3 ms·

"AV1 adoption is accelerating" But before it is widely used and accepted, here's AV2 for you to have compatibility issues with in the wild With the ubiquity o

"AV1 adoption is accelerating" But before it is widely used and accepted, here's AV2 for you to have compatibility issues with in the wild With the ubiquity of h.264 and the patents expiring, will anyone but streamers care?

craftkiller 1y ago
I do. I watch scifi shows over the internet with my friends. We watch it together on a web page with an html5 video element served out of my apartment. I've had to re-encode it to 3 megabits per second (to avoid stutter/buffering) with no b-frames. When initially setting it up, I tested both H264 and AV1 at that bitrate and H264 looked considerably worse. No surprise there, considering H264 is old enough to drink (22 years) whereas AV1 is only 7 years old.
free_bip 1y ago
I understand the point you're trying to make, but I think I can at least sort of understand why they're going with this speed of release cadence. If the release cadence is too slow, you might end up with another JPEG situation where the new codec is undeniably better in every way, but nobody wants to implement it since the old standard was around for so long without any competition.
- dylan604 1y ago
  If you're going to do that, then the new thing must be so much better than the old thing that makes the pain of switching to the new thing worth while. By the time h.265 encoding was trying to gain traction, h.264 encoding speeds were very fast. The image improvement was negligible with the main benefit being smaller file sizes. For the average user, the increased encoding times did not justify that. The switch from MPEG-2 to h.264 had very noticeable quality improvements so it did make it worth while for the slower encodes until h.264 was locked and key code included in CPUs. It was similar to the adoption rates of DVD from VHS compared to Blu-ray from DVD.
  - cubefox 1y ago
    > The image improvement was negligible with the main benefit being smaller file sizes. That's a contradiction because quality improvement and file size improvement are just two sides of the same coin. You can't have a large quality improvement at the same bit rate without having a large file size reduction at the same quality.
    - dylan604 1y ago
      Yes, I stipulate your point, but people are not looking for better encodes at the same bit rate. They are only looking for not worse quality at much lower bit rate. One project I was working on had a hard file size limitation, and switching to h.265 was able to get the size under the limit. At one point, I was told to make it smaller by X MBs. The image wasn't even being looked at with each change. This is why I make the distinction the way that I did.
  - mirashii 1y ago
    > For the average user, the increased encoding times did not justify that. The average user is a consumer of media, not doing encoding themselves. A one time cost for higher encoding to save bandwidth / storage space many times over is almost always going to make some amount of sense. The real issue here is just a standard chicken-and-egg problem. To use a new codec, you need it to be supported in end user devices. To get it to be supported, you need to show demand... for a thing that nobody can use yet.
    - dylan604 1y ago
      The switch from MPEG-2 to H.264 had that in droves with the cable companies updating their set top boxes. That was enough demand to drive hardware development. I don't know how many people have cut the cord as I haven't played in that field to see numbers in quite some time. I'm guessing there won't be a big push to switch out boxes to support AV2. Even shiny round disc sales are plummeting to the point there's not really a need for an at home player to use it either. This really feels like something for streaming only. Those can all be software decoders. Hopefully, they can make it work well on efficiency cores and not require GPU cores for decode. As you say, vast majority of people won't need to encode so sure make a GPU encoder
- throw0101a 1y ago
  > […] but nobody wants to implement it since the old standard was around for so long without any competition. Perhaps: * https://en.wikipedia.org/wiki/Lindy_effect https://en.wikipedia.org/wiki/Lindy_effect
- mrandish 1y ago
  > If the release cadence is too slow... Another reason to get a new video codec standard finalized sooner is that hardware implementation and deployment in mass market consumer SOCs is glacially slow. On the software side, encoder and decoder performance tends to improve meaningfully in the first few years as optimization occurs. And those running large media distribution platforms prefer at least 12-18 months to evaluate and implement a new codec.
kevincox 1y ago
h266 (aka VVC) is seemingly ~~late in development~~ (edit: It has been released but doesn't have much hardware uptake yet). They probably want to ensure that people are aware that it will be matched so that they don't commit to it and AV2 ends up a bit late to the party like AV1 is compared to h265 where it had a notable compatibility lead. If people know that AV2 is coming and competitive they may avoid adopting h266 and wait for the open alternative to ship.
- cogman10 1y ago
  AV1 competes with h266. They were released near the same date. In many ways h266 has already lost the battle as nobody supports it even though it's been around just as long as AV1. h267 is still in development and due to be released in 2028. That's the actual competitor with AV2.
  - kevincox 1y ago
    I guess it depends how you looked at it. Performance-wise AV1 seems more similar to h265. Hardware support wise h266 seems to have only had shipping hardware this year? So I guess neither of these line up 1:1. I tend to see h265 and AV1 competing pretty hard right now so I tend to think of those as one generation and presumably h266 and AV2 will compete as the next generation.
    - cogman10 1y ago
      The way MPEG runs things there are some oddness with the likes of H265. MPEG likes to add little extensions to a codec over time. That's why H265 was first released in 2013 but ultimately had like 10 different extensions after that fact (the last one in 2024). AV1 and VP9/VP8 before that have, in contrast, been pretty much static after they were released. AV1 has had a single errata after it's release. So I could see why you'd see H265 as the competitor. I mostly don't simply because I believe they explicitly stated that they were trying to be competitive with 266. I personally prefer the way AOMedia is running things and I suspect hardware manufacturers do as well. No licenses and AOM is creating open source reference encoders/decoders. They are working very hard to make it easy for manufacturers to be able to pick up the spec and run with it. Keeping the stream standard static for a long period also means manufacturers don't have to worry that they won't get a new extension next year. Content encoders are also reasonably guaranteed that their encoding with today's software still works with yesteryear hardware. MPEG, on the other hand, is paywalling the crap out of everything.
zamadatix 1y ago
Media companies want to save on bandwidth, content types keep changing (e.g. HDR), people want live calls to work whenever, and so on as well. At the same time, GIF still isn't completely gone either just because people needed more than what GIF could give them.
ksec 1y ago
That is what I have been saying for close to 10 years now. We are quite close to Patents free H.264 High Profile. You will always have to provide H.264 as baseline due to compatibility. And that has a cost of storage. Bandwidth cost have been declining fast, and with AI Capex it doesn't seems to be slowing down either. Meanwhile Storage cost hasn't drop, if anything recent trend suggest it may have plateau and went up for HDD. H.264 1080P 2Mbps is good enough for a lot of things. Just like how MPEG-2 is still getting encoder improvement ~30 years later so is H.264 encoder. There are other codec like LCEVC which you can apply on top of H.264 can provide up to 60% bitrate reduction for 4K content. This saves on storage cost and provide enough benefits. It is only in streaming services like Netflix where the catalog of video are low enough they could afford to re-encode it every 5- 8 months and storage cost is minimal. Again a new codec introduction is easily a 10 years task. Higher Speed PON is already being tested, while others are working on NGS-PON2 roll out. 5G Home Broadband with Massive MIMO. The true free and open Video Codec may not be AV1 or AV2, but H.264.

