H.266/VVC Standard Finalized With ~50% Lower Size Compared To H.265

[cl333r]

...I have no problem paying reasonable returns on R&D, 1000 to 1 is not a reasonable return.

How much specifically would be a reasonable return?

[jaxa]

I have been incredibly annoyed at the slow pace of AV1 hardware decode adoption. It wasn't in Raspberry Pi 4, wasn't in the latest AMD/Nvidia/Intel graphics, etc. I think a couple of 8K TVs and SBCs have it.

AOMedia needs to get AV2 out the door so it can beat H.266 and finally reduce MPEG-LA's relevance. They should just throw in all of the complex, hard to encode features that were scrapped from AV1. Google was originally planning to release successors to VP9 at an 18 month pace, and that's what's needed now.

There are two* successors. The other is EVC. It has a royalty-free subset, and it should also be finished pretty soon.

*Or 3 if you count LCEVC.

Nice strategy. I hope they get crushed into oblivion.

[cl333r]

The problem that I suspect these guys are about to discover is that h264 is "good enough" (and interestingly going out of patent soon). That's the reality of why HEVC never really took off. That extra 30% you get only matters if you are YouTube or Netflix...

You forgot the huge army of pirates - which is most people in the world like myself who download h265 on pirate torrents to save space and have a better quality video - patents and fees are irrelevant to us. But yeah, when AV1 becomes mainstream all other solutions will become irrelevant to me.

[jonsmirl]

How much specifically would be a reasonable return?

I don't mind giving them a 300% return which is very good. But the return has to be on the engineering done. Not the Qualcomm model where they do $10M in engineering, then add on $40M in legal fees, and then ask for 100x return on $50M. Any company that employs more lawyers than engineers is going to be a problem.

Also this 300% return is for IP licenses and it would be the total paid by all licensees. If the company builds an actual product and sells it, they can make as much profit as they can get. i would much rather see companies succeed by producing products and selling them rather than by extracting ransom payments by manipulating standards.

[morydris]

I have been incredibly annoyed at the slow pace of AV1 hardware decode adoption. It wasn't in Raspberry Pi 4, wasn't in the latest AMD/Nvidia/Intel graphics, etc. I think a couple of 8K TVs and SBCs have it.

I think you underestimate how long these things take. As it happens, support is included in the upcoming Intel Rocket Lake. As for Raspberry Pi, it has always been using very trailing edge technology (for cost reasons) so it's no surprise it has no AV1 support.

AOMedia needs to get AV2 out the door so it can beat H.266 and finally reduce MPEG-LA's relevance. They should just throw in all of the complex, hard to encode features that were scrapped from AV1. Google was originally planning to release successors to VP9 at an 18 month pace, and that's what's needed now.

Launching AV2 would be counterproductive if you are trying to establish support for AV1. Thankfully it isn't needed. Like any modern encoding standard, where is a VERY big search space for the encoder so we are nowhere near getting the optimum result supported by the AV1 format. This is how x264 could keep improving year after year and that was for H.264; AV1 has far more flexibility than that.

[mppix]

MP3 was brilliant when released. Even today it's still hard to beat.

mp3 was brilliant when it was released because it enabled thousands of songs on inexpensive mp3 players, that did not have the DACs or headphones to hear the difference to a CD (and CD quality).

Today, mp3 is beaten easily by a number of lossy compression formats at a given bitrate, in particular opus and aac/m4a. Then, people with decent audio setups tend to use lossless compression (flac and alac files) because filespace is not a big concern anymore.

[jaxa]

You forgot the huge army of pirates - which is most people in the world like myself who download h265 on pirate torrents to save space and have a better quality video - patents and fees are irrelevant to us. But yeah, when AV1 becomes mainstream all other solutions will become irrelevant to me.

As a frequent YouTube user on sometimes spotty internet connections, I would love additional bitrate reduction. AV1 and future codecs will also be appreciated if streaming VR video takes off, since that is a good use of absurdly high resolutions and framerates that can be compressed well.

