AV2 Is In R&D As The Eventual Successor To The AV1 Video Codec

[phoronix]

Phoronix: AV2 Is In R&D As The Eventual Successor To The AV1 Video Codec

While AV1 adoption is still taking off and finally seeing desktop hardware with AV1 decode, given the time it takes to develop a new high-end video codec it shouldn't come as too much surprise that "AV2" is already being explored...

http://www.phoronix.com/scan.php?page=news_item&px=AV2-Video-Codec-In-Research

[birdie]

By all indications it looks like AV1 will remain reserved entirely for Google and content delivery because the codec is so complex, end users are unlikely to encode their videos into it. Quite a number of patents have expired during the AV1 timeframe, so AV2 might become a lot more efficient in terms of encoding/decoding complexity while not being more efficient in terms of compression ratio. This opinion will not be welcome here but HEVC looks a lot better than AV1: a similar compression efficiency yet a lot easier (up to 20 times) to encode and decode.

The other major issue I have with AV1 is that I expected AV1 to beat AVC/H.264 (first released around 2004, i.e. 16 years ago!) in terms of near-lossless encoding and it mostly hasn't happened. If you want transparent FullHD encodes without any perceivable detail loss, H.264 is still largely unbeatable which is just bonkers. Hopefully AV2 will address this.

Of course new codes like AV1/VVC are a lot more efficient for high-resolution content only it takes so much space most people haven't embraced it.

(Edit: I meant VVC, not HEVC, sorry).

[cl333r]

By all indications it looks like AV1 will remain reserved entirely for Google and content delivery because the codec is so complex, end users are unlikely to encode their videos into it. Quite a number of patents have expired during the AV1 timeframe, so AV2 might become a lot more efficient in terms of encoding/decoding complexity while not being more efficient in terms of compression ratio.

Yeah, it's common sense - last I heard AV1 encoding is too expensive.

This opinion will not be welcome here but HEVC looks a lot better than AV1: a similar compression efficiency yet a lot easier (up to 20 times) to encode and decode.

Yeah, for me for now and the next few years - for sure, because I bought a video card that hw decodes HEVC and I don't plan on upgrading it in the next few years. And, patents aside, HEVC is awesome, I even encoded a tv series to HEVC and shared it [1], sorry if it's not legal to post the link to it.

[1]

http://rutor.info/torrent/758488

[bug77]

By all indications it looks like AV1 will remain reserved entirely for Google and content delivery because the codec is so complex, end users are unlikely to encode their videos into it. Quite a number of patents have expired during the AV1 timeframe, so AV2 might become a lot more efficient in terms of encoding/decoding complexity while not being more efficient in terms of compression ratio. This opinion will not be welcome here but HEVC looks a lot better than AV1: a similar compression efficiency yet a lot easier (up to 20 times) to encode and decode.

The other major issue I have with AV1 is that I expected AV1 to beat AVC/H.264 (first released around 2004, i.e. 16 years ago!) in terms of near-lossless encoding and it mostly hasn't happened. If you want transparent FullHD encodes without any perceivable detail loss, H.264 is still largely unbeatable which is just bonkers. Hopefully AV2 will address this.

Of course new codes like AV1/HEVC are a lot more efficient for high-resolution content only it takes so much space most people haven't embraced it.

I mean seriously, your vacation videos are of so high-quality, H264/265 is weighing them down?
Of course content delivery is the primary client for these.

But yes, overall these are huge, complex beasts, mostly going into intractable problem territory... There will always be downsides.

[Min1123]

I don't blame the complexity of the codec, I blame the GPU manufacturers. Intel, on some of their chips with GPUs, supports VP9 encoding at 8-bit, in a few newer ones VP9 at 10-bit. AMD and NVidia only support the proprietary codecs for encoding, not any of the FOSS codecs.

Google pushes for VP8, VP9, and AV1, because they have to pay licensing costs to encode videos in HEVC and H264. I didn't actually see where this ( https://www.osnews.com/story/22787/m...-license-h264/ ) changed, so I attempt to leave that shit alone, not on technical merits, but on ethical merits.

