libavif 1.0 Released With Stable ABI, Experimental AV2 Support

[Artim]

The first one is barely out and they're already trying to replace it.

It's called progress. That's how humans are supposed to work.

There's always room for improvement and it's very likely that some techniques that where discussed for AV1 where seemed too complicated and have been delayed. This way raw CPU power has time to improve to be able to handle those, and who knows which new instruction set additions in the meantime have been positioned to help with that. So of course they have to make gradual improvements or nobody would be able to use it. And seeing how long AV1 took and still takes to get hardware support, it's about time to finalize AV2 so Chip Designers can get to work on hardware support.

[Silent]

To be fair, in the real world, people are still happy with h264 and that's 20 years old.

The reason h264 is popular is not only because people are happy with it, but because AV1 hardware barely hits market for cheap CPUs, not supported by all low-and-mid AMD GPUs and only AM5 CPUs, and so on. If someone buy good but not top PC 2-3 years ago it still can be without hardware support - how it can be popular then? Intel have it more reasonable but still...

[Toggleton]

if av2 is based on av1, I doubt it will take as long as av1 to get widely available hardware support for encode/decode

AV1 is based on VP9(it was called for a long time VP10). Would guess they will try to make AV2 easy to be support by Hardware vendors. As a lesson learned from the AV1 cycle
Would see this support for AV2 more as easy way to test around the new tools but no where near feature freeze.

[Artim]

AV1 is based on VP9(it was called for a long time VP10). Would guess they will try to make AV2 easy to be support by Hardware vendors. As a lesson learned from the AV1 cycle
Would see this support for AV2 more as easy way to test around the new tools but no where near feature freeze.

That's the same promise they gave with AV1. The biggest point in it was that all relevant hardware vendors were involved from the get-go, so they could immediately start developing hardware support and give input to make support easier. The second point was that it wasn't basically developed by Google on their own. That's also why Apple threw a tantrum for years refusing to support VP9, until they caved in a year ago or two. So don't get your hopes too high.

[Jabberwocky]

It's called progress. That's how humans are supposed to work.

There's always room for improvement and it's very likely that some techniques that where discussed for AV1 where seemed too complicated and have been delayed. This way raw CPU power has time to improve to be able to handle those, and who knows which new instruction set additions in the meantime have been positioned to help with that. So of course they have to make gradual improvements or nobody would be able to use it. And seeing how long AV1 took and still takes to get hardware support, it's about time to finalize AV2 so Chip Designers can get to work on hardware support.

In theory: Yes.

Meanwhile in practice: Fuu yeah! Give us another one. Fill those landfills with more ewaste!

[Artim]

In theory: Yes.

Meanwhile in practice: Fuu yeah! Give us another one. Fill those landfills with more ewaste!

Uhm...how about no? There is no ewaste being produced. Any hardware implementation (I know of), wether in a CPU/SoC or a GPU has multiple hardware codec. Taking Intel 13th Gen Core i processors, they support MPEG-2 decode while having dropped encode, as they have all VC-1 and VP8 support, but have full support for H.264, H.265, and VP9 and AV1 up to 10 bit, with 12 bit decode only support, and JPEG.

Also, it takes many years until a new hardware codec generation is available. Full VP9 Support was introduced first in 2016 with Apollo Lake Celerons, Pentiums and Atoms. AV1 was first fully supported with Meteor Lake in 2022, if you ignore that both VP9 and AV1 can only encode up to 10 bit in hardware. Thats 6 years inbetween. So even if you are stupid enough to always buy the first generation Intel Chip that comes with hardware encoding and decoding of the latest version of the codec, 6 years of use isn't what I would call e waste. And nobody is keeping you from selling the old CPU or Laptop. Accept maybe AMDs simetimes used Vendor Lock, that's what e waste looks like.

Also hardware accelerators are also shipped in dedicated GPUs. 6 years for a GPU is quite a long time if you are big in gaming.

[Old Grouch]

Personally, I like the Kolmogorov codec that's in development. Guaranteed minimal size of the compressed bitstream, but right now, the encoding times are a bit unpredictable.

