Xiph & Mozilla's Daala codec

[Jamaika]

All beautifully. So far codec is a sham. Message is displayed:
None of the functions don't work.
av_interleaved_write_frame(): Broken pipe
Error writing trailer of pipe:: Broken pipe
It took some time to understand that these patterns were caused by successive accumulation of "noise" coming from the interaction between the lapping filters and the rounding that was taking place when the 12-bit "intermediate" images were being rounded down to 8 bits for use as reference in future frames. Trying to change how the rounding was done or how the lapping was done went nowhere, originally causing some concerns over the viability of lapping itself. Fortunately, two solutions eventually emerged. One was to simply store all reference frames in 12 bits rather than 8. This requires more memory and more memory bandwidth, but also improves coding efficiency for all sequences (even those that did not trigger the problem). It's something that's also needed anyway for 10-bit and 12-bit input support.
About importing files 10/12 bit we can forget.

[Nintendo Maniac 64]

About importing files 10/12 bit we can forget.
10bit or more is necessary for good HDR, otherwise you'll have visible color banding.

[Kurtnoise]

frankly, what did you smoke ?

[wiak]

AV1, VP10, and Daala are video-only codecs, but I have a glimmer of hope that somebody will do like my fellow Fabrice Bellard did with HEVC and BPG.
Or maybe the Alliance for OpenMedia will also propose a new still image codec, with Microsoft, Mozilla and Google on board, it could be doable to initiate a change. JPEG seems unbeatable though.



Yeah Github was being a **** yesterday because the files are too heavy when I update :/
well heard of WebP? that was basicly a VP8 frame, pretty sure someone at aomedia will do the same to AV1 and call it WebP v2, much like they did with WebM and added VP8, VP9

i expect the webm v3 spec to include AV1 with Opus

WebM v1 = VP8 + Vorbis
WebM v2 = VP9 + Opus
WebM v3 = AV1 + Opus

Here's a new Daala demo (https://people.xiph.org/~jm/daala/revisiting/). This one is actually revisiting all of our previous technology demos to see what worked, what didn't and how things changed.
thanks btw would been cool if there was AV1 demo pages too :)

[benwaggoner]

VP8 isn't competitive against HEVC for still images. The power of intra-frame prediction is huge, as are the variety and flexibility of block sizes. Plus the ability to losslessly encode individual TUs for graphics or text.


Sent from my iPad using Tapatalk

[Nintendo Maniac 64]

VP8 isn't competitive against HEVC for still images.

...no dip? Who was arguing that?

[mzso]

By the way, how far past jpeg are we with HEVC? Too bad the IETF doesn't push for a new image format too like with video. If it became widespread on the internet it'd soon seep down to digital photography too.

JPEG is surpassed technology, yet it's pretty much the only lossy image format worth mentioning.

[Jamaika]

JPEG is surpassed technology, yet it's pretty much the only lossy image format worth mentioning.
Off the topic at Daala, but ...

There was a lot of advertising for a five years to replace JPEG. He had to enter photos WEBP with codec VP9/VP10. He was BPG (HEVC), which is already very old and not updated. Advertising codecs FLIF.
The truth is that hardware manufacturers don't want to replace JPEG. JPEG evolved. He became a container 8/10/12/16 depth bit. It is with compression mode progress, which took over the technology of Adobe DNG. There isn't colormatrix bt2020 and has an old packer libzip 1.2.7. (options not supported by ffmpeg)

The program which implemented a partial identification of BPG image is MediaInfo, it is not too much.
VP8 isn't competitive against HEVC for still images. The power of intra-frame prediction is huge, as are the variety and flexibility of block sizes. Plus the ability to losslessly encode individual TUs for graphics or text.
There are no miracles. They strive for perfection.