But yeah, there's something exciting about storing a feature length 720p movie at CD-ROM sizes.

I think you underestimate how long these things take. As it happens, support is included in the upcoming Intel Rocket Lake. As for Raspberry Pi, it has always been using very trailing edge technology (for cost reasons) so it's no surprise it has no AV1 support.

I have read the leaks that say that 12-bit AV1 is supported in Rocket Lake, although I'm not interested in buying that.

I mention RPi 4 because it was released just before AV1 started to appear in devices, and is often used for TV/video duties. It's just bad timing for anyone who wants to future proof with devices that support AV1 decode. I probably won't buy any new hardware that doesn't have AV1 decode, which could mean waiting until 2021 or later in many cases. Maybe RDNA2 and Nvidia Ampere won't support it.

Forgot to add: patent trolling might be part of why adoption has been slow, rather than technical reasons.

https://mux.com/blog/did-sisvel-just...-patents-down/
https://www.streamingmedia.com/Artic...ticleID=139636
https://rethinkresearch.biz/articles...-are-in-doubt/

Launching AV2 would be counterproductive if you are trying to establish support for AV1. Thankfully it isn't needed. Like any modern encoding standard, where is a VERY big search space for the encoder so we are nowhere near getting the optimum result supported by the AV1 format. This is how x264 could keep improving year after year and that was for H.264; AV1 has far more flexibility than that.

I assume that AV2 rollout and adoption will be at least somewhat faster than AV1 since it will be building on AV1, and the two will be substantially similar. As opposed to VP9 vs. AV1 (which combined features of VP10, Daala, and Thor).

[cl333r]

I don't mind giving them a 300% return which is very good. But the return has to be on the engineering done. Not the Qualcomm model where they do $10M in engineering, then add on $40M in legal fees, and then ask for 100x return on $50M. Any company that employs more lawyers than engineers is going to be a problem.

Also this 300% return is for IP licenses and it would be the total paid by all licensees. If the company builds an actual product and sells it, they can make as much profit as they can get. i would much rather see companies succeed by producing products and selling them rather than by extracting ransom payments by manipulating standards.

Well I agree, but this will never change as long as it's legal to do what they do now. Think of this - if you you were the head of Qualcomm would you say "Hi, I've decided to cut fees 100 fold because it's immoral what we're doing and so we'll earn only 1M on fees this year instead of 100M, have a nice day dear shareholders and employees", if you do this kind of stuff you'll get kicked from any company. My point is we're against it because we're not part of the extortion system but if we were we'd support it, that's the bitter truth.

[jonsmirl]

Well I agree, but this will never change as long as it's legal to do what they do now. Think of this - if you you were the head of Qualcomm would you say "Hi, I've decided to cut fees 100 fold because it's immoral what we're doing and so we'll earn only 1M on fees this year instead of 100M, have a nice day dear shareholders and employees", if you do this kind of stuff you'll get kicked from any company. My point is we're against it because we're not part of the extortion system but if we were we'd support it, that's the bitter truth.

$6 billion a year and counting for Qualcomm. It's why the US has the most expensive phones and cell service in the developed world. Qualcomm in US courts currently for antitrust but so far they have avoided anything bad happening to them. Its not going to stop, Qualcomm already has 5G locked up in the US too.

[Michael_S]

I buy 4k discs to rip and convert for my plex server. I'll use whatever compression standard is widely supported by my phone, tablet and setup box. I don't have caps and my phone service is unlimited (especially through fast vpn).

What do you use to rip your 4K disks? I've been ripping regular 1080p Blu Rays with MakeMKV and then reencoding to H.265 because it seems to be the most efficient codec that all of my devices support. I use Kodi, not Plex.

MakeMKV support ripping 4K Blu Rays but you need LibreDrive hardware and I don't have any (yet).