A Video Codec Licensing Update
There have been several interesting IP-related developments in the codec market in the last few months that you should know about, including third-party HEVC pricing evaluations and the Media Coding Industry Forum’s (MC-IF ) attempt to influence licensing for the upcoming Versatile Video Codec (VVC). Let’s jump right in.
MPEG LA Bares its Teeth
On November 16, 2018, MPEG LA announced that the District Court in Düsseldorf, Germany ruled that cell phones distributed by Huawei’s and ZTE’s German subsidiaries infringed H.264 patents held by two patent holders in MPEG LA’s AVC Patent Portfolio License. As reported by IP watchdog site Patently Apple, the judgment was particularly onerous; not only did the court order the two companies to stop selling the offending mobile phones in Germany, they ordered that “all such products in their possession or the possession of third parties must be recalled and destroyed.”
The MPEG LA release also yielded two other interesting nuggets. First was that MPEG LA expects verdicts in additional cases brought other H.264 IP owners against Huawei and ZTE before the same German court in December, though I haven’t seen any notice of these. For perspective, MPEG LA packages patent pools that offer a single license for patents owned by members of the group. MPEG LA can’t sue companies that infringe patents held by group members; that’s up to the individual members. Two members of the MPEG LA group sued and won in Germany, with others also suing. You would expect these same members to sue Huawei and ZTE in other countries that enforce IP rights and to sue other infringing phone vendors in Germany and other countries.
Huawei and ZTE can avoid these suits by licensing the technology from MPEG LA, and perhaps they will. Their only alternatives are to sell phones without H.264 playback, which would be uncompetitive, or to exit these markets.
The other major data point from the MPEG LA release was that the court found that that the “AVC License is fair, reasonable and nondiscriminatory.” This language relates to the FRAND (fair, reasonable, and non-discriminatory) terms to which many IP contributors to a standard like H.264 and HEVC agree to ensure commercially reasonable royalty policies.
For the record, the MPEG LA license charges $0.20/unit with an annual cap of $9.75 million, which the German court found to be fair and reasonable. This is not the first time a court has found the MPEG LA license fair and reasonable. In Microsoft vs. Motorola, the court used the MPEG LA $0.20 royalty as the baseline to compute what was fair and reasonable for the Motorola IP, implicitly concluding that $0.20/unit was FRAND for H.264, albeit at the low end of the pricing spectrum.
Why is this relevant? Because Unified Patents just released a comprehensive objective economic evaluation report that finds that based upon this $0.20 value for H.264, a fair and reasonable charge for an HEVC license would most likely fall between $0.08 and $0.28 per unit, depending on the use case and device type. This compares to a minimum $2.07/unit estimated to be charged by the three HEVC patent pools (MPEG LA, HEVC Advance, and Velos) with many other potential licensors not yet in a patent pool (See Table 1). The report was prepared by Dr. Mario A. Lopez from economic consulting firm Edgeworth Economics and Martin Bader of law firm Sheppard Mullin, and will be made available to customers of Unified’s Video Codec Zone.
Table 1. Known and unknown royalties for HEVC.
The rationale behind the valuation is simple and logical. If you assume that HEVC is 50% more efficient than H.264, this reduces storage and bandwidth costs by 50%. However, since H.264 was standardized, costs for storage and bandwidth have dropped by a far greater percentage. If HEVC became available at the same time as H.264, perhaps you could argue that it was worth 100% more than H.264, but at today’s prices for storage and bandwidth, the value is much lower.
According to Unified, the report authors have provided similar valuations in many patent matters over the last ten years. Though the report has no independent legal value, it’s a preview of an approach that an independent economic expert witness might take in a patent infringement suit. You would expect the HEVC patent pools to offer a similar analysis, as HEVC Advance does here.
The HEVC Advance report points to the highest FRAND rate considered in the Microsoft case, $1.50 in 2003 dollars, doubles it to account for the bandwidth savings produced by HEVC’s greater efficiency, and adds $1.10 for inflation. This doesn’t take into account Unified Patent’s argument that the value of these bandwidth savings has dropped precipitously. As an example, according to this historical report I found online, “the average cost per gigabyte of streaming video transferred in 2003 was $2.15.” Today, AWS CloudFront pricing starts at $0.085/GB, and drops as low as $0.02 per GB, a reduction of 99%. In 2003, HEVC would have saved you $1.075/GB compared to H.264; today it’s $0.01/GB.