[Jabberwocky]

Uhm...how about no? There is no ewaste being produced. Any hardware implementation (I know of), wether in a CPU/SoC or a GPU has multiple hardware codec. Taking Intel 13th Gen Core i processors, they support MPEG-2 decode while having dropped encode, as they have all VC-1 and VP8 support, but have full support for H.264, H.265, and VP9 and AV1 up to 10 bit, with 12 bit decode only support, and JPEG.

Also, it takes many years until a new hardware codec generation is available. Full VP9 Support was introduced first in 2016 with Apollo Lake Celerons, Pentiums and Atoms. AV1 was first fully supported with Meteor Lake in 2022, if you ignore that both VP9 and AV1 can only encode up to 10 bit in hardware. Thats 6 years inbetween. So even if you are stupid enough to always buy the first generation Intel Chip that comes with hardware encoding and decoding of the latest version of the codec, 6 years of use isn't what I would call e waste. And nobody is keeping you from selling the old CPU or Laptop. Accept maybe AMDs simetimes used Vendor Lock, that's what e waste looks like.

Also hardware accelerators are also shipped in dedicated GPUs. 6 years for a GPU is quite a long time if you are big in gaming.

You are fixated on a tiny part of the market who will likely sell or ethically dispose their hardware.

In the real world most consumers have no idea what you just said. They look at the box of their phone, smart TV, firestick, mediabox, tvbox, projector, digital camera, IPcam, USBcam, NUC, SBC, laptop, etc... they look for marketing terms like HD Ready, Full HD, 4k, HVEC, AV1 or whatever other jargon someone they know mentioned. They throw their old cheap devices in the garbage and go on with life.

If AV2 is going to be in development for a decade+ then this argument isn't relevant. In that case AV1 could help produce less waste if the term becomes popular enough. If AV1 becomes too popular there's obviously still concerns about AV1 being challenged in court, but there isn't much that we can do about that now.

[Artim]

You are fixated on a tiny part of the market who will likely sell or ethically dispose their hardware.

In the real world most consumers have no idea what you just said. They look at the box of their phone, smart TV, firestick, mediabox, tvbox, projector, digital camera, IPcam, USBcam, NUC, SBC, laptop, etc... they look for marketing terms like HD Ready, Full HD, 4k, HVEC, AV1 or whatever other jargon someone they know mentioned. They throw their old cheap devices in the garbage and go on with life.

If AV2 is going to be in development for a decade+ then this argument isn't relevant. In that case AV1 could help produce less waste if the term becomes popular enough. If AV1 becomes too popular there's obviously still concerns about AV1 being challenged in court, but there isn't much that we can do about that now.

That's just ridiculous crap. People ususally know more about recycling than about any of those technical terms. 4k and depending on their age HD Ready and Full HD are the only ones they will recognize. If their relatives don't tell them to look out for HEVC or AV1, they won't. You also won't see those advertized that much. That's why e.g. where I live they had to put a DVB-T2 Logo on devices to make sure people can identify HEVC compatible DVB-T2 equipment, as they can't tell people to look out for HEVC. And about everywhere else DVB-T2 uses AVC, not HEVC.

Of course, the development of AV2 won't take a decade anymore, but it's also highly unlikely to be released this year, probably even next year. And even then it will take years for hardware support to emerge. So there might be roughly about a decade inbetween.

And also, the actual reason people usually throw out their old stuff isn't because they would know about video codecs, but because they buy cheap crap with planned obsolescence built into the product. If the device doesn't tell them "no support, buy generation xy" they won't, because people are just too cheap. If it still works, they'll use it.

[Quackdoc]

AV1 is based on VP9(it was called for a long time VP10). Would guess they will try to make AV2 easy to be support by Hardware vendors. As a lesson learned from the AV1 cycle
Would see this support for AV2 more as easy way to test around the new tools but no where near feature freeze.

av1 is based on a plethora of techniques, not just vp9, multiple technques came together, it is also built majorly upon daala and thor