Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
[dupe] AV1 Released (aomedia.org)
73 points by FullyFunctional on June 26, 2018 | hide | past | favorite | 18 comments


So, in case people wonder what is happening, wrt to the timeline of the AV1 codec, here is some info.

This is final version of the bitstream and the specification of the codec. It will not change anymore (until AV1.1 or AV2).

There are not much features changes since the "release" that was announced before NAB (early April), but now everything should be clean and checked (checking that the spec actually matches the reference decoder, decoder model, profiles, ...), and a few bugfixes went in already.

Now, in a couple of weeks, you should have MP4 and MKV mappings finalized (yes, that means the files you've seen so far are invalid). And then it will be available to everyone (if you can encode it, of course).


> if you can encode it, of course

And here lies a really big issue I have with modern codec standardisation and development - it will be unfortunately many years to come until we see sufficient optimisations to make at a minimum realtime encode and decode of AV1 feasible. Ten years ago when Moore's law was actually a thing this can-kicking may have been acceptable but now I sincerely believe codec development needs to be more iterative, and to consider hardware optimisations (in particular, having even a HDL variant as part of reference codec) from the initial outset.

VVC I suspect will repeat history in many ways, by virtue of how these codec standards bodies operate.


Direct link to 1.0.0 spec (PDF): https://github.com/AOMediaCodec/av1-spec/releases/download/v...

Unofficial working copy (HTML): https://aomediacodec.github.io/av1-spec/

Source (of spec): https://github.com/AOMediaCodec/av1-spec/tree/v1.0.0

AV1 is a video codec developed by a consortium of companies, broadly descending from Google's VP9 while incorporating significant enhancements from Cisco and Xiph.Org. The participant companies are numerous and include big chipmakers and content delivery networks, banding together to perpetuate a royalty-free video codec that mounts a competition against MPEG's efforts, whose rules don't exclude techniques that are patented.

This is news because previous for previous codecs in this bloodline, like VP8 and VP9, the bitstream has never been frozen [1][2]. This wishiwashiness didn't preclude hardware implementations, and certainly not software implementations, but was very much a departure of the orthodox, old-hat practice in the standard-speccing space. With AV1, they're trying an approach that competes more directly with the MPEG specs, by having a bitstream that's not a moving target and is more formal than just a document embarrassingly marked as draft [2].

[1] http://blog.webmproject.org/2010/06/future-of-vp8-bitstream.... [2] https://www.webmproject.org/vp9/#draft-vp9-bitstream-and-dec...


No. VP8 and VP9 bitstreams are frozen and stayed frozen when they decided to freeze it (they made that sure with automatic tests in libvpx that calculated the bitstream checksum) otherwise old VP8/VP9 files would stop playing. The difference is only with the approach as with VP8 and VP9 the "reference" decoder is the specification, which is not what many people like for various reasons (for example there were some errors in the implementation, which needed to be implemented by everyone because of that).

The first link you posted about VP8 just writes that they'll create an experimental branch (back in 2010) where incompatible bitstream changes can go in. Guess what eventually happened with those changes? VP9

The second link is the link to the unfinished VP9 specification, however with VP9 they still had that "reference" decoder code is the specification approach. As VP9 started to gain traction, the demand for the specification became greater, so they post-hoc started writing one but as you can see, they never finished it. Still this doesn't mean the bitstream isn't frozen.


Since this is just a link to a source tree and I wondered myself what it was, I'll add that AV1 is an open, royalty-free video codec:

https://en.wikipedia.org/wiki/AV1

Opus is the audio equivalent.


AV1 is a collaboration between lots of huge players combining their individual state of the art attempts (Mozilla, Google, Cisco, etc). It's really encouraging, actually.

From what I can tell Opus had development contributed by multiple groups, but significantly less.


We've updated the link from https://aomedia.googlesource.com/aom/+/v1.0.0 to the official page, but we're happy to update it again if someone can suggest better.


Why is it marked as dupe?


I just did that, because it looks like this story had significant discussion recently: https://news.ycombinator.com/item?id=16697415. (And https://news.ycombinator.com/item?id=16796127 also.)

Sorry, I didn't see these comments or I'd have said so sooner!


It's not a dupe. The previous release was a marketing fluff. Now people can actually start encoding video without having to worry about having to do everything over. That is a big deal. The previous announcement was a complete non-event.

Incidentally, Bunny's NeTV2 [1] might be a convenient platform for a hardware accelerated codec.

[1] https://www.crowdsupply.com/alphamax/netv2


Those 2 discussions are about the "release" that was more a beta, than a release. This current discussion is the actual release.


AV1 has lightfield encoding support, which should allow for impressive mobile VR content once hardware decoders are available! Annex D of the spec has more information: https://github.com/AOMediaCodec/av1-spec/blob/master/annex.d...

The general technique was described back in 2000: https://ieeexplore.ieee.org/document/838165/


The title should be: “AV1 Released (for real this time)”

Previous discussion: https://news.ycombinator.com/item?id=16796127


Previous discussion is not the same as this one. This is the actual release.



Which for some reasons isn't marked a dupe.... sigh


How slow is libaom right now?


Any hardware vendors announced timelines on hardware codec support yet?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: