[HEVC] x265.EXE: mingw builds

[Selur]

MPC-BE developers are NOT using the ffmpeg/libav code correctly

they (or the lav-filter they use) might be simply using an older libav version,...

[Gomorrite]

--- SNIPP ---
No, you were right. At least in my case changing the cropping in Avisynth solved the problem. Thanks.

Sín embargo, I didn't have to use mod-8 resolutions to make x265 encode properly with a "non-complex" command-line.

Therefore, I am two-times right, and you are two-times wrong

But I had tried most command lines proposed in this thread or found elsewhere, from the simplest to the complex ones, and every single one gave me the same result. It was only the change in the cropping that made it work. So there has to be something else that made your test different from mine.

[El Heggunte]

......................

But I had tried most command lines proposed in this thread or found elsewhere, from the simplest to the complex ones, and every single one gave me the same result.

As I tried to say, "that's too vague" (and not very-helpful as well).

Example of "bad" command-line below:

avs2yuv 0ddres.avs -o -|5259.exe --y4m --bitrate 1200 --tune ssim --ssim --output 0ddrestest.265 -

0ddres.avs: 628x514, 25 fps, 25 frames
y4m  [info]: 628x514 25Hz C420, unknown frame count

x265 [info]: HEVC encoder version 0.5+259-9d74638c3640
x265 [info]: build info [Windows][GCC 4.8.2][32 bit] 8bpp
x265 [info]: Main profile, Level-3 (Main tier)

x265 [info]: CU size                      : 64
x265 [info]: Max RQT depth inter / intra  : 1 / 1
x265 [info]: ME / range / subpel / merge  : star / 60 / 5 / 3
x265 [info]: Keyframe min / max           : 250 / 250
x265 [info]: Rate Control                 : ABR-1200 kbps
x265 [info]: Lookahead / bframes / badapt : 40 / 3 / 1
x265 [info]: tools: rect amp rd=0 ref=3 lft sao-lcu sign-hide aq
x265 [info]: frame I: 1       PSNR Mean: Y:33.590 U:43.473 V SSIM Mean: 0.965
x265 [info]: frame P: 6       PSNR Mean: Y:40.808 U:45.430 V SSIM Mean: 0.980
x265 [info]: frame B: 16      PSNR Mean: Y:40.878 U:45.431 V SSIM Mean: 0.980
x265 [info]: global : 23      PSNR Mean: Y:40.543 U:45.346 V SSIM Mean: 0.979

encoded 23 frames in 119.31s (0.19 fps), 960.85 kb/s, Global PSNR: 41.686
Global SSIM: 0.979

www.mediafire.com/download/kxo9h1m4vr2au4z/0ddrestest.png.zip

Example of "good" command-line below:

avs2y4m 0ddres.avs -o -|5259.exe --y4m --me 1 --no-wpp --fast-cbf --hash 0 --max-merge 1 --no-tskip --tu-inter-depth 2 --tu-intra-depth 1 --merange 32 --keyint 12 --qp 21 --bframes 0 --rc-lookahead 0 -o 0ddreztest.265 -

0ddres.avs: 628x514, 25 fps, 25 frames
y4m  [info]: 628x514 25Hz C420, unknown frame count

x265 [info]: HEVC encoder version 0.5+259-9d74638c3640
x265 [info]: build info [Windows][GCC 4.8.2][32 bit] 8bpp
x265 [info]: Main profile, Level-3 (Main tier)

x265 [info]: CU size                      : 64
x265 [info]: Max RQT depth inter / intra  : 2 / 1
x265 [info]: ME / range / subpel / merge  : hex / 32 / 5 / 1
x265 [info]: Keyframe min / max           : 12 / 12
x265 [info]: Rate Control                 : CQP-21
x265 [info]: Lookahead / bframes / badapt : 0 / 0 / 1
x265 [info]: tools: rect amp cfm rd=0 ref=3 lft sao-lcu sign-hide
x265 [info]: frame I: 3       PSNR Mean: Y:48.457 U:49.609 V:49.546
x265 [info]: frame P: 22      PSNR Mean: Y:48.457 U:49.609 V:49.546
x265 [info]: global : 25      PSNR Mean: Y:48.457 U:49.609 V:49.546

encoded 25 frames in 99.84s (0.25 fps), 1280.43 kb/s, Global PSNR: 48.737

www.mediafire.com/download/02u9dwfu46awnfh/0ddreztest.png.zip

[Mephesto]

Mmmmm, dat ass.... yummy

[El Heggunte]

Mephesto --- unprofessional and anti-sportive behavior shall not be tolerated

@ all: considering that not everybody likes to frequent Doom9's forum,
I have quoted a relevant post by Tom Vaughan, so that everyone can be well-informed

I've been playing around with quality and scenario tuning of x265 (using 0.5+259-9d74638c3640). I have a few questions I haven't seen answered yet.

Ben,
Answers from our lead developer inline below...

* Does x265 CRF generally match the perceptual quality of x264 CRF? Or are they on a different visual scale due to QP being handled differently or anything? I note the default is 28, which is higher than x264's.

CRF is a feature contributed by an open-source developer, back-porting this from x264, so we are roughly the same. However in general, I believe CRF really wants a decent adaptive quant feature and ours is still incomplete. It would really benefit from mb-tree but that is not even started yet.

* When using CRF, should --tune be left alone to let Rate Factor work, or will it still try to default to --tune PSNR, in which case using --tune SSIM will probably offer better VQ?

the --tune PSNR and --tune SSIM switches currently do nothing. We are still working through this as part of the AQ work.

* Was --weightp turned on by default? The Evaluator's guide suggests no, but the --preset parameters suggest yes

It is not, yet. We're waiting for weightp lookahead analysis before turning it on by default. We're currently *also* turning it off for ultrafast preset just to make sure when the default is switched on, the ultrafast preset still disables it. I expect this to be switched on by default this week.

* Does --aq-mode default to off due to quality concerns or what?

Basically, AQ, CRF, --tune SSIM and --weightp are all unfinished features in the middle of construction

* is --merange in pixels of final frame, or a half-size frame, or what?

It is in pixels of the final frame.

Tom

[Mephesto]

Anti-sportive, what's that mean?I was only praising the model in the vid, god bless x265 for decoding the hottie properly.
Compressing that should be a crime against humanity, for real.

On another note, will CRF/QP be phased out in favor of that new QPRD method H265 standardized, or has this already been somewhat implemented in x265?

[Selur]

new QPRD method H265 standardized

got a link for that?
QPRD ?=? quater pixel rate distortion

btw may be we should start a thread with video compression related acronyms like:
CABAC: Context-Adaptive Binary Arithmetic Coding
CB: Coding Block
CTB: Coding Tree Block
CTU: Coding Tree Unit
CU: Coding Unit
HEVC: High Efficiency Video Coding
JCT-VC: Joint Collaborative Team on Video Coding
MC: Motion Compensation
ME: Motion Estimation
MOS: Mean Opinion Score
PB: Prediction Block
PU: Prediction Unit
RDO: Rate Distortion Optimization
RDOQ: Rate Distortion Optimized Quantization
SAD: Sum of Absolute Differences
SAO: Sample Adaptive Offset
SATD: Sum of Absolute Transformed Differences (SATD)
SIMD: Single Instruction Multiple Data
SSD: Sum of Squared Difference
SSE: Streaming SIMD Extensions
TB: Transform Block
TU: Transform Unit
CRF: Constant Rate Factor
CQ: Constant Quantizer
SSIM: Structural Similarity Index
PSNR: Peak Signal-to-Noise Ratio

[Mephesto]

RDO Quantization, I held onto this that I read years ago and remembered it as the new CRF which employs RDO by default or something. Looks like I misread, nvm.
http://www.h265.net/2009/06/quantization-techniques-in-jmkta-part-4.html

[El Heggunte]

^
^ Yes$S, and that thread should become a sticky as well.

[Selur]

Looks like I misread, nvm.

Okay, I was wondering if I missed something.

[tnti]

@ El HeggunteThat is the problem caused by WPP Options
You can try to issue the video sent to the mailing list X265
help fix the problem

[Selur]

btw. is there some sort of user-mailinglist for x265? (normally a dev-mailinglist is ment for developers only,..)