[jntesteves]

That roadmap page is a description of the phases of deployment of the codec. It does not states in which phase AV1 is right now, but it's easy to imply from the descriptions in the roadmap that it is not in the last phase, but most importantly it does not state that AV2 is in development or R&D right now. I think this news is just bogus.

[ksec]

I don't blame the complexity of the codec, I blame the GPU manufacturers. Intel, on some of their chips with GPUs, supports VP9 encoding at 8-bit, in a few newer ones VP9 at 10-bit. AMD and NVidia only support the proprietary codecs for encoding, not any of the FOSS codecs.

Google pushes for VP8, VP9, and AV1, because they have to pay licensing costs to encode videos in HEVC and H264. I didn't actually see where this ( https://www.osnews.com/story/22787/m...-license-h264/ ) changed, so I attempt to leave that shit alone, not on technical merits, but on ethical merits.

1. Google does not play per Stream, both on HEVC and H.264. Only on Encoder.
2. There is a Cap on H.264 and H.265. So little that sales of Pixel have reach that cap and grant Google the use of both codec in infinite use.
3. Providing Support of Hardware Encoding Cost money to Hardware Vendors. Hence the little FOSS codec support. Not to mention there needs to be IP to licenses from Hardware Design in the first place,

What people dont understand ( And I dont understand why they dont understand ) is that the cost efficiency has to reach a point. If Av1 encoding complexity meant they are paying a sum of money greater than say HEVC or VVC, they will need to work on it because not every is Google that can waste money.

And before any one comes with an argument of saving bandwidth ( I guess that is why they dont understand the cost factor ), bandwidth cost is very little and dropping faster than Hardware Performance / dollar.

And on the last point. If you think paying licensing to work and research done by other companies is unethical than I think we can end the discussions. It is perfectly fine that you are against patents or thinking the cost of licensing is unfair and expensive. It is entirely different matter that it is ethical to use other people work, benefiting from it and not paying them.

And on the AV2. I surely hope they learned their lesson and takes the time to work on it. But judging from the current state of things it doesn't seems anyone there care about efficiency. Continue to throw in Machine Learning or all sort of expensive ( but exciting ) tools.

[MadeUpName]

The AO-media encoder is a dog. But ffmpeg will be dropping their new release shortly that supports a couple of new encoders, I am hoping those will be better as I believe both of them will be multi-threaded as opposed t the AO-media one that is single-threaded. We will know shortly. As for AV1 just being a crap codec image wise there are way to many variables in how you use an encoder for that statement to be valid on it's face. We need to know what settings you are using on both encoders, what bit rates etc. You have to be encoding the exact same file.

If we get a decent encoder my company will move to AV1 for 4K+ for the same reasons the big players are.

[Orphis]

This opinion will not be welcome here but HEVC looks a lot better than AV1: a similar compression efficiency yet a lot easier (up to 20 times) to encode and decode.

You're not comparing the same generation of codecs: H264 is on the same level as VP8, H265 (HEVC) compares to VP9 and H266 to AV1.

Each generation is a lot more complex to encode since you have more tools to choose from. But it doesn't necessarily mean higher quality, it means better quality for the same bitrate.
That is useful for both extremes: 4k and 8k which need to fit in current medium and also super lower quality streams that need to be more reliable (start fast, work on bad connection), aka conferencing.

[Min1123]

And on the last point. If you think paying licensing to work and research done by other companies is unethical than I think we can end the discussions. It is perfectly fine that you are against patents or thinking the cost of licensing is unfair and expensive. It is entirely different matter that it is ethical to use other people work, benefiting from it and not paying them.

Absolutely not, my understanding is that if I film something with a proprietary codec, the content is mine, but I technically owe a licensing fee that I cannot pay to the patent pool, because it only has the actual agreements set up with corporations, and so my video is in a state of limbo by the license not being paid, and by the rights to encode with that codec not being transferable from the device manufacturer to me. I wanted to find a reference to that, but it is such an old thing (dating to original arguments about Theora), that I couldn't find a reference (in a reasonable amount of time) and couldn't cite one.
Hopefully that information is inaccurate or obsolete, but I don't have citations to know so, and I'm not going to spend hours of reading legalese to try to confirm/deny it.