AV1 Released
aomedia.org
aomedia.org
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).
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.
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...
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.
https://en.wikipedia.org/wiki/AV1
Opus is the audio equivalent.
From what I can tell Opus had development contributed by multiple groups, but significantly less.
Sorry, I didn't see these comments or I'd have said so sooner!
Incidentally, Bunny's NeTV2 [1] might be a convenient platform for a hardware accelerated codec.
The general technique was described back in 2000: https://ieeexplore.ieee.org/document/838165/
Previous discussion: https://news.ycombinator.com/item?id=16796127