Experts Panel Discusses Codecs, 8K and More
Rethink Research hosted a panel discussion around the adoption and use cases for the latest crop of codecs in 8k.
- Carl Furgusson – Mediakind
- Jan de Cock – Synamedia
- Janne Pelkonen – IdeaNova Technologies
- Fabio Murra – V-Nova
Most of these speakers have efforts in HEVC, VVC, AV1 and LCEVC – more on the broadcast side vs OTT. The webinar took place before IBC 2022, so the speakers first talked about what they would show at the event (except for IdeaNova, which does in-flight entertainment).
The discussion quickly turned to VVC – its use cases, time frame for rollout, licensing issues and more. Mediakind says they are investing in developing VVC. Still, they see adoption coming later rather than soon, with 8K broadcast being the primary (and singular) use case for it (although they did not show anything 8K-related at IBC). Codec adoption seems to be on a decade-long cycle from development to chipset availability to mass adoption. And the pandemic has delayed things even more – certainly on the silicon side for these new codecs.
Synamedia sees VVC as the format of the future, but the royalty issues that dogged HEVC are cropping up again for VVC, creating uncertainty around adoption and delays.
The V-Nova perspective on VVC is to look for where there will be traction. DVB has added VVC to its toolbox, and ATSC 3.0 will also look at it. But the leader here may be Brazil and its TV 3.0 initiative. VVC is specified along with LCEVC as the scalable element and will be used for HD, 4K and 8K. The test will begin for real-time encoding and decoding in early 2023 (but perhaps not for 8K yet).
IdeaNova is more focused on AV1 than VVC for in-flight entertainment, with the need to support a wide range of platforms and devices available today.
Furgusson delved deeper into the licensing issue, defining three main areas where IP holders can extract fees: the device maker; the service provider; and the people in the middle, like the browser provider. The Sisvel pool for AV1 focuses on the device maker, as do the HEVC and VVC pools. He explained that there are many more service providers than device makers making reporting and royalty collection more complex. The browser crowd is a mix of big and deep pockets, which is also more complicated.
V-Nova is adopting a service provider licensing model, which means there is no fee to add LCEVC to silicon, encoders or playback devices. There is a capped fee for larger service providers to minimise the reporting concerns. V-Nova also sells implementation of its technology as software, CPU or GPU implementations, or as FGPA or ASIC options.
In EVC adoption, all agreed there was very little investment in this codec. It is unclear whether it can support a use case that the other codecs can’t.
For MediaKind, the bigger question he posed was whether any codecs could replace MPEG-4 for streaming applications. The business case for an encoding company in the OTT space is complex as the big streamers develop their own encoders. So, there have to be enough sales to justify the multi-million-dollar investments needed to make a profit. An audience question stated that AV1 did not support live. It does, said Murra. Live AV1, like SVT-AV1, takes more computational power, and some computationally-intensive tools need to be turned off, which will impact quality to some degree. He noted that V-Nova worked with Intel to do around a million encodes for AV1 with and without LCEVC to show this trade-off space with a clear difference in performance (or compute or bit rate) when using LCEVC. (See this IBC V-Nova video for more – https://youtu.be/ESyigwf-8Lo).
Synamedia noted that HEVC already works well for live streaming, and they are showing an 8K demo at IBC. AV1 will help push bitrates down, but VVC will help significantly.
Furgusson said that the availability of silicon decoders would ultimately determine the timeline for adopting any codecs (the just-announced iPhone 14 does not support AV1, for example).
The panel then turned to initiatives to reduce power consumption in the video encoding and delivery ecosystem. De Cock said the encoder power per channel is going down yearly. Synamedia says they are down 60% in the last two years. They are also implementing a just-in-time encoding and packaging strategy for delivery to the CDNs; there are more power savings there too. But efforts should be made too on the decoder side for hardware assistance to reduce power at the playback device.
Furgusson distinguished between the choices available to encoders and, therefore, power consumption for live vs offline content. For live, they are usually limited by the hardware available on-premises or the nodes available for cloud encoding. That limits the processing power, which means some bit rate and/or quality trade-offs. Over time, they get better at improving quality and reducing the bit rate for these same assets. But it is true that, in general, processing increases over time and, with it, power consumption.
Murra did not call AV1 a green codec, but he did note that when using it with LCEVC, you get bit rate savings and lower compute for the same quality, which means less power consumption. Their paper also looked at the decode side, noting that LCEVC can prolong a phone’s battery life by 50% on specific devices.
The whole panel discussion can be viewed here.
What about AVS3?