Comparison of codecs: bitrate=3000kbps
Warning. Bitrate is false, applies to all one hundred frames and not the individual.
Original:
Daala codec v0.0-1598-g7290550
daalaenc.exe --b-frames 4 --keyframe-rate 50.000 --complexity 5 --video-rate-target 3000k --soft-target --output "image.ogg" - {functions disabled: bit-depth=10+}
Selur codec vpxenc v1.5.0-936-g6f397b8
vpxenc.exe --bit-depth=8 --input-bit-depth=8 --threads=4 --i420 --profile=0 --best --codec=vp9 --fps=50000/1000 --cpu-used=0 --passes=1 --pass=1 --drop-frame=0 --disable-kf --target-bitrate=3000
--end-usage=vbr {recommended: --end-usage=cbr --gf-cbr-boost=200} --aq-mode=1 --output="image.webm" -
Komisar codec x264 r2705kMod
x264.exe --demuxer y4m --preset veryslow -tune stillimage --input-depth 8 --input-res 1920x1080 --input-csp i420 --output-depth 8 --fps 50000/1000 --keyint 250 --bitrate 3000 --output "image.h264" -
El Heggunte GCC 5.3.0 v2.0+2
x265.exe --y4m --no-info --preset veryslow --no-open-gop --input-depth 8 --input-res 1920x1080 --input-csp i420 --output-depth 8 --fps 50000/1000 --keyint 250 --bitrate 3000 --output "image.h265" -
nwgat.ninja (https://awesome.nwgat.ninja/aomedia) aomenc.exe GCC 5.3.0 v0.1-daf841b (12.07.2016)
aomenc.exe --threads=4 --i420 --profile=0 --best --codec=av1 --fps=50000/1000 --cpu-used=0 --passes=1 --pass=1 --drop-frame=0 --disable-kf --target-bitrate=3000
--end-usage=vbr {recommended: --end-usage=cbr --gf-cbr-boost=300} --aq-mode=1 --output="image.webm" - {functions disabled: pass=2, bit-depth=10+}
Themselves evaluate codecs. :)

[BadFrame]

By the way, how far past jpeg are we with HEVC?

Wyohknott at github offers a wide range of comparisons, AOM AV1, BGP, Daala, Webp, FLIF, VP10 ... and of course jpeg

Here's a comparison between OAM AV1 and HEVC:

http://wyohknott.github.io/image-formats-comparison/#mascot&aom=m&bpg=m

There are lots of images to compare with as well, just choose your combinations from the drop down lists.

[mandarinka]

well heard of WebP? that was basicly a VP8 frame, pretty sure someone at aomedia will do the same to AV1 and call it WebP v2, much like they did with WebM and added VP8, VP9

VP9 never made it into webp, which is a shame. Their excuse for not doing it was "we don't want fragmentation" which was IMHO pretty bad idea/mistake, because webp still has very little support (and where it is, it is software/browser based, so what are we even talking about?) and hence you can still update it like that.

[Quikee]

The truth is that hardware manufacturers don't want to replace JPEG. JPEG evolved. He became a container 8/10/12/16 depth bit. It is with compression mode progress, which took over the technology of Adobe DNG. There isn't colormatrix bt2020 and has an old packer libzip 1.2.7. (options not supported by ffmpeg)

JPEG wasn't replaced because it has a good enough compression ratio, has low complexity and its support is practically 100%. Also in all the years there was no viable alternative available - by that I mean a format that has a much better 50%+ compression ratio, is free (no patent issues), standarized, has reasonable complexity and has support in majority of browsers.

JPEG/JFIF - the standard format that has largest support didn't evolve much. Yes, it become faster to encode/decode, it gained quality/size because of more clever compression techniques that don't require to break the bitstream. However it didn't gain support for 10+ bit or anything such. Even the support for arithmetic coding (which should be part of the standard) instead of huffman can be problematic so you don't want to use it (and nobody does) for compatibility reasons.

[Jamaika]

However it didn't gain support for 10+ bit or anything such. Even the support for arithmetic coding (which should be part of the standard) instead of huffman can be problematic so you don't want to use it (and nobody does) for compatibility reasons.
Can see more and more converters JPEG 10+ bit on standard 9.2 (https://libraries.io/nuget/libjpeg/9.2.0.1).
12-bit JPEG codec on CUDA (http://www.fastcompression.com/products/jpeg/12-bit-jpeg-codec.htm)

[foxyshadis]

Can see more and more converters JPEG 10+ bit on standard 9.2 (https://libraries.io/nuget/libjpeg/9.2.0.1).
12-bit JPEG codec on CUDA (http://www.fastcompression.com/products/jpeg/12-bit-jpeg-codec.htm)

libjpeg isn't the JPEG standard, and in fact libjpeg7+ have been shunned because they break JPEG compatibility for no good reason, and the guy's frankly a little nutty. The recent 10-bit support is actually a good reason, but again, broken compatibility -- everyone writing and using decoders is too conservative to care about anything but the same JPEGs everyone's been using for over 20 years, and so no minor variant of it will ever matter.

DejaVu might've been successful years ago if it wasn't a giant pile of patents. Otherwise, storage and network speeds have improved too much, and the benefits of deep color too slim, for the masses to ever demand an update.