craftkiller 1y ago

I do. I watch scifi shows over the internet with my friends. We watch it together on a web page with an html5 video element served out of my apartment. I've had to re-encode it to 3 megabits per second (to avoid stutter/buffering) with no b-frames. When initially setting it up, I tested both H264 and AV1 at that bitrate and H264 looked considerably worse. No surprise there, considering H264 is old enough to drink (22 years) whereas AV1 is only 7 years old.

free_bip 1y ago

I understand the point you're trying to make, but I think I can at least sort of understand why they're going with this speed of release cadence. If the release cadence is too slow, you might end up with another JPEG situation where the new codec is undeniably better in every way, but nobody wants to implement it since the old standard was around for so long without any competition.

dylan604 1y ago
If you're going to do that, then the new thing must be so much better than the old thing that makes the pain of switching to the new thing worth while. By the time h.265 encoding was trying to gain traction, h.264 encoding speeds were very fast. The image improvement was negligible with the main benefit being smaller file sizes. For the average user, the increased encoding times did not justify that. The switch from MPEG-2 to h.264 had very noticeable quality improvements so it did make it worth while for the slower encodes until h.264 was locked and key code included in CPUs. It was similar to the adoption rates of DVD from VHS compared to Blu-ray from DVD.
- cubefox 1y ago
  > The image improvement was negligible with the main benefit being smaller file sizes. That's a contradiction because quality improvement and file size improvement are just two sides of the same coin. You can't have a large quality improvement at the same bit rate without having a large file size reduction at the same quality.
  - dylan604 1y ago
    Yes, I stipulate your point, but people are not looking for better encodes at the same bit rate. They are only looking for not worse quality at much lower bit rate. One project I was working on had a hard file size limitation, and switching to h.265 was able to get the size under the limit. At one point, I was told to make it smaller by X MBs. The image wasn't even being looked at with each change. This is why I make the distinction the way that I did.
- mirashii 1y ago
  > For the average user, the increased encoding times did not justify that. The average user is a consumer of media, not doing encoding themselves. A one time cost for higher encoding to save bandwidth / storage space many times over is almost always going to make some amount of sense. The real issue here is just a standard chicken-and-egg problem. To use a new codec, you need it to be supported in end user devices. To get it to be supported, you need to show demand... for a thing that nobody can use yet.
  - dylan604 1y ago
    The switch from MPEG-2 to H.264 had that in droves with the cable companies updating their set top boxes. That was enough demand to drive hardware development. I don't know how many people have cut the cord as I haven't played in that field to see numbers in quite some time. I'm guessing there won't be a big push to switch out boxes to support AV2. Even shiny round disc sales are plummeting to the point there's not really a need for an at home player to use it either. This really feels like something for streaming only. Those can all be software decoders. Hopefully, they can make it work well on efficiency cores and not require GPU cores for decode. As you say, vast majority of people won't need to encode so sure make a GPU encoder
throw0101a 1y ago
> […] but nobody wants to implement it since the old standard was around for so long without any competition. Perhaps: * https://en.wikipedia.org/wiki/Lindy_effect https://en.wikipedia.org/wiki/Lindy_effect
mrandish 1y ago
> If the release cadence is too slow... Another reason to get a new video codec standard finalized sooner is that hardware implementation and deployment in mass market consumer SOCs is glacially slow. On the software side, encoder and decoder performance tends to improve meaningfully in the first few years as optimization occurs. And those running large media distribution platforms prefer at least 12-18 months to evaluate and implement a new codec.

