October 2020 is shaping up to be the biggest month in V-Nova’s rollercoaster history, one which has seen the UK-based compression vendor pivot from being the target of much resistance from the video industry, to being instrumental in the delivery of a new standardized codec in the form of LCEVC (low complexity enhancement video coding).
In less than three months’ time, LCEVC (aka MPEG-5 Part 2) will be subject to editorial review from the standards body powers that be – but what do we know about the critical licensing terms for LCEVC? Faultline managed to scratch below the surface of the unfinished LCEVC licensing structure ahead of completion of the final standard – with V-Nova CEO and co-founder Gudio Meardi promising, “We will get it right, don’t worry.”
“It will be cheaper than royalty free, because royalty free isn’t royalty free. People must pay a fair price to access and use LCEVC, but the standard will be made available and everyone will be able to use it. For the first year, we are more interested in building traction than we are about immediate economic return. We have a responsibility to be a standard now,” confessed a familiarly honest Meardi.
V-Nova has come such a long way from the days of being told not to bother driving a standards-based approach to compression. Even Google tried to talk V-Nova down, having tried and failed to standardize VP9, so clearly this is not a question of money and power, but one of a unique combination of V-Nova’s engineering tenacity and the persuasive skills of Meardi himself. This is the same Meardi whom Faultline has argued with in the past about why V-Nova doesn’t have any customers. Today, he is almost unrecognizable in his relaxed (but still eccentric) attitude.
From now until October, V-Nova is effectively in limbo, working away on some key deployments with customers and getting creative with a lot of independent third parties. Publicly available libraries will be made available online for everyone to access come October/November time once reviews have completed, and the new lcevc.com website just landed this week to help out on the marketing side of things.
We tentatively slipped in how the MC-IF (Media Coding Industry Forum), the group promoting VVC (versatile video coding), only last week said how VVC’s licensing structure would be based on the FRAND (fair, reasonable and non-discriminatory) principle. Meardi’s initial skepticism of FRAND bordered on dismissal, before deciding, “Within and beyond FRAND, we are committed to making LCEVC easily licensable and fair. LCEVC has 6 tools versus 100 tools for VVC, so the chance of hitting on a submarine patent is very low.”
Meardi’s musings echo Faultline’s from last week, when we highlighted how it was a tall order for the MC-IF to avoid a repeat of the patent bloodbath between HEVC Advance, MPEG-LA, and Velos Media. A potential silver lining is that most patent holders will still be hung up with HEVC and AVC, considering the two are still responsible for processing 90% of all internet-delivered video bits to over 10 billion active devices worldwide. The continued dominance of HEVC and AVC today is precisely why LCEVC is so well positioned to triumph, proving its worth for enhancing established codecs, reducing encoding and CDN costs by as much as 50%, while also significantly reducing computational complexity of next-generation codecs like the power-hungry AV1 (which sucks approximately 50x more power than HEVC).
Results of a recent test show LCEVC x265 winning against native x265 in 99% of the clips on Netflix’s VMAF (Video Multi-method Assessment Fusion) quality metric, and 61% on PSNR (peak signal-to-noise ratio). Here, the combined LCEVC-HEVC stream was lighter to encode and decode than the native format at the same resolution.
So, once LCEVC is given the final stamp of approval and licensing terms are sorted, commercial deployments will swiftly follow, arriving before the end of the year – months ahead of commercial hardware encoders for the newly launched VVC. Just to be clear, comparing LCEVC and VVC is really a moot point, and convincing people as such will be one of Meardi’s biggest challenges in the months ahead. LCEVC is designed for software-only environments, while VVC is a hardware-based codec being driven by the likes of Qualcomm.
The journey into LCEVC, which we must also reiterate does not compete directly with AVC, HEVC, AV1, VP9, EVC or VVC, is reflective of V-Nova’s evolution into a compression company that no longer rivals the likes of Harmonic, Ateme or Bitmovin, but is capable of partnering them. These very companies resisted V-Nova’s approaches as a non-standard proprietary codec around the 2015 to 2016 timeframe, adamant to prove their own video compression resources as more than adequate for the job.
Meardi’s vision for the future is therefore to influence existing customers of these vendors, to the extent that a customer pushes the likes of Harmonic to integrate LCEVC libraries. In an ideal world, V-Nova would convince Harmonic to market the technology directly to its customers, once Harmonic eventually realizes that this technology would actually help it to sell more of its own products. A few years ago, V-Nova would have been laughed out of the room for suggesting such a scenario, but LCEVC has changed everything – to the extent we can accept a future where vendors (perhaps not one as resistant as Harmonic) freely market LCEVC as part of their product portfolios.
Meardi likened his company’s changing business model to Dolby – in the sense that V-Nova sees encoding vendors for LCEVC in the same way Dolby sees them for penetration of AAC audio, Dolby AC-3 or Dolby Atmos.