[Quikee]

libjpeg isn't the JPEG standard, and in fact libjpeg7+ have been shunned because they break JPEG compatibility for no good reason, and the guy's frankly a little nutty. The recent 10-bit support is actually a good reason, but again, broken compatibility -- everyone writing and using decoders is too conservative to care about anything but the same JPEGs everyone's been using for over 20 years, and so no minor variant of it will ever matter.

Yes, the mostly used open source library is libjpeg-turbo which is generally an fork of an old version of libjpeg with performance improvements added. libjpeg-turbo stopped following changes from libjpeg when they started with adding non-standard stuff and breaking ABI for no reason, so very few projects actually use libjpeg library today.

There is also JPEG-XT which is relatively new. The point of this format is to be compatible with JPEG and can be decompressed as a standard JPEG file. Additionally to that it would have extensions embedded in the container like high bit depth and lossless support which a JPEG-XT supported decoder could use. Nice idea but I don't think it will catch on..

[Jamaika]

libjpeg isn't the JPEG standard, and in fact libjpeg7+ have been shunned because they break JPEG compatibility for no good reason, ...
You're right, but recently a lot of advertising on 10bit MOV MJPEG. It is fiction, but I let myself be fooled.

http://www.acrovid.com/footagestudio.htm
http://www.cinemartin.com/next/

10bit lossy is Cineform, but it isn't MJPG.
How do I find a compact camera with 10bit JPEG images, I add a link.
http://www.shikino.co.jp/eng/products/JPEG-IP%20Leaflet_Ver1.01E_HP.pdf
http://www.shikino.co.jp/eng/products/
http://www.shikino.co.jp/eng/products/ip-1.html
http://www.sansmirror.com/cameras/a-note-about-camera-reviews/panasonic-camera-reviews/panasonic-gx7-review.html

[BadFrame]

There is also JPEG-XT which is relatively new. The point of this format is to be compatible with JPEG and can be decompressed as a standard JPEG file. Additionally to that it would have extensions embedded in the container like high bit depth and lossless support which a JPEG-XT supported decoder could use. Nice idea but I don't think it will catch on..

I'd say not a chance, given that a lot of the standard is patented and subject to royalties.

Like you said, for a new format to come and replace jpeg, it needs to be royalty free (ain't no way the 'web' is going to start paying royalties for serving images), and a worthwhile improvement in compression, the ~50% you mentioned sounds reasonable.

I also think that widely supported hardware accelerated decoding would be needed, since the new format would most likely be more expensive cpu-wise.

I believe the only realistic option would be if AV1 was also offered as a image codec, Intel, AMD, ARM, NVidia are all aboard to support the AV1 codec, and Android SoC manufacturers like Qualcomm, Samsung will of course support it, and it's royalty free which means it can gain traction on the web and in apps (as in being supported).

As for the existing offerings, WebP was never good enough in lossy mode (lossless is great though), BGP is based on HEVC and thus patented/royalties = dead in the water, FLIF and Daala are promising but I can't see them ever gain hardware accelerated support.

[Quikee]

You're right, but recently a lot of advertising on 10bit MOV MJPEG. It is fiction, but I let myself be fooled.

It's not that it can't be done and looks like some try to add support for 10bit, but doing this is like adding 10bit support to MPEG2 - it can be done but you can't use any existing MPEG2 equipment to play it back - so what's the point of doing it.


10bit lossy is Cineform, but it isn't MJPG.
How do I find a compact camera with 10bit JPEG images, I add a link.

They use JPEG (extended) for 12bit which I think is JPEG XT (XT - extended) that I described in my previous post. So again - you will need a special decoder for such images to get all 12bits back, with a standard JPEG decoder you will get only 8bits.

They also talk about JPEG XR which is a completely different codec.

[Quikee]

I'd say not a chance, given that a lot of the standard is patented and subject to royalties.

I didn't look into royalties as I assumed JPEG (the committee) learned its lessons regarding patents.


I also think that widely supported hardware accelerated decoding would be needed, since the new format would most likely be more expensive cpu-wise.

Sure, that would be a bonus but I don't think this is required for success of a image codec. I think hardware manufacturers would fill this gap regardless if they are on-board from the start or not.


I believe the only realistic option would be if AV1 was also offered as a image codec, Intel, AMD, ARM, NVidia are all aboard to support the AV1 codec, and Android SoC manufacturers like Qualcomm, Samsung will of course support it, and it's royalty free which means it can gain traction on the web and in apps (as in being supported).

Yes, a new image codec from AOM would probably be successful but it doesn't need to be based on AV1. It could be based on Daala. A image codec is not priority now for AOM but I think they will looking into it when AV1 is launched.

