A new codec is not another layer of the onion
Why the technical independence of video codecs matters for patent licensing
By Giovanni Ballocca and Matteo Sabattini
It is tempting to describe the history of video coding as a simple evolution: MPEG-2 became AVC; AVC became HEVC; HEVC became VVC. In this telling, each major generation made it possible to transmit more or better video through finite networks and storage. Such a perspective is understandable. Standards bodies themselves refer to newer codecs as ‘successors’, and the performance of new codecs is routinely measured against that of their predecessors. Technology policy also borrows the language of generations and, almost without noticing, begins to imagine a cumulative stack.
But the word ‘generation’ can be misleading. A VVC codec is not an HEVC codec with another layer added on top. A decoder does not need to implement HEVC to decode a VVC bitstream and an AV1 decoder needs neither.
This technical reality has important implications for licensing: we should be cautious about treating distinct video codecs as though they were cumulative layers of a single technology stack. Each codec should be evaluated on its own technical contribution, legal framework and economic value.
A remarkable history of standardisation
Video coding standards are not simply documents that describe a technology. They are the result of an organised process in which companies, universities and research institutions submit competing technical solutions, those solutions are evaluated and a common specification emerges. When patented technology is selected, the relevant IPR policies create a route to access through (fair), reasonable and non-discriminatory (F/RAND) terms. This bargain gives implementers confidence that standard essential technology will be licensable, while allowing innovators to earn a return and reinvest in the next cycle of research. The patent system has been part of this success, not an accidental complication of it: contributors have invested heavily in compression research because successful inventions can be patented and licensed.
Patent pools became a practical complement to that bargain. Video standards can read on patents owned by many different organisations, so standardisation creates a need for efficient access to everything essential to implementing the chosen specification. The MPEG-2 pool was a landmark. In 1997, the US Department of Justice reviewed the proposed licensing arrangement and recognised the efficiencies of a one-stop licence covering complementary standard essential patents (SEPs). Importantly, the structure was built around technical essentiality: an independent expert determined whether patents were truly necessary to comply with MPEG-2.
The model proved influential. Pools and portfolio licensing programmes subsequently developed around AVC, HEVC, VVC and other technologies. Their central economic proposition is straightforward: instead of negotiating separately with many patent owners for patents that must all be practised to implement one standard, implementers can obtain substantial coverage through a single transaction, while patent owners can license efficiently across a broad market. But this does not mean that every codec belongs in the same basket.
‘Successor’ does not imply ‘cumulative’
The distinction becomes clearer when video coding is compared with technologies in which successive generations are visibly cumulative. Consider Wi-Fi: a Wi-Fi 7 product is ordinarily designed to communicate in an installed ecosystem that includes Wi-Fi 6 and earlier devices; and later releases build on a substantial body of shared architecture, procedures and specifications. The commercial product therefore resembles an onion – capabilities accumulate around a common interoperable system.
Video codecs are different. HEVC and VVC certainly share a research lineage. Both use the broad hybrid video-coding paradigm and VVC’s designers learned much from AVC and HEVC, just as new aircraft incorporate lessons learned from older aircraft. But the similarities end when it comes to implementation.
VVC defines its own normative bitstream syntax and decoding process. A VVC bitstream is decoded according to the VVC spec, while a HEVC bitstream is decoded according to the HEVC spec. A compliant VVC decoder can be implemented as a standalone decoder, an HEVC decoder cannot decode VVC and a VVC decoder cannot decode HEVC, simply because they use (at least partially) different coding tools. The same is true of AV1, which is a different codec under a different specification. Supporting several of these formats involves providing support for several codecs.
From a purely technological perspective, a manufacturer could build a television that decodes VVC but not HEVC. In practice, however, that would be a poor commercial choice, because enormous libraries of existing content are encoded in AVC, HEVC, AV1 and older formats. Legacy format support is valuable, but it is a market and compatibility decision, not a technical prerequisite for decoding VVC. Equally, interoperability with older content is a market requirement, not a technical dependency of the new decoder.
This distinction is evident in real products. Devices often contain multiple codec implementations because the content ecosystem is heterogeneous and manufacturers support whichever formats their customers’ content and services require. The presence of several decoders in the same television, smartphone or chipset does not turn those codecs into one technical standard, any more than the installation of several applications on the same computer turns them into one application. A multi-codec television contains several doors to several libraries, not one ever-thicker door.
AV1: different both in origin and in legal character
AV1 adds another distinction, this time legal and institutional as well as technical. It was developed by the Alliance for Open Media (AOM), a private industry consortium founded in 2015, rather than through the traditional International Telecommunication Union/International Organization for Standardization/International Electrotechnical Commission standardisation process. AOM published AV1 in 2018 and built its intellectual property framework around a royalty-free patent licence.
That description requires precision. Under the AOMedia Patent License, a licensor grants a no-charge licence to its “Necessary Claims” for qualifying AV1 implementations, subject to the terms of that licence. This is a meaningful commitment and an important part of the AV1 model. But no private consortium can license patents that it does not own or control. An innovator outside AOM that never accepted the AOM patent policy does not lose its patent rights merely because AV1 was marketed as or intended to be royalty free. If AV1 practises that third party’s patented invention, the AOM licence cannot supply the missing permission and the patent owner remains entitled to decide how to exercise its rights.
For that reason, ‘royalty free’ is not a technical property of the AV1 specification – it is, at most, a description of the commitments made by particular parties for particular rights – and a ‘royalty-free codec’ is not the same thing as a codec for which no patent royalties can ever be due. Sisvel’s AV1 licensing programme, for example, offers licences to patents held by innovators whose relevant rights are not made available simply by downloading or implementing the specification.