Will a court find this argument persuasive? Who knows, and to be fair, the HEVC Advance report offers several other pricing justifications. But Unified Patents is emerging as a tech savvy organization funded by many H.264 and HEVC users that can quickly and affordably act to help any HEVC user sued by an HEVC IP owner. Unified is also actively challenging patents to help avoid overcharging and other abusive practices.
MC-IF Making VVC FRAND-ly
Another organization is attempting to proactively steer royalty policies for the next generation MPEG codec towards a more reasonable footing. On Monday I chaired a panel for theMedia Coding Industry Forum (MC-IF). The title of the event was the Future of Video Codec Licensing: Avoiding the tragedy of the commons. By way of background, the MC-IF is a group formed to “work on the non-technical aspects of deployment of media standards, notably including licensing.” Members listed on the forum’s website include IP owners, codec and encoder vendors, several consumer electronics vendors, one HEVC patent pool, and a range of other companies. Looking around the audience at the event, I saw all these categories represented as well as several companies with large streaming content properties.
The “tragedy of the commons,” refers to the concept “that individual actions in a group, while individually rational or even optimal, might result in an outcome that is far from optimal for any member, and generally undesirable for the group.” In plain language, MC-IF was formed to help ensure that the next MPEG codec, the Versatile Video Codec (VVC), launches with a cohesive, transparent, and affordable royalty structure that promotes its acceptance. The obvious contrast is HEVC, that came with a fractured, opaque, and overreaching structure that even the head of MPEG has called broken.
My panel was titled Industry Needs and Opportunities for the Ecosystem. The entire meeting was held under the Chatham House Rules, which means that I can’t identify the speakers or their companies without their approval. In general, the three speakers were service or technology providers and discussed their experience with attempting to deploy HEVC internally or for their customers. One declined to be quoted in this article even without attribution.
Another speaker, a VP for a supplier to broadcasters, noted that several of the company’s customers were experimenting with AV1 in favor of HEVC. When I asked about the perceived technology risk relating to AV1, the response was “many of our customers perceive the technology risk to be less for AV1 than HEVC, primarily because AV1 is seen as a known technology with known risks. In contrast, not only are there unanswered questions about licensing from the HEVC pools, there are many, many HEVC IP owners who haven’t joined a pool or announced whether they will seek royalties or not. Perception matters to these customers and the perception is that there are too many unknowns about HEVC.”
Divideon’s Middle Ground
The third speaker, Jonatan Samuelsson from Divideon, who agreed to be identified, presented the compromise shown in Figure 1. On the left, Samuelsson posited that royalty-free technologies came with a high IP risk, though this contradicts the real-world experience expressed by my other speaker.
More importantly, he also asserted that royalty-free codecs like AV1 often had performance compromises because they couldn’t use effective tools and techniques that are patented by others. In essence, open-source codecs either have to use IP owned by one of their contributors or open-source techniques, or have to reinvent the wheel to avoid stepping on IP owned by non-members. While certainly few in the Alliance for Open Media (AOM) would agree that this compromises performance, none of the early deployments by Netflix or YouTube have pushed AV1’s quality envelope, though my tests showed AV1 to be about 18% more efficient than HEVC.
Figure 1. The pros and cons of royalty-free and unregulated FRAND codecs, and an alternative balanced approach.
Interestingly, during lunch, a representative from a content company privately shared that even if AV1 isn’t dramatically superior in encoding performance to HEVC or VVC, it could still provide significant value as a single codec that can be played anywhere. Though HEVC is available widely in Smart TVs and iOS devices, it’s only available on 78% of iOS devices and 57% of Android devices, and won’t play in Chrome or Firefox on any platform even if the underlying operating system does support HEVC.
In contrast, AV1 already plays efficiently in Firefox and Chrome and is being shipped by a prominent social media vendorto iOS and Android devices for playback in their app. If AV1 is widely deployed on Smart TVs, OTT boxes, and Apple devices, it could ultimately replace H.264 as the single codec that plays everywhere. Given that VP9 has achieved good penetration in these markets and that Apple, Netflix, YouTube, Amazon, and Hulu are all AOM members who plan to distribute AV1-encoded content, you would expect AV1 playback to be very widely adopted.
