Format/Codec Wars - AV2 vs. VVC / H.266

  • Want to keep track of this thread?
    Accounts can bookmark posts, watch threads for updates, and jump back to where you stopped reading.
    Create account
I think it might be too early for AV2 given that AV1 is still getting off the ground.

The whole reason 264 has been around so long is because it was the default for so damn long. AV1 is getting close to that level of penetration, being used in YouTube, Netflix, and with devices being built with support for it.

Others mentioned Webp, but Perhaps a better example might be HDMI and USB. ie. A usable standard that became the default, but is now causing headaches because of so many compatible-but-not-really additions. When I go into a shop, it's almost impossible to tell which HDMI cable you should buy. As far as normies are concerned, they're all the same until they get home and suddenly their supermarket brand HDMI cable won't carry the signal of their 4k player or fancy new console.
 
HEVC sucks monkey balls the moment you have to work on a file encoded in it. I'm yet to find one program that can clip my HEVC encoded movie torrents without shitting the bed and not forcing me to re-encode the whole fucking thing in x264 just to clip it.

x264 is the ol' reliable.
Give Avidemux2 a try. It's basically just a GUI wrapper for FFMPEG and it works great for basic things like clipping. If you only cut on I-frames, you can export the truncated stream directly and avoid transcoding entirely.
 
HEVC sucks monkey balls the moment you have to work on a file encoded in it. I'm yet to find one program that can clip my HEVC encoded movie torrents without shitting the bed and not forcing me to re-encode the whole fucking thing in x264 just to clip it.

x264 is the ol' reliable.

Give Avidemux2 a try. It's basically just a GUI wrapper for FFMPEG and it works great for basic things like clipping. If you only cut on I-frames, you can export the truncated stream directly and avoid transcoding entirely.

I would also suggest LosslessCut I haven't had any issues with it so far. Although obviously it has the limitation that you have to cut your clip at keyframes. If I really need to clip between keyframes, what I usually do is make a longer clip with LosslessCut, then cut that further with AviDemux or KdenLive and re-encode it.
 
Can you open WEBP in photoshop now? What about AIF?
despite their historically horrible support for common formats like png, i would have reason to believe that in modern times they simply #include (and pass some linker flags for) the permissively-licensed codec implementations for these formats in a reasonable time frame
of course i don't use photoshop and would never touch it even for research purposes, so i might be wrong

adding to that question you should have asked if it supports jpeg xl or heic (the hevc version of avif, gay patentshit, avoid if possible, nobody uses it at all except apple)
 
I convert most news article images and my text screenshots with GIMP to 75% quality .webp before uploading them here. KF can convert them for you, but *if* the originals are still stored, that's a lot of space savings. It's also a faster upload when file uploads are slow.

I think it might be too early for AV2 given that AV1 is still getting off the ground.

The whole reason 264 has been around so long is because it was the default for so damn long. AV1 is getting close to that level of penetration, being used in YouTube, Netflix, and with devices being built with support for it.
It's never too early to start, since it will probably take at least 2 years to take off (I would be surprised if RDNA5 and RTX 60-series included an AV2 hardware decoder). I don't think H.266 is a big threat if it's only around 10% better on average because the patent/licensing situation has been the albatross around MPEG's neck since H.265. But it's important to keep pushing the state-of-the-art forward as HD/UHD video consumption continues to explode.

Before VP10 was cancelled and rolled into AV1, Google's plan was to have an 18-month gap between codec releases. That turned out to be hopelessly unrealistic, but shows their ambition to quote, "get codec development to Web speed."

Like with AV1, the early adopters of AV2 will be the big AOMedia members like Google and Netflix. They will get their hands on (or design themselves) expensive ASICs to encode the video quickly, and have it ready to serve to users as the hardware decoders percolate. A 30% average reduction in bitrate required per AV2-ready user saves them money. It will take until the 2030s before plebeians contemplate using AV2 to encode their own content.
 
I convert most news article images and my text screenshots with GIMP to 75% quality .webp before uploading them here. KF can convert them for you, but *if* the originals are still stored, that's a lot of space savings. It's also a faster upload when file uploads are slow.
lossless webp can actually be a lot better than lossy webp when saving images with big flat colors like screenshots of text
this is due to how the discrete whatever the fuck transforms in the lossy mode work, which throw away the regular lines and crisp colors in the screenshots in favor of noisy frequency distributions that end up harder to compress in the end
It will take until the 2030s before plebeians contemplate using AV2 to encode their own content.
software encoding for av1 took a while but it eventually got sort of ok (sort of, it's still quite slow)
maybe av2 will be similarly structured and the software encoder will have more improvements earlier as a result
who knows. they might have even specifically tried to optimize the software encoding speeds. you know how they like the idea of saving money on live streaming video from underpowered client hardware too, as well as how nice it would be if their webrtc implementation made the video streams look really good on still-shitty rural amerifat internet and things of that nature
 
despite their historically horrible support for common formats like png, i would have reason to believe that in modern times they simply #include (and pass some linker flags for) the permissively-licensed codec implementations for these formats in a reasonable time frame
idk when WEBP became a standard but I'll check if Photoshop supports it now... Oh it does, only took a couple of years.
 
