AOMedia's "AVM" Repository Serves As Reference Implementation For Eventual AV1 Successor

[direc85]

[...] if I can't use Firefox with VA-API at all?

Well, apparently the bug has been fixed: https://bugzilla.mozilla.org/show_bug.cgi?id=1751363 Can't wait for FF102 and testing it with both Intel and AMD hardware!

Michael This might be of newsworthy (once it's actually released).

[Danny3]

How about fixing and completing the acceleration of current-gen video codecs on current-gen hardware first? What use is a new shiny codec if I can't use Firefox with VA-API at all?

Indeed, they should help Firefox have proper high performance decoding on all platforms first!

[smitty3268]

Release? By definition patents can't be disclosed. They are disclosures.

Companies are granted a time-limited monopoly in return for disclosing scientific and technical developments to the public. Thus they can continue to use it as if it had remained secret, yet the public still benefits. At least, that's how it was supposed to work.

The list of what patents they believe AV1 infringes is what was being referred to as needing to be disclosed, not the details of the patents themselves.

It's easy enough to claim that something violates your patents, but you need to make it clear exactly which patents those are before anyone can verify such claims and figure out how serious they really are or if it's trivial to work around them.

[Developer12]

Ah yes, Sisvel. The entity that said they'd disclose all patents needed for AV1 and then didn't do anything.

Release? By definition patents can't be disclosed. They are disclosures.

Companies are granted a time-limited monopoly in return for disclosing scientific and technical developments to the public. Thus they can continue to use it as if it had remained secret, yet the public still benefits. At least, that's how it was supposed to work.

[brucethemoose]

Well, when "AI" gets involved with compression, we can probably expect shiny pictures after decompression that may just not resemble the semantics of the original image. Like, in a much simpler way, the infamous Xerox compression did.
We already see this in video games, were artificial neural networks are used to "guess" details that just are not there in an to-be-upscaled rendered image. Often those "guessed" details are plausible, sometimes they are just nonsense.
For a video game output, adding some nonsense to a picture is not as problematic though, then let's say guessing stuff in a war crime video.

Again, AI doesn't equal aggressive upscaling. In fact, I think its more useful in the denoising/noise synthesis steps and other things like contextual parameter adjustment, block matching, and other bits of tricky decision making. I don't see much "hallucination" risk there.

But resolution doublers can still behave conservatively, and the kind of super res feature AOM is thinking about is very different than traditional image doubling anyway... otherwise there would be no need to integrate it into the codec.

Indeed, I'm mostly using H.264 for video encoding, because it is way faster to compress in software than any of the later standards, and I don't mind spending just a little more size to get decent quality. Those pushing for new video codecs are usually only those people who want to lower bandwidth cost for "streaming", and they lower them to the extent that the gain in coding efficiency is more than compensated.

The thing about H.264 is that its missing some really useful features, like (widely supported) bit depths higher than 8 bits, or better rate control modes on the available encoders. Even if you encode at crazy high bitrates, its going to cost some quality.

For quick local encodes, hardware HEVC is best IMO.

And for what its worth, YouTube has put all the savings from AV1 into increasing quality at the same bitrate rather than reducing the bitrate of the stream. Netflix seemingly uses it to save storage space for offline playback (albeit only at low res for now).

[arQon]

Those pushing for new video codecs are usually only those people who want to lower bandwidth cost for "streaming", and they lower them to the extent that the gain in coding efficiency is more than compensated.

Right: the point is to reduce opex for YouTube, Netflix, etc, and in the short term that happens by pushing some of the cost onto the user in the form of higher-power decoding (though for users on mobile devices or other "pay-per-byte" connections, the saved data cost outweighs the additional electricity, the same as it does for the CDNs). Modern codecs aren't really intended for archival or LAN use though, and the math rarely works out for those cases.

Codec development is a very slow process these days. It needs to be pretty much "done" about 5 years ahead of when you hope for mainstream adoption, because between patent trolls and SoC support it takes that long to have any presence at all, let alone wide support. The actual development starts several years ahead of that, but even that stage is usually more about bureaucracy than technology, because each of a dozen or so companies are all trying to wedge their own patents into the spec so that they can profit from it more than the others. The "eventual" Michael used in the headline is very much the right term: AV1 is barely a blip on the radar so far, and still has years to go before it becomes commonly used.

The nicest thing about where we currently are though, of course, is that MPEG-LA can go f**k itself. That's not to say that AANG won't simply replace them in the exact same role a few years from now, but in the short term at least I expect the MAD analogy someone offered earlier in the thread is a pretty good one.

BTW, the situation with MS and HEVC is so funny that it deserves special mention. MS's greed over HEVC patents is what created AOM in the first place - and then they *pulled* HEVC support from W10 a couple of years after release because they didn't want to pay the licensing cost for it themselves...

[dwagner]

Indeed, I'm mostly using H.264 for video encoding, because it is way faster to compress in software than any of the later standards, and I don't mind spending just a little more size to get decent quality. Those pushing for new video codecs are usually only those people who want to lower bandwidth cost for "streaming", and they lower them to the extent that the gain in coding efficiency is more than compensated.

[OneTimeShot]

It's probably good to keep developing video codecs, but after a certain point... well... h264 is 20 years old and still going strong. There's only so much that a video codec needs to do.

Video patents are expiring and we just need mass adoption of a patent-free codec now. That could just be AV1... or well... h264 in a bit.

[dwagner]

I know AI gets thrown around as a buzzword and slapped on things that don't deserve the term, but there are stages of the AV1 pipeline that it could really dramatically help.

Well, when "AI" gets involved with compression, we can probably expect shiny pictures after decompression that may just not resemble the semantics of the original image. Like, in a much simpler way, the infamous Xerox compression did.
We already see this in video games, were artificial neural networks are used to "guess" details that just are not there in an to-be-upscaled rendered image. Often those "guessed" details are plausible, sometimes they are just nonsense.
For a video game output, adding some nonsense to a picture is not as problematic though, then let's say guessing stuff in a war crime video.

[Jan Ozer]

FYI to all - Sisvel posted a patent list for AV1 and VP9 back in May 2020. https://www.sisvel.com/blog/audio-vi...9-patent-pools

The list is here. https://www.sisvel.com/images/docume...ntList_AV1.pdf