One suspects that southern regions occupied by reddish-toned individuals with horns might experience sustained sub-zero temperatures before Google enables accessible HEVC or VVC playback in Android or Chrome, effectively preventing any single MPEG codec from ever playing everywhere. On the other hand, another lunch companion wondered if Google’s refusal to implement HEVC playback in Chrome, even when available in the operating system, was a potential antitrust violation. While this obviously hearkens back to Real Network’s successful Antitrust suit against Microsoft, the speaker isn’t an attorney and I haven’t independently vetted this claim. I’m merely sharing this as an example of the typical lunch conversations held during the meeting.
Back on Samuelsson’s analysis, on the right the Figure shows problems relating to HEVC’s Unregulated FRAND policy, which basically amount to uncertainty and higher costs. Indeed, one technology vendor from the audience expressed a preference to work with standard-based technologies but reported that for his organization and many customers, HEVC’s cost structure and uncertainty was not only a bar to adoption, but actually encouraged experimentation with open-source technologies like AV1.
Samuelsson then proposed a middle ground where individual encoding techniques could be included in a codec with bitstream tags that allow users to include or exclude technology from different providers to meet difference performance and cost goals. This is the approach taken by Samuelsson’s company Divideon with its xvc codec.
Interestingly, MPEG co-founder and chairman Leonardo Chiariglione advocated a similar approach in a 2018 blog post entitled,A crisis, the causes and a solution. In the post, Chiariglione proposed developing “coding tools with ‘clear ownership’, unlike today’s tools which are often the result of contributions with possibly very different weights... [And to]...embed in the standard the capability to switch coding tools on and off. The work of patent pools would be greatly simplified because they could define profiles with technologies that are “available” because they would know who owns which tools. Users could switch on tools once they become usable, e.g. because the relevant owner has joined a patent pool.” Samuelsson’s proposal was the topic of much positive discussion, though several attendees indicated that the approach was easier to implement in software than in hardware.
Will MC-IF help VVC IP owners avoid the mistakes of the past? It’s impossible to say at this point, but MC-IF certainly has their work cut out for them. Though teenagers typically morph into adults at some point and abandon their petulant and self-centered ways, you really can’t say the same for at least one of the three HEVC patent pools and the many HEVC IP owners who haven’t joined a pool or announced their licensing-related intentions. Specifically, unlike MPEG LA and HEVC Advance, Velos doesn’t supply a patent list or royalty rates on their website and won’t share them without a nondisclosure agreement. Nearly six years after HEVC was finalized, Velos still won’t clarify whether they will seek a royalty on HEVC-encoded content, despite this being a very significant bar to publisher adoption.
Scanning the licensee lists from MPEG LA and HEVC Advance (Velos doesn’t have one), reveals many significant HEVC users or deployers missing from one or both lists, including Amazon, Adobe, Google, Intel, Microsoft, Netflix, NVIDIA, all on neither list, and Apple and Cisco (both on MPEG LA, not HEVC Advance). Not surprisingly, all of these companies are AOM members. While these companies may have signed agreements or made other arrangements, from the data we can see, it appears that we’re many years and many expensive lawsuits from figuring out which IP is integral to HEVC and what a FRAND cost will be. To expect many of these same IP owners to come together and avoid the tragedy of the commons seems unlikely, though the MC-IF is undoubtedly a step in the right direction.
As recent lawsuits regarding DASH-related patents make clear, royalty payments on adaptive bitrate streaming are inevitable—even if it's not yet clear who'll be paying who, or how much
BitSave promises 70% bitrate savings via machine learning, and currently works with H.264, with H.264, H.265, and VP9 on the way
Advances in video codec technology driven by the explosion in video over IP are creating heated competition and a fragmented market. Will one winner emerge or is this the new normal?
Ahead of Streaming Media West, a meeting of codec experts offers new developments in leading-edge codecs, as well as field reports from companies already using them.
A new research report highlights how H.264/H.265 workflows have become integrated, as well as discussing VP9 and AV1 as possible AVC replacements
HEVC Advance is eliminating subscription and title-by-title royalty fees for non-physical content distribution, and it's adding reduced royalty rates for low-end consumer devices.
Thanks to a fractured HEVC licensing system companies no longer have the financial incentive to innovate, but Leonardo Chiariglione suggests steps to reverse the damage.
New Unisphere report indicates that OTT services and other video publishers are making the move to HEVC, but are equally interested in AV1