idk when WEBP became a standard
it was out in 2011, it was mostly figured out by around 2014 i guess, and truly became stable in 2018 (read: the reference encoder's 1.0 release)
apparently browsers started supporting it en masse by 2020. i would say there is no reason to ever use a png or jpg as your main format for serving images these days, unless you're just doing it for fun or using a png as an intermediate format (they work quite well for that!)
couple of years
:optimistic:

it's actually hilarious how these professional software packages fail to add basic things like file formats for years and years and the zero-budget retards in the free software world support them in 5 minutes after like 4 things happen:
  1. reference implementation comes out and arrives in repositories
  2. it gets hooked into gdk-pixbuf and whatever the qt equivalent of gdk-pixbuf is and the changes arrive in repositories
  3. application authors add a dialog menu for it (longest part of the process)
  4. they make a new minor point release and it arrives in repositories
 
Last edited:
have it ready to serve to users as the hardware decoders percolate.
That's really the issue imo. AV1 is still a niche format when it comes to usability. Desktop computers have the power to software it if needed, but things like TVs and maybe phones? I'm not sure.
 
Desktop computers have the power to software it if needed, but things like TVs and maybe phones? I'm not sure.
idk
tvs and phones these days usually include decently powerful arm socs with at least 4 cores and a decent few gigabytes of ram

and recent codecs seem to be only partially focused on making the bitrate as low as possible. there seems to be quite a heavy focus on making them have low algorithmic complexity, just look at opus and how it somehow manages to be fast as fuck and have a good quality/bitrate ratio at the same time
also take into consideration that compressing things harder in the encoder doesn't really affect decoder time all that much. if you put a png through one of those png crushers for a fucking hour, it will decode exactly as fast as the 4 byte larger png you only put through the crusher for 10 seconds
there are just a shitload of very complicated factors going on here, and we don't know many of them from the handful of vmaf benchmarks they've posted so far

it's possible they are optimizing for decoder efficiency, and you will be able to decode higher resolution streams than you really have any right to be decoding

there are also other techniques, that were not around in 2014 or whenever the fuck. we are at a point where basically every computer you can buy supports vulkan somehow, which means they might be trying to make it work well on compute shaders too (fun fact: ffmpeg has actually been getting a few vulkan compute shader codecs. it's like hardware acceleration but without waiting 9 years for the gpu manufacturers to get off their asses!)
 
Here's an article about AV2 from a guy who is far more critical, from a few weeks ago before most of the specifics were known
first he makes the assumption that av2 will be vastly more complicated than av1:
It is also likely to be more complex than AV1, which means a heavier load for playback.
then he goes on to just completely assume that it will have higher codec complexity:
If AV1 strained low-end Android devices, AV2 will be worse.

in general this article is just a meandering pile of shit that doesn't ever say anything we didn't already know. i can't even tell what he's trying to get at with this article of internet pollution
 
And yet niggers still won't stop using barely compressed h264 instead of even a reasonably efficient codec like HEVC, because in nigger logic bigger file = better quality.
Unfortunately, most normie-tier video playback devices don't support any codec newer than h.264/x264. If you happened to be the designated pirate in your family, Grandma's only way of watching videos is on a TV and she can't work out how to watch her stories via Netflix or Tubi, you're going to end up giving her a whole bunch of USB flash drives with MP4s on them.

Maybe newer devices support HEVC, but it takes years for manufacturers to catch up. I'm pretty sure there were TVs, DVD players etc that didn't support any video format newer than XviD/DivX AVI right up until the mid '10s.
 
Last edited:
Unfortunately, most normie-tier video playback devices don't support any codec newer than h.264/x264.
you'd expect them to support 2-year-old (less ancient) formats given the fact that most of them run android or some crap
unless, of course, it's another one of those proprietary software moments
it's amazing enough that they let you watch things from removable media in the first place
 
it's possible they are optimizing for decoder efficiency, and you will be able to decode higher resolution streams than you really have any right to be decoding
I think they're primarily optimizing for streaming services like Netflix and Youtube. 10% off the filesize/bandwidth is huge for them but not that great for the home consumer considering how long it will take to do an efficient encode.

Maybe newer devices support HEVC, but it takes years for manufacturers to catch up. I'm pretty sure there were TVs, DVD players etc that didn't support any video format newer than XviD/DivX AVI right up until the mid '10s.
I feel like devices like have been better at supporting the codecs and weaker when it comes containers and their features.
 
how long it will take to do an efficient encode.
i said decoder efficiency, as in "your shitbox can decode this video like it decodes an h.264 video"
some very competitive formats like opus and zstandard have really fucking good decoder (and sometimes encoder) efficiency

they might be paying a little bit of attention to encoding performance as well as decoding performance too, we'll see. encoding performance is actually pretty important because normies fucking love streaming facecam videos of themselves with 9 other normies and a really good codec is essential for being able to do this over shitty lte cell towers
 
I haven't checked but it's highly likely most of my video collection is h264. Should I bother encoding it all to h265? I'm not considering AV1 for fears of incompatibility with some devices (possibly unfounded). Another fear, also possibly unfounded, is that I would feel compelled to check every file thoroughly afterwards, just in case the re-encoding fucks the files up somehow. Not sure I wanna commit to that.
 
Back
Top Bottom