He bundled V-Nova’s future target customers into two simple buckets:
- The big giants or the nimble players that create video workflows by themselves by assembling open source and best-of-breed components.
- The rest of the market, which has fewer internal integration skills and so buys turnkey solutions from third-party vendors (e.g. encoder vendors, online video platforms, etc.).
Set 1 is growing by the day, according to Meardi, as more companies discover the efficiency gains of building a custom video workflow – either in the cloud or on-prem – compared to buying professional. Set 2, meanwhile, was where he admits to V-Nova making some crucial mistakes in the 2016 to 2018 era, wasting time and resources targeting customers like Sky Italia (V-Nova’s biggest long-term backer), when in reality these types of customers were too caught up with the likes of Harmonic, Ateme, AWS Elemental, and Ericsson/MediaKind.
V-Nova now has its sights firmly on set 1.
It’s time to talk about the elephant in the room – hardware. While LCEVC is software retrofittable, the technology can and is being used in hardware-based deployments as seen with FPGAs from Xilinx, which have ported the Perseus Plus codec from V-Nova, which LCEVC is based on, since early last year. This is designed to improve scale for large deployments and encoding efficiency for real time services, especially where some form of UHD is involved. There are two versions, Perseus XSA to accelerate encoding for existing live or VoD pipelines where codecs have already been deployed, and Perseus XDE, which incorporates both the V-Nova software and HEVC codec on the same FPGA card for new services.
Meardi added some extra color this week, explaining that LCEVC in Xilinx FPGA comes into play for both “head” and “tail” in different ways, where head video represents the 10% of video feeds that make up 70% of an operator’s CDN consumption, while the tail pertains to the remaining 90% of video feeds accounting for the 30% or so of excess CDN bandwidth.
“With head content, LCEVC provides by far the best compression, and by the way it impacts 100% of the delivery, because it allows to enhance both H.264 and new codecs. For instance, with Xilinx we also offer LCEVC VP9, LCEVC HEVC and LCEVC AV1, so operators can enhance all components of a multi-codec workflow. For tail content, LCEVC H.264 provides by far the highest density and cheapest transcoding costs,” he explained.
LCEVC therefore is being pitched as a sort of savior for hardware designers, as it uses general purpose processing hardware components such as SIMD (single instruction, multiple data) instructions and GPUs – so does not require new hardware blocks.
Software-centricity in short means LCEVC can arrive commercially much earlier than other codecs. Yet still there are experts in the room who underestimate the problem with hardware-based codecs. AV1, EVC or VVC can of course be deployed in software at low resolutions, best suited for powerful PCs with a permanent power supply. But the caveat is that the compression benefits of these next-gen codecs are only fully realized at high resolutions – while we all know that the present and future of video streaming is with battery-powered devices. AV1, EVC and VVC could therefore be left powering a niche installed base as low as 10%.
We couldn’t get on the phone to Meardi without dropping in some comments from Leonardo Chiariglione, who recently left his position at the MPEG Group and made sure the ship went down with him, as the standards body transitions instead to the ISO. One particular Chiariglione post said how LCEVC could compete with VVC in terms of quality, in appropriate conditions, if added as a layer on top of EVC. But he also said that if EVC succeeds in its own right and patent holders agree to a decent license, then VVC has a chance of success. Of course, LCEVC is exactly that – an enhancement to other codecs – but the fact that it is being pitted against VVC is a perpetual thorn in Meardi’s side.
“LCEVC is completely neutral to all these codecs, and it enhances also VVC. In addition, there is such a big choice when you have to move to a new hardware codec, that you lose all support for existing codecs,” he said. This boils down to what devices your subscribers have, and Meardi reckons that more than half the devices people use in five years’ time will be manufactured pre-2020 – meaning that the circulation of hardware come 2025 will still be hugely reliant on H.264 and H.265.
An amusing recent move for Meardi was Google’s decision to add support for AV1 on videoconferencing for mobile, predicting that no handsets are capable of processing the power-guzzling AV1 codec without spontaneously combusting. While instead LCEVC in combination with AV1 may actually make frequent use of AV1 feasibly, since it would work at lower resolutions.
We concede our assessment of LCEVC is largely flattering, so we should emphasize that the job is far from done. V-Nova has a Herculean task ahead in convincing decision makers (primarily the open source advocates in set 1) that LCEVC is not an alternative codec – but a turbocharger codec. Meardi’s frank assessment is that much of the industry remains “very confused by the codec mess.” Our concern is that V-Nova’s emphasis on marketing LCEVC over making money from LCEVC could potentially plunge the company into financial difficulties if decision makers cannot be swayed about the technology’s disruptive potential.
As a teasing closing item, V-Nova has some rather left field plans up its sleeve for LCEVC beyond the media industry, which we will hopefully be able to report on in more detail later this year. Think the David Attenborough of video compression.