It would be more accurate, and less confusing, to describe AV1 as a proprietary consortium specification promoted under a royalty-free policy among participating parties. That formulation recognises both the engineering achievement and the limits of the promise. It also reminds implementers that source-code availability or a permissive software licence does not address the separate question of third-party patent rights.
Licensing as good citizenship, not inherited obligation
This distinction changes the starting point for licensing. F/RAND is a bargain. In the traditional standards context, a participant may commit to make SEPs available on F/RAND terms under the applicable standards development organisation (SDO) policy. That commitment facilitates the adoption of the common standard while preserving compensation for patented innovation.
An outside patent owner whose technology is used by a private consortium specification has made no such bargain. It has promised neither a F/RAND licence nor a royalty-free licence. It may not even know that its inventions are being used until it conducts expensive technical analysis and claim mapping from outside the consortium.
When such an owner nevertheless chooses to place its patents in a transparent, independently evaluated, one-stop licensing programme, it is enabling licensing voluntarily. One might call this being a good citizen of the innovation ecosystem. The owner could instead rely on ordinary patent rights and, without any further legal constraints, decide on whether and how to license and seek the remedies that patent law makes available. Offering a predictable licence converts this potential friction into access.
Voluntary discipline should not be mistaken for a pre-existing F/RAND duty. A well-governed pool for a proprietary codec may adopt familiar safeguards because they enhance trust and efficiency, but those safeguards are chosen to make the market work. They do not arise from an SDO bargain that the outside patent owner has never entered into.
Different technologies, different bargains
These points lead to a broader principle for codec licensing, because two questions should not be collapsed:
The first is legal: what commitments attach to the patents? Patents that are subject to an SDO F/RAND undertaking should be licensed consistently with that undertaking. Patents whose owners have made no such commitment should not be subject to F/RAND constraints by association simply because a licensing administrator places them next to F/RAND-encumbered patents.
The second question is technical and economic: what exactly is being licensed and what value does it provide? HEVC, VVC and AV1 are separate codecs. A product may implement all three, but each provides an independent capability to encode or decode a particular format, and their value propositions, adoption patterns, patent landscapes and legal origins can differ materially. If VVC decoding required HEVC decoding as an embedded layer, there might be a stronger argument that the two form one inseparable cumulative package. But that is not how codec support works. A commercial bundle does not convert technically independent decoding paths into one technology.
This matters for pool design. A licensee should tell the implementer what technology is being licensed, why the covered patents are relevant, what obligations govern the programme and how the price was established. Combining HEVC, VVC and AV1 into one undifferentiated vehicle can obscure all four questions. HEVC and VVC are SDO standards associated with F/RAND commitments, whereas AV1 is a consortium specification for which outside owners may have made no commitment at all. The same patent family may even be relevant to more than one codec while carrying different licensing obligations, depending on the path through which the technology entered each specification.
This is particularly problematic when F/RAND-based SDO standards and non-F/RAND private consortium technologies are mixed. As Mattia Fogliacco has argued, doing so risks imposing F/RAND rate discipline, non-discrimination obligations and remedial constraints on patent rights whose owners never accepted them. It can also obscure the different work required to identify and establish the use of third-party patents in a privately developed codec.
Separation does not reject convenience. There is nothing inherently wrong with giving customers an efficient way to have both. The danger arises when convenience becomes conceptual commingling: when separate technologies are treated as one cumulative stack or when the legal obligations applicable to one class of patents are silently imported to another. The better approach is separation – keep the technologies analytically distinct, determine essentiality codec by codec, preserve the legal commitments that actually attach to each portfolio and price each technology in a way that reflects its own value. Modularity is the better one-stop shop: separate programmes, separately explained and valued, made easy to obtain through a coordinated interface.
Value, not chronology
A new codec should be valued for what it contributes. The mere fact that VVC came after HEVC does not mean that the value of VVC patents should be treated as an incremental surcharge on a HEVC licence. Nor does the fact that a device happens to implement HEVC, VVC and AV1 mean that the patents covering those codecs necessarily form one economically indivisible package.
For any codec, the relevant question is what value the technology creates. This can include substantial bandwidth savings at equivalent quality, improved picture quality at a given bitrate, reduced storage requirements, the ability to serve more users with the same network capacity, support for new content formats and use cases, lower compute and implementation costs, a wider scale of deployment, and greater numbers and more types of users or services enabled. A licensing rate should be capable of reflecting these benefits, together with the strength and breadth of the patented technology that produces them. It should not be mechanically determined by the number of older codecs that happen to coexist in the same product.
This does not predetermine any particular rate. It sets the right analytical frame: royalty design should follow the technology, the rights and the market value delivered, not a misleading model in which every new codec is simply viewed as another layer added to the previous one.
Not another layer
A television can decode VVC without decoding HEVC and it can decode AV1 without decoding either. Manufacturers implement multiple codecs because users need access to content in multiple formats, not because one decoder technically requires the other. Licensing policy should begin from this reality. MPEG’s history illustrates the value of open standardisation, F/RAND assurances and efficient pools; and pools function best when they reduce friction around genuinely complementary rights. F/RAND works because patent owners made a particular standardisation commitment. AV1 shows that a consortium’s royalty-free policy applies only to the patent rights actually committed to it and cannot bind the rest of the world.
None of these propositions justifies the treatment of every video codec as part of a single undifferentiated licensing stack. A codec should stand on its own – technically, legally and economically – and its licensing should reflect the value it creates, with clear boundaries and transparent explanations.
Giovanni Ballocca is Sisvel’s Chief Technology Officer. Matteo Sabattini is the firm’s Executive Head of Government Affairs
