Anyone tested VP9 10bit?

[Selur]

(Thinking about adding support for it to Hybrid, but not sure if it's really worth the effort atm., since last time when I tested it 10bit encoding wasn't stable and since mult threading doesn't seem to be anything Google is interested in atm. VP9 is rarely used by me.)

-> Just wanted to ask if anyone tested the 10bit (and 4:2:2) encoding in VP9 and compared it to x264 and x265?

Cu Selur

[Marchand]

Okay.
I will perform some tests and then compare.
Very well remembered.

E.T: ffmpeg version N-67586-g3e1ac10 Copyright (c) 2000-2014 the FFmpeg developers<br/>
built on Nov 12 2014 22:01:54 with gcc 4.9.2 (GCC)<br/>

Incompatible pixel format 'yuv422p' for codec 'libvpx-vp9', auto-selecting format 'yuv420p'<br/>
Incompatible pixel format 'yuv444p' for codec 'libvpx-vp9', auto-selecting format 'yuv420p'<br/>

[Marchand]

Google's Web-video ambitions bump into hard reality -

by Stephen Shankland
@stshank
September 12, 2014 11:17 AM PDT


The company's technical prowess and free VP9 licensing haven't been enough to dent the fortunes of rival compression format HEVC. But Google's already moving on to VP10.

VP10 improvements

If Google's VPx improves faster than HEVC and whatever follows, that would undoubtedly help its prospects. That's one of the hopes for VP10, Ramamoorthy said.

"Higher quality per bit" is always a priority, he said, meaning that Google wants to make even better use of available network capacity.

Another priority is simplicity. Google wants to make the codec easier to use without tradeoffs and optimization hassles. Today, those wanting to stream video have to worry too much about whether it's going to mobile devices or PCs or whether it's headed for fast or slow networks. "These are all complexities because of legacy [technology]," he said.

He wouldn't comment on VP10 arrival times, but said Google is working hard to speed up development. The gap between VP8 and VP9 was 24 months, but he hopes Google can get iteration cycles down to 18 months.

That can be tough given that Google also wants more parties involved in codec development.

"We want to get more industry participation with the next effort. We want to find how to get that without sacrificing the speed of VP9," Ramamoorthy said. He showed little enthusiasm for the more formal standardization process but appreciates how it offers a chance for multiple parties to define technology that's useful for all of them. "If that can be achieved without the current cons of the standardization process, we are definitely interested in that," he said.

If Google can make its 18-month goal, that would mean VP10 would arrive about the end of this year, with another 18 months or so for that version of the codec to arrive in chips. Google expects the first VP9-equipped mobile devices to appear in 2015.

But Google's got a lot of work to do to convince skeptics unimpressed with its codecs so far. The middling success of VP9 affects VP10's prospects.

"We have yet to look at VP10," Encoding.com's Heil said, "and will probably not do so unless VP9 shows some more traction."

Source with more information:

VP10 is coming soon. I'll wait.

[Selur]

Incompatible pixel format 'yuv422p' for codec 'libvpx-vp9', auto-selecting format 'yuv420p'<br/>
Incompatible pixel format 'yuv444p' for codec 'libvpx-vp9', auto-selecting format 'yuv420p'<br/>

Then the libvpx version you are using isn't based on the highbitdepth branch of the source but the main branch.

[Marchand]

OK Selur

Profile 3 adds support for 4:2:2 chroma subsampling, 4:4:4 chroma subsampling, and alpha channel support

I will try with the Hybrid

[Selur]

Hybrid doesn't come with a highbitdepth vp9 build either.
My main motivation behind this thread was to gather some information if it is worth looking into it at this point in time or if I should wait till highbitdepth branch and master branch are merged,..

[Marchand]

Hybrid doesn't come with a highbitdepth vp9 build either.
My main motivation behind this thread was to gather some information if it is worth looking into it at this point in time or if I should wait till highbitdepth branch and master branch are merged,..

Hummm...OK

I'll find another way to test the VP9 - 10bits. 4.2.2 .... now I was curious about it. hehehehe

[Selur]

Any news in this regard? Especially in the decoder section?
I can encode 10bit&12bit content, but since libav doesn't support decoding of such files, it's really a pain the check the output. (having to use vpxdec)

[Jamaika]

(Thinking about adding support for it to Hybrid, but not sure if it's really worth the effort atm., since last time when I tested it 10bit encoding wasn't stable and since mult threading doesn't seem to be anything Google is interested in atm. VP9 is rarely used by me.)