[Jamaika]

So again - you will need a special decoder for such images to get all 12bits back, with a standard JPEG decoder you will get only 8bits.
It's just me not surprised. It is worse with the comfort of work. :(
It remains to use the DNG.
https://photographylife.com/dng-vs-raw

[BadFrame]

I didn't look into royalties as I assumed JPEG (the committee) learned its lessons regarding patents.

One would think so with the history of jpeg and unused potential due to patents, however that is not the case.

From what I've read, not even XR baseline profile is guaranteed to be free from royalties (!), it's clear to me that the jpeg committee has lost any relevance, attempting to introduce a patent-burdened image codec today when even video is going towards royalty free is just insane.


Sure, that would be a bonus but I don't think this is required for success of a image codec. I think hardware manufacturers would fill this gap regardless if they are on-board from the start or not.
Depending on your definition of success, I think I disagree, as my definition would be replacing jpeg as the de facto (lossy) image format. Given how mobile is increasingly becoming the way people consume web content, I think the chance of gaining the necessary traction without hardware decoding is very unlikely.


Yes, a new image codec from AOM would probably be successful but it doesn't need to be based on AV1. It could be based on Daala.

Sure, Mozilla/Daala are part of the AOM so it's certainly not impossible, my doubt again comes back to hardware support, AV1 will be hardware supported, which most likely means it should be easier to implement a hardware accelerated image codec based upon it.

A image codec is not priority now for AOM but I think they will looking into it when AV1 is launched.

Here's hoping.

Of course, AV1 itself also needs to survive what I expect to be a massive patent aggression from the MPEG LA group, they won't give up their patent cash cow business model of : 'pick a bunch of patents from the pool, implement a slight improvement on the current video codec standard, charge royalties, rinse and repeat' : without a fight.

That said, with the companies behind AOM, I think we have the best chance ever of seeing royalty laden video codecs be history.

[mandarinka]

they won't give up their patent cash cow business model of : 'pick a bunch of patents from the pool, implement a slight improvement on the current video codec standard, charge royalties, rinse and repeat' : without a fight

It is fine to be a fan of royalty free codecs, but I think you give the MPEG technology a little bit too small credit.
HEVC isn't just a small slight improvement. And are we forgetting already that VP9 is basically just a copy of HEVC with some nerfs (worse reference structure, weird hacky b-frames, lack of weighted prediction or SAO, IIRC)? On2 still did it the usual way: take ideas from the MPEG standard, and obfuscate them a bit to not be directly exposed to patent lawyers.

And currently, VP9 is the format that AV1 is being built on! So I'd say we should get a bit more mature on the freetard/fanboy feelings toward MPEG. Not to mention when the best video coding technology to this day (x264) was based on their codec.


Also, MPEG-LA royalties are sane and okay. It is the greedy companies that split off from MPEG-LA pool that went overboard and caused a SNAFU.

[dapperdan]

... So I'd say we should get a bit more mature on the freetard ...

Yes, maturity is definitely required here.



Also, MPEG-LA royalties are sane and okay. It is the greedy companies that split off from MPEG-LA pool that went overboard and caused a SNAFU.

MPEG is not the same as MPEG-LA (despite the confusing name), but it is MPEG policies (e.g. being patent-blind) that create the opportunity for both MPEG-LA and competing entities to stuff the proposals with their patents without any credible legal requirement to give access to the patents under sensible terms.

This has continually caused problems, e.g. the big drama when Apple held Quicktime 6 hostage until they dropped per-use royalties on AAC all the way up to the current HEVC nonsense (which again caused Apple remove any mention of HEVC from their iPhone page). You know it's gone too far when Apple, who generally have no big issue with patent encumbered formats, is calling you out.

[BadFrame]

HEVC isn't just a small slight improvement.
Actually, considering how old h264 is and the amount of extra cpu time HEVC needs in order to actually see those benefits, I can't say I'm overly impressed with the improvements. That said this will most likely hold just as true for AV1.

And are we forgetting already that VP9 is basically just a copy of HEVC with some nerfs

Need something to back this up with.


On2 still did it the usual way: take ideas from the MPEG standard, and obfuscate them a bit to not be directly exposed to patent lawyers.

Need something to back this up with.


And currently, VP9 is the format that AV1 is being built on!

From what I've read, it's built on VP10, which was supposed to be the VP9 successor, and is now to become AV1.

So I'd say we should get a bit more mature on the freetard/fanboy feelings toward MPEG.

