More Details Emerge About AV2
We often write about codecs on the 8K Association website as they are a key part of the 8K ecosystem. The future of digital media is via streaming as packaged media seems unlikely to make a comeback. Codecs are the piece of technology that make it feasible to send high resolution, high frame rate and high dynamic range content with wide color gamut over networks.
8K Produces a Lot of Data
As Spin Digital pointed out when we interviewed them at IBC, an uncompressed 8K 10 bit 60P frame from a camera, with no compression is producing 55.62 Gbps or 6.95 GB/s of raw data. On the other hand, to stream that content (as Intel did for the Paris Olympics), you need to get under 50 Mbps – that’s an order of magnitude compression of more than 1,000:1. It’s not easy.
Various strategies such as color subsampling, to reflect that the eye has less resolution in color than in luminance, and mezzanine codecs such as as JPEG XS can be used in production to maintain very high fidelity to the original but at lower bit rates. JPEG XS and color sub-sampling might get you down to around 8 Gbps, which may be OK in a 10G Ethernet network system, but not in streaming. If we were aiming for 30Mbps average (which Intel demonstrated), that would still mean a compression ratio of 266:1 being needed.
While production codecs like JPEG XS are “visually lossless”, to get to these extreme levels of compression, some element of data loss is inevitable, but the aim of different codecs is to minimize the loss of visual quality at a particular compression level.
Codecs – a Very Brief History
The MPEG group was set up by ISO and IEC to develop codecs in the late ‘80s. The group produced the first standardized and widely adopted codec – MPEG-1 which was for VHS-grade video compression and had a compression ratio for video of around 26:1. (It also included the massively widespread MP3 audio codec). It was released in 1991.
MPEG-2 came out in the second half of the ‘90s and was very widely adopted for digital TV in standard definition as well as being the main codec used for DVDs. It is still used in basic STBs and by broadcasters.
With the move to higher resolution, new codecs were needed and MPEG-4 (H.264 or AVC) was the next development and was released in 2004. It was used for TV, Blu-ray and was widely adopted for streaming. H.264 could support 8K.
Next to come along was H.265 – also known as High Efficiency Video Coding (HEVC) and that arrived in 2013. HEVC has been adopted for HD STBs and is widely used in devices from smartphones to computers.
The most recent standard in this series is H.266 also known as Versatile Video Coding (VVC) which further improved quality and supports up to 16K resolutions as well as HDR and wide color gamuts. The Intel/Spin Digital demonstrations at the Paris Olympics used VVC encoding achieving compression ratios of over 1,000:1. The standard was finalized in 2020.
Trade Offs for Codec Choice
In general, each of these generations of codec is expected to boost the coding efficiency by up to around 50% at the same visual quality. This comes at the cost of increased complexity, especially in the encoder.
There are also a lot of issues related to the licensing of these codecs. That’s the topic of many careers, and we can’t cover it here, but while the patents on which MPEG-1 was built have all expired, the costs of using the different codecs is a significant issue. The HEVC format uses many thousands of patented techniques. Samsung alone contributed over 4,000.
Why Pay?
Because of the substantial costs to pay licences for the more recent MPEG technologies, the Alliance for Open Media (AOMedia) was formed and it developed a codec called AV1. It was a development of the VP9 codec which had been developed by Google, but had the support of more companies. Amazon. Cisco, Intel, Microsoft, Mozilla and Netflix were founders along with Google. There are now 48 members with Samsung, Meta, Nvidia and Tencent now on the Steering Committee.
The key point of AV1, which was released in 2018 and competes really with HEVC in broad terms, is that it is an open royalty-free format. It is designed for streaming and web-based operations. Reassurance that AV1 is a good technical solution came from the award of an ‘Engineering Emmy’ award from the National Academy of Television Arts & Sciences (NATAS) at the end of September 2025.
Difficult to Migrate
Choosing a codec is a big decision for a TV operator. streaming service or operating system developer. They are complex technologies and once adopted, you need to have the systems to encode and also to decode the technology. Typically, earlier adopters of a codec may use software decoders (or encoders) to support the technology while a base of hardware-enabled devices is established. However, developing the devices and low cost hardware can take years and more years to be established in the market. Typically, hardware decoders can be much more power and cost efficient than software.
(If you read this article on the day of publication (October 21st 2025), you will still have time to register for a webinar on AOMedia)
We reported an article published in 2024 on the issue of the time taken to adopt new codecs by specialist, Jan Ozer. A quick summary? New hardware takes 3-5 years to be adopted with enough support for a market to become addressable.
Why Are We Talking About This Today?
We’re writing about this topic today, because AOMedia has said that it will release its AV2 codec, designed to compete with VVC, at the end of 2025. We reported on this in September around the time of IBC, but the group has now given more details. The organization is claiming an improvement in coding efficiency of around 30%. That would allow streamers to use fewer bits to deliver content with the same quality. Alternatively, AOMedia says that the there are 28.6% improvement in PSNR and 32.6% improvement in VMAF measurements (visual quality metrics). That could mean that streamers could deliver higher quality content with existing bitrates.
AV2, like VVC, adds other features such as support for multi-layer video and immersive formats, although details of that are yet to be clear. There is a 25 minute video on YouTube from Andrey Norkin of Netflix and another with the test results by Ryan Lei of Meta. These mainly covered resolutions up to 4K and Lei’s presentation also included some early subjective tests. There’s also a third video on AV1 Film Grain Synthesis, but we’ll return to that another day!
