AOmedia next (AV2) codec

[rwill]

Well, I am just happy that they stopped also making a completely new container formats with every major MPEG codec release.

We had Program Streams with Mpeg-1, Transport Streams with Mpeg-2 and that extended Quicktime Container with Mpeg-4.

Since then, nothing new, and thats totally fine I guess.

[Z2697]

Given long enough time average Joe's will know MP4 is not codec.

[Ritsuka]

Even if we had a different file extension for every container + video combination, what about audio and subtitles? It's not possible to do a separate extension for each video/audio/subtitles/whatever combination.

[excellentswordfight]

Originally Posted by kurkosdr

Before HEVC came along, pretty much everyone was following the above de facto rules, and mp4 files were universally playable on any "HD" player as a result. This is why mp4 files with HEVC not playing in "HD" players are difficult to explain to Average Joes and Janes, mp4 had evolved into a de facto standard for H.264 video before that and they expect the filetype to be universally playable.

Joes and Janes dont touch videofiles anymore, they use streaming and are on phones and tablets. People dont burn videofiles to dvds, or transfer them to usb drives to put in playbackdevices anymore to view some home video or whatever. Heck, people barely view their media in file managers at all, its in their icloud, in their instagram/snapchat library etc. They have moved on, its no point in starting doing this now, for their sake.

Originally Posted by Z2697

Given long enough time average Joe's will know MP4 is not codec.

They dont need to know it anymore, and at least in my experience power users have actually started to understand this. There has never been more resources about this stuff, when getting in to video production youtube and the internet are full or resources were this is explained good enough.

[kurkosdr]

Originally Posted by benwaggoner

I expect we'll eventually see some pretty darn good 2160p HDR 24p encodes in the 2-4 Mbps average bitrate range. AFGS1 means that we'll see the biggest savings in the content that need the highest bitrates today.

Personally, I am eager to see how 2160p HDR 24p video at 4Mbps average bitrate will look like in any format (for example AV2 or ECM), outside of specialized cases such as videos with very low motion throughout their duration. Let's assume there is no film grain either, just lots of bonafide detail that a 2160p video shot with a decent camera ought to have. I would very much like to see those claims of 4Mbps average bitrate.

At some point, you hit the limits of entropy. AV1 and VVC already lean heavily on post-processing filters to hide the damage caused by compression (leading to tons of blur that removes detail and also a "melting" effect in motion scenes), let's see how AV2 or ECM improve on that (or "improve" by leaning even more heavily on post-processing and creating an even more unnatural image).

[excellentswordfight]

Originally Posted by kurkosdr

Personally, I am eager to see how 2160p HDR 24p video at 4Mbps average bitrate will look like in any format (for example AV2 or ECM), outside of specialized cases such as videos with very low motion throughout their duration. Let's assume there is no film grain either, just lots of bonafide detail that a 2160p video shot with a decent camera ought to have. I would very much like to see those claims of 4Mbps average bitrate.

Well, for that scenario, even x265 can look decent at 4Mbps. Modern clean digital sources compresses really really well.

https://limewire.com/d/RFTwE#7AXmHV0j1p

But even so not sure if it makes sense as it will probably look better at 1080p at that compression. But new codecs have also been less sensitive to bits/pixel, at some point maybe we do the whole abr at the same res.

[Z2697]

So called "post-processing" filters are actually very important.
Think the deblocking filter, since it's introduction in AVC it has become THE standard feature.
In reality most of which is implemented as in-loop filter, so that they actually contain some information of how the source should look like.

Blocking is basically an impossible (or too expensive) problem to solve without the in-loop deblocking filter.
With large blocks the next eye irritating artifact is "DCT ringing".

[kurkosdr]

Originally Posted by Z2697

So called "post-processing" filters are actually very important.
Think the deblocking filter, since it's introduction in AVC it has become THE standard feature.
In reality most of which is implemented as in-loop filter, so that they actually contain some information of how the source should look like.

Blocking is basically an impossible (or too expensive) problem to solve without the in-loop deblocking filter.
With large blocks the next eye irritating artifact is "DCT ringing".

Yes, deblocking signalling (and post-processing signalling in general) is a good thing, but my point is, we've reached diminishing returns with HEVC and VP9. Any further post-processing introduced since then is basically a tool to game the metrics, it results in a more pleasing picture but at the expense of detail, which means the bitrate savings are fraudulent (just use a lower resolution if you aren't gonna keep the detail). That's why I am eager to see those claims of 2160p24 HDR at 4Mbps or even 6Mbps with AV2 or ECM being tested. Maybe on some types of content, but not on content with actual temporal or spatial (non-film-grain) detail.

[benwaggoner]

Originally Posted by kurkosdr

Personally, I am eager to see how 2160p HDR 24p video at 4Mbps average bitrate will look like in any format (for example AV2 or ECM), outside of specialized cases such as videos with very low motion throughout their duration. Let's assume there is no film grain either, just lots of bonafide detail that a 2160p video shot with a decent camera ought to have. I would very much like to see those claims of 4Mbps average bitrate.

There are plenty of movies that look good in a 4 Mbps ABR with HEVC 2160p24. Peaks will be higher, but without grain or a lot of very complex detail, it's pretty achievable. Motion blur can hide many sins, and many bits can be moved out of end credits elsewhere. Not using x265 default parameters or anything, but with proper tuning it can be done, especially with CGI content with literally no kind of grain or noise.

At some point, you hit the limits of entropy. AV1 and VVC already lean heavily on post-processing filters to hide the damage caused by compression (leading to tons of blur that removes detail and also a "melting" effect in motion scenes), let's see how AV2 or ECM improve on that (or "improve" by leaning even more heavily on post-processing and creating an even more unnatural image).[/QUOTE]

[excellentswordfight]

Originally Posted by kurkosdr

That's why I am eager to see those claims of 2160p24 HDR at 4Mbps or even 6Mbps with AV2 or ECM being tested. Maybe on some types of content, but not on content with actual temporal or spatial (non-film-grain) detail.

See above, again not AV2, but I dont see the claim to be unreasonable given whats achievable with HEVC.

Originally Posted by excellentswordfight

Well, for that scenario, even x265 can look decent at 4Mbps. Modern clean digital sources compresses really really well.

https://limewire.com/d/RFTwE#7AXmHV0j1p

STEM2 is a very good soruce for this case, although it seems like this have shifted quite a bit in recent years were its more common for streaming first titles to have grain, its a title that image wise is very close to what we used to see from a lot of netflix originals.

[hajj_3]

libavif v1.4 is out: https://github.com/AOMediaCodec/liba...ses/tag/v1.4.0