Are you seriously pretending to be mature while using words like 'freetard' ? And then accusing others of being 'fanboys' while sounding like a MPEG LA advertising pamphlet...


Not to mention when the best video coding technology to this day (x264) was based on their codec.

No argument that h264 is a great codec, but with the vast pool of patents at the MPEG LA disposal, and the hinderance they pose for alternate codecs to appear which in turn could have been better had they not have to work around what is often ridiculously broad patents, it's not surprising.


Also, MPEG-LA royalties are sane and okay.

That's your opinion.

[Quikee]

Depending on your definition of success, I think I disagree, as my definition would be replacing jpeg as the de facto (lossy) image format. Given how mobile is increasingly becoming the way people consume web content, I think the chance of gaining the necessary traction without hardware decoding is very unlikely.

I would agree if we would talk about a video codec but for image codec I can't say that hardware support is the deciding factor. Of course it is important that the codec is designed with hardware implementation in mind as it can't be successful otherwise but I don't think it would be the feature that is important initially.

Consider that we don't need to decompress images at constant rate of at least 21 FPS so even most mobile should be able to handle it reasonably fast and if you factor the download time I think it wouldn't in most cases cause a considerable lag at page loading.

[Quikee]

It is fine to be a fan of royalty free codecs, but I think you give the MPEG technology a little bit too small credit.
HEVC isn't just a small slight improvement.

I agree MPEG did a good job with HEVC, however they were totally agnostic regarding patents and I think they need some "beating" for this mistake and I think AOM with AV1 will do this. This will force them to require some patent commitments from participants or their technology won't be adapted in the standard. Maybe we would also see a standard that is partially or completely royalty free.

And are we forgetting already that VP9 is basically just a copy of HEVC with some nerfs (worse reference structure, weird hacky b-frames, lack of weighted prediction or SAO, IIRC)? On2 still did it the usual way: take ideas from the MPEG standard, and obfuscate them a bit to not be directly exposed to patent lawyers.

I'm not an expert and don't have a much knowledge of video coding techniques but at least I find VP9 "hacky b-frames" quite ingenious solution. It generally makes the decoder agnostic of what kind of frame (past or future) is currently decoding as this is not part of the bitstream. A future frame is only flaged for not showing and I think this makes it a more flexible concept than b-frames. You could also have a frame that is neither past nor the future by maybe a hybrid of more frames - however I don't know if this would be beneficial.

libvpx doesn't exploit all the possibilities that the format has to offer and I think it would come much closer to the quality of HEVC if an alternative implementation emerges and starts exploiting such things (my hope is Eve will better show what VP9 is capable of).

[mzso]

I believe the only realistic option would be if AV1 was also offered as a image codec, Intel, AMD, ARM, NVidia are all aboard to support the AV1 codec, and Android SoC manufacturers like Qualcomm, Samsung will of course support it, and it's royalty free which means it can gain traction on the web and in apps (as in being supported).

Hopefully people in the AOM come to the same conclusion, and they'll pester IETF into standardizing a still image format too.

[mzso]

I agree MPEG did a good job with HEVC, however they were totally agnostic regarding patents and I think they need some "beating" for this mistake and I think AOM with AV1 will do this.

It's countries like the US that truly need a good beating, because they allow software patents. There wouldn't be any need for this pathetic comedy with "free" formats and patented formats. They could just take HEVC and improve upon it...

[mandarinka]

IMHO the only problem with HEVC is that payments got too high because of HEVC Advance and then Technicolor or which company it was that went rogue as the third party. The mistake was not having/enforcing binding agreements that would guarantee sane royalties/terms of use.

Otherwise, I don't see a problem with codec not being completely beer-free if it means better technology/compression (and indirectly, incentive for further development). H.264 showed it works well, although some people think it is some catastrophe or something. :rolleyes:

[BadFrame]

I
Otherwise, I don't see a problem with codec not being completely beer-free if it means better technology/compression (and indirectly, incentive for further development).

I disagree with the notion that software patents and the subsequent royalty fees are somehow the main driving force for better technology, if this was the case we would never have seen great audio codecs like Opus and FLAC, and had it not been for having to navigate around the broad video encoding patent minefield, efforts like Daala would have had an entirely different outlook.

Instead I'd argue that in reality software patents obstructs and slows down technology, and I think the MPEG LA pool is a perfect example of that, they dictate when and how much progress in video codecs will be made available by holding a vast array of encoding techniques locked under what is again aggressively broad software patents.

