26 Jun 2017

[Blog] AV1

Hello, this is Erik with a post about the up and coming AV1 video codec. While none of the elements that will eventually be accepted into the final design are currently fixed (and the specification of the codec might not even be finalized this year), more and more companies have started working on support already. This post is meant to give a more in-depth view of what is currently happening, and how everything will most likely eventually fall into place.


Internet Video Codec (NetVC) is a working group within the Internet Engineering Task Force (IETF). The goal of the group is to develop and monitor the requirements for the standardization of a new royalty-free video codec.

The current version of their draft specifies a variety of requirements of the next generation royalty-free codec. These range from low-latency real-time collaboration to 4k IPTV, and from adaptive bit rate to the existence of a real-time software encoder.

While these requirements are not bound to a specific codec, it determines what would be needed for a codec to be deemed acceptable. Techniques from the major contenders have been taken into account whilst setting up this list in 2015, with the Alliance for Open Media (AOMedia) forming shortly after to develop the AV1 codec to comply to these requirements in a joint effort between the Daala, Thor and VP10 teams.


The AOMedia has been formed to streamline development of a new open codec, as multiple groups were working simultaneously on different new codecs. AV1 is the first codec to be released by the alliance. With many of the influential groups within the internet video area participating in its development, it will be set-up to compete with HEVC in regard to compression and visual quality while remaining completely royalty free.

Because steering completely clear of the use of patents when designing a codec is a daunting task, the AOMedia group has decided to provide a license to use their IP in any project using AV1, as long as the user does not sue for patent infringement. This, in combination with a thorough Intellectual Property Rights review, means that AV1 will be free of royalties. This should give it a big edge over the main competitor HEVC, for which there are multiple patent pools one needs to pay license fees to, with an unknown amount of patents not contained in these pools.

Decided is to take the development of the VP10 codec as developed by Google as the base for AV1. In addition to this, AV1 will contain techniques from both Daala (by Xiph and Mozilla) and Thor (by Cisco).

The AV1 codec is developed to be used in combination with the opus audio codec, and wrapped in WebM for HTML5 media and WebRTC for real time uses. As most browsers already support both the WebM format and opus codec, this immediately generates a large potential user base.


Development of AV1 is based mainly around the efforts of Google's VP10. Built upon VP9, VP10 was geared towards better compression while optimizing visual qualities in the encoded bitstream. With the formation of AOMedia, Google decided to drop VP10 development, and instead focus on solely the development of AV1.

Building upon a VP9 core, contributors can submit so called experiments to the repository, which can then be enabled and tested by the community. Based on whether the experiment is considered worthwhile, it enters IPR review, and after a successful pass there, it will be enabled by default and added to the specification. Most experiments have come from the respective developments from VP10, Daala and Thor, but other ideas are welcomed as well. As of yet, no experiments have been finalized.


Multiple tests have been run to compare AV1 to both H.264 and HEVC, with varying results over time. However, with no final selection of experiments, these performance measures have been made not only with different settings, but with completely different experiments enabled.

While it is good to see how the codec is developing over time, a real comparison between AV1 and any other codec can not be reliably made until the bitstream is frozen.

What's next?

With the end of the year targeted for finalizing the bitstream, the amount of people interested in the codec will probably only grow. With the variety of companies that AOMedia consists of, the general assumption is that the adoption of the codec and hardware support for encoding/decoding will be made available in a matter of months after the bitstream is finalized, rather than years.

While VP10 has been completely replaced by the AV1 codec, the same does not seem to hold for Thor and Daala. Both projects still see respective development, which does not seem limited to the features that will eventually be incorporated into AV1.

And that concludes it for now. Our next post will be by Carina, who will show how to create an access control system in PHP with our triggers.