[tnti]

#1 have

I do not know you did not pay attention to see

[tnti]

Did not add the HEVC support MKVToolNix

Using FFmpeg patch from DIVX


l-smash in the new version seems to disable support for HEVC


Of course, I did not personally done a test compile l-smash


I just listen to the mouth from a friend

[Selur]

@tnti: assuming that was ment for me: #1 does mention different emails (one being the devel-mailing list email).
Which of these is ment for users?
-> will probably stick with mentioning bugs to the bug tracker (https://bitbucket.org/multicoreware/x265/issues) since that seems to be the only appropriate way.

[tnti]

bug tracker response speed is very slow

This is just my guess


Just let developers see the error on the line this is the ultimate goal no matter what method you use

[Selur]

bug tracker response speed is very slow

which is a shame, since the bug tracker should be the first thing developers check (otherwise the bug tracker should be removed)

Just let developers see the error on the line this is the ultimate goal no matter what method you use

sure the goal is to show the developers the problem, but since it's unclear how to and where to report a bug properly the whole thing is a mess.

[tnti]

You can change the WPP alone option without changing other options look

You can determine if the problem lies in the WPP

The BUG exists in many versions

bug tracker Not user-friendly for other languages


English poor user does not know how to submit BUG

[LigH.de]

Mmmmm, dat ass.... yummy

Using erotic material in scientific image compression algorithm tests has some tradition: e.g. the world famous head of Lenna was a cropped Playboy centerfold image of Lena Söderberg.

[El Heggunte]

Mmmmm, dat ass.... yummy

Using erotic material in scientific image compression algorithm tests has some tradition: e.g. the world famous head of Lenna was a cropped Playboy centerfold image of Lena Söderberg.

Thanks for the reminder , and BTW,

http://en.wikipedia.org/wiki/China_Girl_(film_term)

[poisondeathray]

So maybe El Heggunte should post the source so we can inspect it more closely

But next time don't post something that looks like jailbait . Don't need no FBI crashing down doors LOL

[El Heggunte]

{{ edited }}

[vhelp]

i didn't see any pics, nor any pm's of sources / links, cheap bstrd.. i'm upset. but i saw lena. actually, old news though. have that (original) huge png somewhere on my old win98 machine from years so. old news ligh. anyway.

anyone had any bad experience with mpc-be showing/not showing at random ?

when i rt-click a video to play, its there one day, and not the next, and so on. today its not avail..no wait, 20 mins later, its avail

[El Heggunte]

i didn't see any pics, nor any pm's of sources / links, cheap bstrd.. i'm upset.

Just read post#903 again. I replaced the image attachments with coded hyperlinks to Mediafire, because poisondeathray has a point.

[vhelp]

edit...i moved this topic to the "Software Playing" section.

[rhaz]

I get stuck when compiling at 23% simply nothing happens, and Task Manager shows yasm.exe at 0% CPU usage.


Update: OK this can be deleted. My security tools were isolating various processes and etc. Sorry, false alert.

And thanks for this great tool!

[LigH.de]

Compilers generating CPU optimized code can be highly suspicious to resident antivirus tools...

[Selur]

btw. reported the out-of-sync presets and the problems compiling (against my better judgment) to the dev-mailinglist.

[rhaz]

Okay I am totally new to this. Can someone make short guide on how to encode my blurays?

This is what I did:
- Extract video from .m2ts to .mkv using eac3to.
- Now I need to convert AVC everytime to YUV or Y4M? I converted 1 minute long .mkv sample extracted from .m2ts resulting the size less than 200MB. Then I used this command:

ffmpeg -i test.avs -an -pix_fmt yuv420p -f rawvideo -y test.yuv

and I got .yuv in the size of almost 2GB! Also so how long would it take take to convert entire BluRay to .yuv and what size would it be? Is this normal all this? And is yuv420p the only option? I mean, my videos I want are at least 720p. And yuv720p didn't work.
- So what next do I do so I could finally encode this sample and get my first x265 encoder ever? Please give me tools and examples how to use x265 command line to encode this .yuv from .avs or something.

Thanks,
Cheers

[rhaz]

OK I managed to make my first encode with simplest line and no extra settings. So my question is how do I get an encode I want, i.e. the output resolution and resizers like lanczos, how do I give settings like these? Adding simple .avs created via MeGUI don't work. And how do you mux .hvec to a container like .mkv?