity. Software-based AV1 encoding using the libaom reference encoder can be 50\u2013100x slower than H.264 encoding at the same quality. Even faster software encoders like SVT-AV1 are significantly slower than H.264 or H.265 software encoding.<\/p>\n

    This is the main reason AV1 adoption for live streaming has lagged behind VOD. Real-time software AV1 encoding at broadcast quality is still impractical for most use cases on typical server hardware.<\/p>\n

    Limited hardware encoder coverage<\/h3>\n

    While hardware AV1 decode support is now widespread, hardware encode<\/strong> support is more limited. Not all devices that can play AV1 can encode it in hardware. Teams building real-time video conferencing or live broadcasting tools need to verify hardware encoder availability across their full target device range before committing to AV1.<\/p>\n

    Higher CPU load on devices without hardware decode<\/h3>\n

    Devices without dedicated AV1 hardware decoders fall back to software decoding, which requires significant CPU resources. On older smartphones, budget devices, or lower-power hardware, software AV1 decoding at 1080p or above can cause dropped frames, stuttering, or high battery drain.<\/p>\n

    Not yet practical for live streaming ingest<\/h3>\n

    Most live streaming infrastructure \u2014 hardware encoders, broadcast software like OBS, RTMP\/SRT ingest pipelines \u2014 defaults to H.264 or H.265. AV1 encoding in real time requires either specialized hardware (NVIDIA RTX 30+, Intel Arc) or accepting lower quality than what software encoding produces offline. For most live streaming applications today, H.264 remains the practical choice for ingest.<\/p>\n

    Uneven mobile support on older and budget devices<\/h3>\n

    Flagship phones from 2021+ typically include AV1 hardware decode, but budget and mid-range devices may not. If your app serves a global audience that includes users on older or lower-cost Android devices, you’ll want to test AV1 playback across representative hardware before deploying it as your primary delivery codec.<\/p>\n

    \n

    Managing codec compatibility, adaptive bitrate renditions, and CDN delivery across device types is the kind of infrastructure problem that a video hosting API<\/a> handles at the platform level \u2014 so your team can focus on building the product instead of the pipeline.<\/p>\n

    \n

    AV1 Hardware and Browser Support<\/h2>\n

    Browser support (2025)<\/h3>\n\n\n\n\n\n\n\n\n\n
    Browser<\/th>\nAV1 Support<\/th>\nSince<\/th>\n<\/tr>\n<\/thead>\n
    Chrome<\/td>\nYes<\/td>\nChrome 70 (2018)<\/td>\n<\/tr>\n
    Firefox<\/td>\nYes<\/td>\nFirefox 67 (2019)<\/td>\n<\/tr>\n
    Edge<\/td>\nYes<\/td>\nChromium Edge (2020)<\/td>\n<\/tr>\n
    Opera<\/td>\nYes<\/td>\n2019<\/td>\n<\/tr>\n
    Safari<\/td>\nYes<\/td>\nSafari 16, macOS Ventura (2022)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    Desktop GPU support<\/h3>\n\n\n\n\n\n\n\n\n\n\n\n
    GPU Family<\/th>\nHardware Decode<\/th>\nHardware Encode<\/th>\n<\/tr>\n<\/thead>\n
    NVIDIA RTX 20-series (Turing)<\/td>\nYes<\/td>\nNo<\/td>\n<\/tr>\n
    NVIDIA RTX 30-series (Ampere)+<\/td>\nYes<\/td>\nYes (NVENC)<\/td>\n<\/tr>\n
    AMD RDNA 2 (RX 6000-series)+<\/td>\nYes<\/td>\nNo<\/td>\n<\/tr>\n
    AMD RDNA 3 (RX 7000-series)+<\/td>\nYes<\/td>\nYes<\/td>\n<\/tr>\n
    Intel Tiger Lake (11th gen)+<\/td>\nYes<\/td>\nPartial<\/td>\n<\/tr>\n
    Intel Arc<\/td>\nYes<\/td>\nYes (QSV)<\/td>\n<\/tr>\n
    Apple M1\/M2\/M3+<\/td>\nYes<\/td>\nYes (VideoToolbox)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    Mobile chipset support<\/h3>\n\n\n\n\n\n\n\n\n\n\n
    Chipset<\/th>\nAV1 Decode<\/th>\nAV1 Encode<\/th>\n<\/tr>\n<\/thead>\n
    Qualcomm Snapdragon 888+<\/td>\nYes<\/td>\nNo<\/td>\n<\/tr>\n
    Qualcomm Snapdragon 8 Gen 2+<\/td>\nYes<\/td>\nYes<\/td>\n<\/tr>\n
    Apple A15 (iPhone 13+)<\/td>\nYes<\/td>\nNo<\/td>\n<\/tr>\n
    Apple A17 Pro (iPhone 15 Pro+)<\/td>\nYes<\/td>\nYes<\/td>\n<\/tr>\n
    MediaTek Dimensity 1000+<\/td>\nYes<\/td>\nNo<\/td>\n<\/tr>\n
    MediaTek Dimensity 9000+<\/td>\nYes<\/td>\nYes<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    A practical rule: for devices manufactured after 2022, assume AV1 hardware decode is likely available. For encoding support, verify against specific chipset documentation.<\/p>\n

    How to Encode with AV1<\/h2>\n

    There are three main open-source software AV1 encoders used in production, all available through FFmpeg. Understanding the differences helps you pick the right one for your workflow.<\/p>\n

    libaom (aomenc)<\/h3>\n

    The reference AV1 encoder, developed by Google and the Alliance for Open Media. It produces the highest-quality AV1 output at a given file size but is extremely slow \u2014 often 50\u2013100x slower than H.264. Useful for research, quality benchmarking, and offline archival encoding where time is not a constraint.<\/p>\n

    ffmpeg -i input.mp4 -c:v libaom-av1 -crf 30 -b:v 0 output.webm\r\n<\/code><\/pre>\n
    -crf 30<\/code> sets constant quality (lower = better quality, larger file). -b:v 0<\/code> disables bitrate targeting so CRF mode is active.<\/p>\n
    SVT-AV1<\/h3>\n
    Developed by Intel in collaboration with Netflix, SVT-AV1 is the fastest multi-threaded software AV1 encoder. It scales well across CPU cores and is now used in production by Netflix and other large platforms. Quality is slightly lower than libaom at the same file size, but encoding speed is dramatically higher \u2014 making it the practical choice for production VOD workflows.<\/p>\n
    ffmpeg -i input.mp4 -c:v libsvtav1 -crf 28 -preset 6 output.mp4\r\n<\/code><\/pre>\n
    -preset<\/code> controls the speed\/quality trade-off (0 = slowest\/best quality, 12 = fastest). Preset 6\u20138 is a reasonable starting point for production encoding.<\/p>\n
    rav1e<\/h3>\n
    A Rust-based AV1 encoder focused on correctness and safety. Less commonly used in production than SVT-AV1, but respected in the open-source community.<\/p>\n
    ffmpeg -i input.mp4 -c:v librav1e -qp 80 output.mp4\r\n<\/code><\/pre>\n
    Hardware encoders<\/h3>\n
    For real-time or near-real-time AV1 encoding, hardware encoders are the practical option:<\/p>\n