One perception of the Alliance for Open Media is that of an aggressive body comprising technology and internet behemoths intent on disruptive monopolization of the entertainment industry; another is that advanced codecs should be a commodity. What we don’t expect to hear is how the royalty free AV1 patent licensing group is on the ultra-defensive, having set up a World Council and Legal Defense Fund – preparing for all manner of sticky situations which may arise from the plethora of patented intellectual property within the alliance.
According to AOMedia’s Executive Director Gabe Frost, of Microsoft, the idea is for every member of the alliance to sign a patent agreement before joining, applying the same policy as the W3C (World Wide Web Consortium). Crucially, a company does not have to be an AOMedia member to license the technology, meaning if you implement AV1, you then can’t sue others for using it, according to Frost, speaking to Faultline Online Reporter.
“We are creating an environment where commercial deployments can exist and thrive, not trying to build a codec to end them all,” said Frost.
AOMedia, driven by Google, Netflix, Microsoft, Intel, Amazon, Cisco and more, has drafted in legal and engineering experts to assess the AV1 processing conducted independently by member companies, an arduous and rigorous process, assured Frost.
Tightening up defenses for potential patent litigation cases is the ugly but essential side of the business. The interesting aspect is where AOMedia sees future disruption through AV1 support occurring. Following the official launch of AV1 a few weeks ago, the alliance is now waiting patiently for the critical silicon stage – as the fledgling codec embarks on its first phase of product adoption.
The initial phase involves bitstream specification to unblock silicon implementations, using an unoptimized version of the reference software codec, from which reference streams for production validation are then made available and binding specifications for popular media formats and encryption. Into Phase 2 we will see the first commercial AV1 codec projects emerging, through adoption by content creation tools and desktop browsers, using a better optimized reference codec based on open source tools and developer frameworks.
Phase 3 is when next-generation silicon in AV1-enabled PC, mobile, TV and set tops will emerge, with commercial video tools for greater optimization and efficiency for scale. This sets the stage for Phase 4, in which AV1 becomes available on all modern silicon, with devices supporting AV1 playback, camera capture and real-time communications.
AOMedia expects to reach Phase 2 later this year, initially in desktop browsers before games consoles and other streaming devices from the end of 2018 into early 2019, planning to reach Phase 4 by 2020. Frost could not divulge whether “other devices” includes smartphones, preferring not to directly address product roadmaps due to the compute versus battery trade off debate around mobile devices. However, Frost said he has seen Netflix content encoded with AV1 on his phone, so we can rest assured it can already be done.
In the meantime, AOMedia is busy addressing swarms of queries coming in from companies preparing to support AV1. French firm Ateme is perhaps the only AOMedia member which we would refer to as a traditional encoding company, but Frost assured us that discussions with others are underway, as many vendors await widespread device support before investing in AV1. “Companies like Beamr are great, but with someone like V-Nova there is some conflict,” said Frost, referring to the second-generation Perseus codec due out in a couple of years. V-Nova has said in the past it doesn’t see AV1 as a competitor, claiming it can work alongside Perseus as a sort of “turbocharger”.
Frost was also dismayed at how AOMedia has been blamed for its repeated delays, insisting the member companies discussed a goal to arrive at and this was reached, apparently conveying the wrong message about AV1 rolling out in early 2017. “There has never been a codec done in under 3 years. Of course we were hoping to launch earlier, but there were no technical setbacks,” said Frost.
AOMedia is caught up in a three-way flux with two HEVC patent licensing pools HEVC Advance and MPEG-LA. MPEG’s chair and co-founder Leonardo Chiariglione recently, and very publicly, slammed AOMedia, essentially describing how the success of AV1 would, in short, destroy the industry – shifting more power into the hands of a few internet and chip giants away from traditional device makers.
“We admire Leonardo so much and everything he has done for the industry. But the governance and rules around his own organization has handcuffed the ability to do stuff in the way AOMedia does. There are some narrow views on innovation, in that monetization typically comes via license fees, but this is just one bullet point. Added value comes from apps and services – pushed up the monetization stack. This is what has hit HEVC,” said Frost. He gave the example of how people criticized Microsoft for the Xbox games console, calling it a loss leader, yet the company has made a lot of money from additional software and services in the gaming market.
HEVC Advance has responded by recently eliminating some royalties, specifically for subscription and title-by-title content distribution. “We would encourage HEVC Advance to go completely royalty free, but we would be surprised,” said Frost. Although AV1 is royalty free, users will still need to pay implementation costs and fees to look after it all.
“The entire world is using AVC, now AV1 is making 65% savings versus AVC. Compression has the most impact in delivering video beyond built-up areas,” said Frost.
AV1 also claims 30% savings against current generation codecs including HEVC and VP9, according to a test by Bitmovin. A recent study by Moscow State University showed that AV1 was between 2,500 and 3,000 times slower than its competitors. However, Frost confirmed our suspicions, saying the test used a 6-month old version of AV1 with no optimization tools, and the research team never reached out to AOMedia. “The code we produce is experimental, meaning the codec comes with different options, so you need to write the code through the lens of different bit streams and apply optimization afterwards. Moscow State University is doing some good work,” highlighted Frost.
A final burning issue to address is that of Samsung’s AOMedia membership, which seemed all but confirmed in early February, but rumors have since tailed off. Frost confirmed Samsung hasn’t backed off and a membership is on the cards, but there a lot of questions to be addressed. “Some 82% of all internet traffic will be video by 2021, mostly from AOMedia member companies, but devices don’t sell content – whether on the big screen or small. Membership is not a prerequisite for content. It’s the end users who are important,” said Frost.