kevincox 1y ago

h266 (aka VVC) is seemingly ~~late in development~~ (edit: It has been released but doesn't have much hardware uptake yet). They probably want to ensure that people are aware that it will be matched so that they don't commit to it and AV2 ends up a bit late to the party like AV1 is compared to h265 where it had a notable compatibility lead. If people know that AV2 is coming and competitive they may avoid adopting h266 and wait for the open alternative to ship.

cogman10 1y ago
AV1 competes with h266. They were released near the same date. In many ways h266 has already lost the battle as nobody supports it even though it's been around just as long as AV1. h267 is still in development and due to be released in 2028. That's the actual competitor with AV2.
- kevincox 1y ago
  I guess it depends how you looked at it. Performance-wise AV1 seems more similar to h265. Hardware support wise h266 seems to have only had shipping hardware this year? So I guess neither of these line up 1:1. I tend to see h265 and AV1 competing pretty hard right now so I tend to think of those as one generation and presumably h266 and AV2 will compete as the next generation.
  - cogman10 1y ago
    The way MPEG runs things there are some oddness with the likes of H265. MPEG likes to add little extensions to a codec over time. That's why H265 was first released in 2013 but ultimately had like 10 different extensions after that fact (the last one in 2024). AV1 and VP9/VP8 before that have, in contrast, been pretty much static after they were released. AV1 has had a single errata after it's release. So I could see why you'd see H265 as the competitor. I mostly don't simply because I believe they explicitly stated that they were trying to be competitive with 266. I personally prefer the way AOMedia is running things and I suspect hardware manufacturers do as well. No licenses and AOM is creating open source reference encoders/decoders. They are working very hard to make it easy for manufacturers to be able to pick up the spec and run with it. Keeping the stream standard static for a long period also means manufacturers don't have to worry that they won't get a new extension next year. Content encoders are also reasonably guaranteed that their encoding with today's software still works with yesteryear hardware. MPEG, on the other hand, is paywalling the crap out of everything.

zamadatix 1y ago

Media companies want to save on bandwidth, content types keep changing (e.g. HDR), people want live calls to work whenever, and so on as well. At the same time, GIF still isn't completely gone either just because people needed more than what GIF could give them.

ksec 1y ago

That is what I have been saying for close to 10 years now. We are quite close to Patents free H.264 High Profile. You will always have to provide H.264 as baseline due to compatibility. And that has a cost of storage. Bandwidth cost have been declining fast, and with AI Capex it doesn't seems to be slowing down either. Meanwhile Storage cost hasn't drop, if anything recent trend suggest it may have plateau and went up for HDD. H.264 1080P 2Mbps is good enough for a lot of things. Just like how MPEG-2 is still getting encoder improvement ~30 years later so is H.264 encoder. There are other codec like LCEVC which you can apply on top of H.264 can provide up to 60% bitrate reduction for 4K content. This saves on storage cost and provide enough benefits. It is only in streaming services like Netflix where the catalog of video are low enough they could afford to re-encode it every 5- 8 months and storage cost is minimal. Again a new codec introduction is easily a 10 years task. Higher Speed PON is already being tested, while others are working on NGS-PON2 roll out. 5G Home Broadband with Massive MIMO. The true free and open Video Codec may not be AV1 or AV2, but H.264.