IMHO we'd be further along in video encoding if we did not have software patents, as we can see with AOM's effort with AV1 and what Google has done all along with the VP series, there is huge commercial incentive beyond cashing in on royalties to improve technology, and yet again even this effort is hampered by software patents, and will likely be attacked by software patents as MPEG LA fights to keep their business model of piecemealing video encoding progress.

I
H.264 showed it works well, although some people think it is some catastrophe or something. :rolleyes:

I don't think it's a catastrophe, but I also don't think the system works 'well'.

Now, I (and I assume you) want to see the best video compression technology possible, in my/your hands as soon as possible, MPEG LA's business model of controlling video compression progress through software patents means we will get what they deem is enough of a step forward to create enough demand for them to kickstart a new codec and royalties cycle, rinse and repeat, continously artificially limiting progress in order for them to milk as much royalties as possible.

Take AOM as a comparison, since their model is not based upon royalties at all, there's no reason whatsoever for them to artificially limit the capacity of their codec, in fact the better the codec is, the more money they save on bandwidth, and the more attractive their services are to customers, but here again the sad concept of software patents rear it's ugly head, as it will limit what AV1 can provide, just as it limits video compression progress as a whole.

[mandarinka]

I disagree with the notion that software patents and the subsequent royalty fees are somehow the main driving force for better technology, if this was the case we would never have seen great audio codecs like Opus and FLAC, and had it not been for having to navigate around the broad video encoding patent minefield, efforts like Daala would have had an entirely different outlook.

Instead I'd argue that in reality software patents obstructs and slows down technology, and I think the MPEG LA pool is a perfect example of that, they dictate when and how much progress in video codecs will be made available by holding a vast array of encoding techniques locked under what is again aggressively broad software patents.

IMHO we'd be further along in video encoding if we did not have software patents, as we can see with AOM's effort with AV1 and what Google has done all along with the VP series, there is huge commercial incentive beyond cashing in on royalties to improve technology, and yet again even this effort is hampered by software patents, and will likely be attacked by software patents as MPEG LA fights to keep their business model of piecemealing video encoding progress.



I don't think it's a catastrophe, but I also don't think the system works 'well'.

Now, I (and I assume you) want to see the best video compression technology possible, in my/your hands as soon as possible, MPEG LA's business model of controlling video compression progress through software patents means we will get what they deem is enough of a step forward to create enough demand for them to kickstart a new codec and royalties cycle, rinse and repeat, continously artificially limiting progress in order for them to milk as much royalties as possible.

Take AOM as a comparison, since their model is not based upon royalties at all, there's no reason whatsoever for them to artificially limit the capacity of their codec, in fact the better the codec is, the more money they save on bandwidth, and the more attractive their services are to customers, but here again the sad concept of software patents rear it's ugly head, as it will limit what AV1 can provide, just as it limits video compression progress as a whole.

You insist that the non-royalty-free codecs limit the pace of overall development of compression. I don't think that is a clearly demonstrable thing at all. Note that while the technology they (MPEG etc) develop is patented, that doesn't mean it is exclusive and other development projects can't use it. If you look into the past, that was not the case - Real (RV40), Sorenson (SVQ3 IIRC?) licensed H.264 IP for their own incompatible codecs. They were allowed to do exactly what you think the baddies make impossible - make their own thing on top of MPEG IP. So as long as you accept to pay, you are not really left out and barred from using the stuff to do your own innovation on top of it.

I would much rather see efforts like AV1/Daala to happen this way, together with the commercial development, so that the best of all camps can be combined. I'd say that it is those projects limiting themselves from use of patented technology, rather then the patented technology makers limiting them, strictly speaking.

Of course, it is very unlikely that Xiph/Mozilla/On2 will change their policy and license patented stuff for themselves (well, Google did that, for VP8/9, actually), given how their whole goal is to be royalty free. But it is their choice, not result of somebody harassing them into it and denying them the option to do otherwise.

[Jamaika]

I disagree with the notion that software patents and the subsequent royalty fees are somehow the main driving force for better technology, if this was the case we would never have seen great audio codecs like Opus and FLAC, and had it not been for having to navigate around the broad video encoding patent minefield, efforts like Daala would have had an entirely different outlook.
...but there are some suggestions.
Video streaming services and video creation companies pay licensing fees for not only the hours and hours of movies and TV shows that viewers watch but also the basic coding technology that enables files to be created and displayed. The cost to license this technology has become a major impediment to innovation for companies trying to make content easily available to millions (billions!) of consumers, which is why Adobe is proud to join the Alliance for Open Media.