For me it does not make sense. 10bit is a larger amount from this information. This means higher bitrate. Where's hardware acceleration graphics card, processor? There is in my case there will be a long time because it is a replacement of the entire computer.
Besides talking about 10bit and codec not have a multi-stage video quality slider. For the "best" days are long and tiring.
The third issue is the muxer. Matroska container, WebM are not recognized by some companies to pirate such as Sony. Neither describe what codec saved parameters. Is a fully commercial? And the main thing is not yet applicable in cameras or recorders.
Although advertising is a container MOX, whose eyes have seen.

[LigH.de]

Some people don't mind "enough bitrate" as long as it preserves enough quality; there are even situations where a "high bit depth" of transformed parameters allows more efficient compression than 8 bit resolution. Certainly it helps avoiding some banding in low detail scenes due to less rounding errors and finer quantization steps, as demonstrated often with x264 and x265. But yes, a higher resolution of encoded parameters requires a more complex decoder too. Hardware support is therefore rather sparse. Yet.

[newpball]

The Premiere WebM plugin supports 10 (and 12) bit encoding.

http://www.webmproject.org/

Problem is that it is much slower than H.264/H.265

A 30 second 1080/60P clip in 10 bit takes me more than 10 minutes to encode!

Parallelization is better than before but still bad, only an average of 30% CPU utilization.

However how do we view 10bit VP9 videos?

[Jamaika]

Can anyone knows where can I download container MOX?
https://www.indiegogo.com/projects/mox-file-format#/updates

[Selur]

and this is related to VP9 and 10bit, how?

[Jamaika]

Indeed, nohow. Muxer MOX is supposed to improve data throughput.
Advertising is that all types of encoding VP9 plays the VLC player. This is not true.
I recently discovered that recreates VP9 v1.4 10bit i444 ffplay.exe v20150613.

ffmpeg.exe -i "video-2007-12-01-14-03-15.mp4" -s 1280x640 -r 23.976 -an -sn -f rawvideo -pix_fmt yuv444p10le - | vpxenc.exe -v --bit-depth=10 --input-bit-depth=10 --i444 --profile=3 --threads=4 --passes=1 --pass=1 --good --codec=vp9 --fps=24000/1001 --cpu-used=3 -w 1280 -h 640 --target-bitrate=2000 - -o "vp90_yuv444p10le.webm"

Edit: MOX muxer costs $ 20, now subject has ceased to interest me.

[Selur]

For me it does not make sense. 10bit is a larger amount from this information. This means higher bitrate.

In a compressed domain a lot more data might only need a bit more space. The whole topic has been discussed when H.264 10bit was supported in x264.

However how do we view 10bit VP9 videos?

ffplay should work

Cu Selur

Ps.: btw. MOX isn't really interesting for most users since it only supports Dirac, OpenEXR, DPX, PNG, and JPEG as video format iirc.

[Jamaika]

In a compressed domain a lot more data might only need a bit more space. The whole topic has been discussed when H.264 10bit was supported in x264.

The subject may have been discussed. Currently codec VP9 under Adobe Premiere me not working properly. This is officially a free beta version 1.0.0.0. No patches.

Firstly, I have not found a command from the command help for this codec in Adobe Premiere CC 2014. Additional commands are useful. As it is not known what they are, for the beginner will be a big problem to guess. Just shall fall, but it will do something unconsciously.
After the second test involved only movies 8bit i420. For the rest I have not found decoder. On my six-year-compression machine it takes longer than on the codec vpxenc, but you do not need to combine with temporary files outside the editor.
Thirdly. I asked parameters such as pictured. The film was for the export of 1-pass, good quality great quality, as for me. The bitrate was reduced from 50Mbps up to 3Mbps. It is a film with very low traffic dynamics.
Fourthly. The film did not keep the original colors although he had asked all the parameters. I put pictures. The pictures were taken MPC-HC 1.7.8.162. MPC-BE Decoder 1.4.4 (beta) 2015-04-05.
Fifthly. The film has not survived bitrate. It was to be size 24MB and 48MB arose. It's not that important in the end is not Bluray.
Case Youtube and revelations from VP9.
http://filmowiec.pl/forum/viewtopic.php?f=12&t=29083&hilit=webm#p166763

[Selur]

So nothing new then.
That 2pass doesn't hit the file size has been known, that decoding of 10bit is only supported by ffplay atm. is known.

[Jamaika]

A month ago, the decoder has not yet acted. I did not know about this information.
https://github.com/FFmpeg/FFmpeg/blob/master/Changelog
The program also Hybrid VP9 I could not play 10bit i444.

[Selur]

Hybrid uses mplayer for preview, so no surprise there either,...

[Jamaika]

I found a player that plays VP9 10bit i444 as someone is interested. It is the latest beta version of MPC-HC 1.7.9.14.
This program does not play color space "sRGB"