35 comments

[ 2.5 ms ] story [ 52.9 ms ] thread
Project description:

  dav2d is the fastest AV2 decoder on all platforms :)
  Targeted to be small, portable and very fast.
If you're out of the loop like me:

  AV2 is the next-generation video coding specification from the Alliance for Open Media (AOMedia). Building on the foundation of AV1, AV2 is engineered to provide superior compression efficiency, enabling high-quality video delivery at significantly lower bitrates. It is optimized for the evolving demands of streaming, broadcasting, and real-time video conferencing. 
- from https://av2.aomedia.org/
> AV2 is the next-generation video coding specification from the Alliance for Open Media

Oh no. Not another one. I presume this one makes lossy better, or faster or both.

Why do you not want another one? We're approaching a decade since AV1 released. There's plenty of improvement to put into a new format.
> There's plenty of improvement to put into a new format

Like what ? Lossy compression has nothing to offer since about 10 years ago.

Is there a flac for video ? Is there a codec able to "guarantee" DVD quality on today's internet ? Those "up to" really sound like "up your's".

> Like what ? Lossy compression has nothing to offer since about 10 years ago.

h.266 is about 1/3 smaller than AV1, so it fits about 50% more video per megabyte at the same quality. AV2 will almost certainly be better than h.266

> Is there a flac for video ?

ProRes is a close equivalent.

> Is there a codec able to "guarantee" DVD quality on today's internet ? Those "up to" really sound like "up your's".

Basically no. You won't find a video distribution codec that guarantees any level of quality because they're all configurable. And youtube in particular is always tweaking the quality dial as low as they can. But if you can get a bitrate that would be passable with h.264, then AV1 will look good and AV2 will look great at the same bitrate.

And they can all beat a DVD with a fraction of the bitrate. You just have to not get greedy.

is there already an av2 encoder for ffmpeg?
i don't even own a dgpu capable of encoding av1 and they're already coming out with av2?
Not on topic, but wow the internet has very quickly devolved into: click -> "making sure you're not a bot", click -> "making sure you're a human", click -> "COOKIES COOKIES COOKIES", click -> "cloudflare something something"
AI is a gift that keeps on giving.

High hardware prices, locked information sources, plenty of AI slop etc.

Their bot-detection page took more than 40 seconds to complete on my low-end smartphone. This sucks.
[flagged]
We must not continue to develop media codecs in memory unsafe languages. Small, auditable sections can opt-out perhaps, but choosing default-unsafe for this type of software is close to professional negligence.
Are you part of any codec development team to use "we" here?
I would even remove the C code and lower the usage of the assembler pre-processor to a basic C pre-processor.

Happy, AV2 decoding already here.

:)

Glorious. Really looking forward to seeing how much better than AV1 it actually turns out to be. It's a shame it'll take a while before we'll have a decent encoder (it took an annoyingly long time until SVT-AV1 was usable).
off topic, but related to the recent github alternative discussion:

Wow, this gitlab instance looked so much cleaner/simpler and less clunky than my past experiences! Also loaded really fast on first page load as well as subsequent actions

Gitlab has been better than github in almost every aspect for a few years already, I don't understand why anyone still bother with github at that point
is there any understanding of how big of an improvment av2 will be over av1?
About 30% better compression than AV1 at equivalent quality. But it'll be a while before it's a good idea to use AV2 in your home media server. (AV1 is still not that broadly supported)
> About 30% better compression than AV1 at equivalent quality.

Claimed by the authors or actually based on anything more than "trust me bro"?

maybe not great naming. Sounds very similar to the rapper D4vd, who was just arrested for murdering a 14 year old girl
Nice.

What's the current state of of Dolby trying too attack AV1 ecosystem (Snapchat more specifically)? I hope there is an organized fight back by AOM against these trolls.

Mostly ASM for performance critical paths is a pattern that never gets old. The VideoLAN team did the same with dav1d and it paid off. Curious how much of dav2d ends up staying C as AV2 matures.
With the first RVA23 boards shipping this month, I find it a mistake to still focus on legacy ISAs like x86 or ARM rather than what will be dominant by the time AV2 is deployed.
Some dude at videolan be like, I've have a cool name idea
A little extra context:

Dav1d was the surprisingly fast assembly implementation of AV1 decoding. Even for something in hand-coded platform-specific-assembly I think the general impression was that they'd done amazing work to really chase down every last bit of potential performance.

It didn't initially exist when AV1 was first rolled out and its arrival was a step change in powering adoption on devices without hardware decoding.

Dav2d is likely to play a similar role, but it exists from the start of AV2 and can build on the work of dav1d, so should have an even bigger effect.

In a weird reverse chicken and egg scenario, having really good software decode that can be deployed will spur on hardware development and adoption due to network effects.

Where do I get example AV2 videos to test it?
I've noticed that a number of comments seem to have missed the recursive naming pattern:

    Dav1d AV1 Decoder
    Dav2d AV2 Decoder
Just like "GNU's Not Unix"