Of course, it is very unlikely that Xiph/Mozilla/On2 will change their policy and license patented stuff for themselves (well, Google did that, for VP8/9, actually), given how their whole goal is to be royalty free. But it is their choice, not result of somebody harassing them into it and denying them the option to do otherwise.
Adobe Joins Alliance for Open Media to Develop Next Generation Video Platform (http://blogs.adobe.com/conversations/2016/06/alliance-for-open-media.html)
Not only for himself royalty free.
Along with other members like Amazon, Cisco, Google, Intel Corporation, Microsoft, Mozilla, Netflix, and many others, Adobe is working to develop technology for open video compression and delivery across numerous devices. As a member of the alliance, Adobe will collaborate with industry leaders to create a leading edge and royalty-free video codec. Bottom line: this means faster and higher resolution video is on its way at a lower cost to the consumer.

Wonder, will it be free plugin like to VP9?

[dapperdan]

Daala update from IETF 96:

https://www.ietf.org/proceedings/96/slides/slides-96-netvc-4.pdf

And video (Daala section starts 1 hour in):

http://recs.conf.meetecho.com/Playout/watch.jsp?recording=IETF96_NETVC&chapter=chapter_1

Probably the headline is that "main development switched to AV1, Daala is primarily being used as a testbed to implement ideas that then get moved to AV1, though it might be revived as a standalone codec at a future date when some of the techniques they've come up with have matured".

[littlepox]

In short, daala is dead.

[LigH]

If so, then it should be buried gracefully, having delivered useful experience for the next step in the evolution of codecs... ;)

[mandarinka]

If so, then it should be buried gracefully, having delivered useful experience for the next step in the evolution of codecs... ;)

To a large degree however, many of its lessons just say what doesn't work and that it is not at all easy to beat established technology (Mpeg, VP9) with new (or more precisely, not used yet) ideas.

[littlepox]

Another thing is the design of encoder is much more challenging than design of format. Today, if I am to encode a video without x264/x265, honestly, I'd use Easy Real Producer. I've seen how it beats those crappy H.264/H.265/VP9 encoders.

R.I.P. daala, hope AV1 can fulfill its destine.

[mzso]

If so, then it should be buried gracefully, having delivered useful experience for the next step in the evolution of codecs... ;)

More like delivered useful experience in circumnavigating obvious solutions to problems that are already patented.

[Nevilne]

Which they succeeded at... don't get the hostility of people here.

[mandarinka]

Which they succeeded at... don't get the hostility of people here.

In case you meant me by the "hostility", I have been "fan" of it probably longer than almost all people on the internet. I followed the development (or preparation) maybe since 2011 or so, since time when there were no demos or public announcements yet, only Xiph wiki page and talk in #Theora.

The only "hostility" is in me not caring about the topic of "encumbered" or "free" things, which lots of people here put into the front. I just want to see technical progress regardless of whichever party will bring it about. Daala was quite interesting in that regard (at least in theory), whereas On2's stuff, not much - hence lack of respect from me.

The fact that lots of stuff projected for Daala failed in practice is true, though (sadly). Their attempt to reconcile intra prediction with lapping/OBMC is probably the most important part there, because it has lead to problems for the whole codec, need to add 1-2 loopfilters that the design originally didn't mean to use etc. Their lapping itself had a tough ride, for a long while they didn't even know if it they should really keep it (I am not up to date on how that ended up). These things are what I meant when I said that thing about nature of Daala's lessons. It doesn't mean I hate them/it. I'm not even disappointed in the sense that I would blame them or think they did a bad job. I know that this happens when you do scientific work - you go through number of theories but most of them are ruled out as wrong, with positive results/inventions being rare.

The jury is still out on AV1, but currently it seems like it will mostly inherit On2 DNA, so I don't expect much there. Hopefully Xiph/Mozilla will still manage to influence it in a good way, though.

[Jamaika]

The jury is still out on AV1, but currently it seems like it will mostly inherit On2 DNA, so I don't expect much there. Hopefully Xiph/Mozilla will still manage to influence it in a good way, though.
To this day I don't know what codec Daala improved the codec AV1. I have no way of comparison VP10 vs AV1. You may find that they are comparable in quality.

[mzso]

Which they succeeded at... don't get the hostility of people here.

You mean toward the degenerate legislation(s) that allow software patents? What is there not to get? :)
Without such they could effort into actual developments instead of circumventions...

[Nevilne]

Just so we're on the same page here I was saying that Daala devs succeeded in making a competitive video codec.
I'm really not sure what you're trying to say - do you think it was somehow daalas fault that we still have software patents?

To mandarinka: personally i'm optimistic about AV1, the spec will be good enough that we will likely get a good encoder software finally (and also some good encoding hardware).

[mzso]

Just so we're on the same page here I was saying that Daala devs succeeded in making a competitive video codec.
I'm really not sure what you're trying to say - do you think it was somehow daalas fault that we still have software patents?

To mandarinka: personally i'm optimistic about AV1, the spec will be good enough that we will likely get a good encoder software finally (and also some good encoding hardware).

I don't know why you don't get that that I was cursing software patents and not Daala...

[CruNcher]

Another thing is the design of encoder is much more challenging than design of format. Today, if I am to encode a video without x264/x265, honestly, I'd use Easy Real Producer. I've seen how it beats those crappy H.264/H.265/VP9 encoders.

R.I.P. daala, hope AV1 can fulfill its destine.

I still hope one day we gonna see those results from Real coming back for now all the NGV Results are buried somewhere deep in some Intel Restroom, though Intel is AOM member as well so who knows ;)

Sure Intel, AMD and Nvidia are mostly only there for getting their Hardware Decoding ready to go but they also have valuable R&D in different parts that might become interesting if they decide to share them :)

So there might be a possibility we gonna see Intel contributing some NGV IP there would be actually no better time then now for it instead of leaving everything taking dust in the Restroom :D

[Jamaika]

Daala finally provides color depth 10 bit. Unfortunately, the lack of support the CPU very limited coding. (max. fullhd) It's very slow. Still need to improve the player because he doesn't work.:rolleyes:
daalaenc.exe --b-frames 4 --keyframe-rate 23.976 --complexity 5 --video-quality 18 --fpr --output "image_rate18.ogg" -
Output #0, yuv4mpegpipe, to 'pipe:':
Metadata:
date : 2016-09-07T08:27:49+02:00
encoder : Lavf57.50.100
Stream #0:0: Video: wrapped_avframe, yuv444p10le, 4096x2304, q=2-31, 200 kb/s, 23.98 fps, 23.98 tbn, 23.98 tbc
Metadata:
encoder : Lavc57.57.101 wrapped_avframe
Stream mapping:
Stream #0:0 -> #0:0 (rawvideo (native) -> wrapped_avframe (native))
Press [q] to stop, [?] for help
File '-' is 4096x2304 23.976 fps 444p10 video.
Compressing...
frame= 7 fps=0.6 q=-0.0 size= 387072kB time=00:00:00.29 bitrate=10860790.5kbits/s speed=0.0241x
https://github.com/xiph/daala/commit/05243557bc3e59872fd043c99dc4c17ca33bcb1b
https://github.com/xiph/daala/commit/4f0cf4203d09399515a824f1056dcbec9f5d1b63
https://github.com/xiph/daala/commit/8af4b2fb8021b5b4da5212c0e25c6066cddf1f41

[Clare]

A poster summarizing Daala's technologies:
https://jmvalin.ca/video/mmsp2016_poster.pdf

[mzso]

A poster summarizing Daala's technologies:
https://jmvalin.ca/video/mmsp2016_poster.pdf

So AV1 is a tweaked(vp9), tweaked (vp10), tweaked(av1) VP8. Is it worth the fuss? They could have just renamed and released VP9, or whichever state was VP10 at that point as AV1.

Then they could have moved on to AV2 with drastic/revolutionary changes.

[Quikee]

So AV1 is a tweaked(vp9), tweaked (vp10), tweaked(av1) VP8. Is it worth the fuss? They could have just renamed and released VP9, or whichever state was VP10 at that point as AV1.

No, I don't know where you've read that. On the poster it clearly says that AV1 is VP9 as base + parts from Daala, Thor (and VP10) and new contributions (work that has been done exclusively on AV1 codebase).

Then they could have moved on to AV2 which drastic/revolutionary changes.

AV1 was never meant to have revolutionary changes - it needs to get to the market rather quickly (to still take advantage of HEVC licensing uncertainty).

[mzso]

On the poster it clearly says that AV1 is VP9 as base + parts from Daala, Thor (and VP10) and new contributions (work that has been done exclusively on AV1 codebase).

AKA. Tweaked VP9, the way I see it. :)


AV1 was never meant to have revolutionary changes - it needs to get to the market rather quickly (to still take advantage of HEVC licensing uncertainty).

Missed my second point too. :) If that was the most significant aspect they should have just released as soon as they got an agreement upon AV1. VP10 + throw in whatever is considered "mature" and gainful from the others. Then start optimizing the encoder/decoder.

[Nevilne]

Well thankfully they don't listen to you