Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Video Encoding > High Efficiency Video Coding (HEVC)

Reply
 
Thread Tools Search this Thread Display Modes
Old 8th December 2019, 16:00   #7241  |  Link
_kermit
Registered User
 
Join Date: Apr 2017
Posts: 23
Quote:
Originally Posted by Boulder View Post
A source clip can always be useful. Easier to see what's really happening.
see if you can access that (uploading):

https://1drv.ms/u/s!AuVhsELqus7qlI9E...GSG-g?e=GuXnag

cheers, roland
_kermit is offline   Reply With Quote
Old 8th December 2019, 17:30   #7242  |  Link
Boulder
Pig on the wing
 
Boulder's Avatar
 
Join Date: Mar 2002
Location: Hollola, Finland
Posts: 4,636
Quote:
Originally Posted by _kermit View Post
see if you can access that (uploading):

https://1drv.ms/u/s!AuVhsELqus7qlI9E...GSG-g?e=GuXnag

cheers, roland
What frame is the one that shows the issue best? I'm not seeing anything specific in my encode at CRF 18, preset slower and slight tweaks.
__________________
And if the band you're in starts playing different tunes
I'll see you on the dark side of the Moon...
Boulder is offline   Reply With Quote
Old 8th December 2019, 19:44   #7243  |  Link
_kermit
Registered User
 
Join Date: Apr 2017
Posts: 23
Quote:
Originally Posted by Boulder View Post
What frame is the one that shows the issue best? I'm not seeing anything specific in my encode at CRF 18, preset slower and slight tweaks.
at 24 seconds, the woman in the middle.

But I've just changed the monitor I play the video on and the other isn't doing it, at least I can't see it.
When using another preset on the first display, I also don't see it anymore.
Can just the preset of a monitor cause that?
(I don't watch them on those, they are just for checking)
_kermit is offline   Reply With Quote
Old 8th December 2019, 21:26   #7244  |  Link
Boulder
Pig on the wing
 
Boulder's Avatar
 
Join Date: Mar 2002
Location: Hollola, Finland
Posts: 4,636
Yes, it's very much possible. Different calibrations or settings can change the image a lot, so you're always on the safe side if you can check it on the display where you will be watching the final video. Of course, for most of us it's not very convenient.
__________________
And if the band you're in starts playing different tunes
I'll see you on the dark side of the Moon...
Boulder is offline   Reply With Quote
Old 8th December 2019, 22:19   #7245  |  Link
_kermit
Registered User
 
Join Date: Apr 2017
Posts: 23
Quote:
Originally Posted by Boulder View Post
Yes, it's very much possible. Different calibrations or settings can change the image a lot, so you're always on the safe side if you can check it on the display where you will be watching the final video. Of course, for most of us it's not very convenient.
so, I guess all the fuss for nothing. Sorry for that.
_kermit is offline   Reply With Quote
Old 9th December 2019, 00:05   #7246  |  Link
markiemarcus
Registered User
 
Join Date: May 2018
Posts: 16
Quote:
Originally Posted by _kermit View Post
at 24 seconds, the woman in the middle.

But I've just changed the monitor I play the video on and the other isn't doing it, at least I can't see it.
When using another preset on the first display, I also don't see it anymore.
Can just the preset of a monitor cause that?
(I don't watch them on those, they are just for checking)
It's quite a difficult scene, especially leading up to that point. It's prone to grain freezing in the exterior shots and the level is low enough that RDOQ makes it appear more coarse and boxy than it actually is. I've run into this on animated sources that make heavy use of dither. The establishing shot of the bar @ 9 seconds in the red/brown sky is great example. RDOQ level 2 makes a mess of it, even at the default value. If you increase either (Psy-rd or Psy-RDOQ), the situation gets even worse. I'm getting more pleasing and accurate results from RDOQ Level 0, maxing out Psy-rd, and using a lower CRF to arrive at the same file size. Is RDOQ really supposed to behave like this? Because it also blows up the filesize so I can't see how it can be efficiency related.

Default RDOQ and Psy settings:
https://i.ibb.co/vBbXhkh/Psy2-RDOQ1.png

RDOQ Level 0, Psy-rd 5:
https://i.ibb.co/cFK5JhV/RDOQLevel0-Psy5.png

Aq mode 1 is coming out ahead for me. Default AQ strength seems fine.

Grain motion in the exterior shots is definitely helped with --rd 6.

Last edited by markiemarcus; 9th December 2019 at 00:58.
markiemarcus is offline   Reply With Quote
Old 9th December 2019, 00:14   #7247  |  Link
markiemarcus
Registered User
 
Join Date: May 2018
Posts: 16
Quote:
Originally Posted by _kermit View Post
so, I guess all the fuss for nothing. Sorry for that.
I read somewhere on here a while back that 10-bit playback on 8-bit displays can be a little unpredictable.
markiemarcus is offline   Reply With Quote
Old 9th December 2019, 00:58   #7248  |  Link
_kermit
Registered User
 
Join Date: Apr 2017
Posts: 23
Quote:
Originally Posted by markiemarcus View Post
I read somewhere on here a while back that 10-bit playback on 8-bit displays can be a little unpredictable.
that does make sense
_kermit is offline   Reply With Quote
Old 9th December 2019, 07:39   #7249  |  Link
froggy1
ffx264/ffhevc author
 
froggy1's Avatar
 
Join Date: May 2007
Location: Belgium
Posts: 1,552
@_kermit

I can't see anything in your picture you posted (haven't checked the sample yet). Maybe it's a monitor calibration issue?
__________________
ffx264--ffhevc--ffxvid
froggy1 is offline   Reply With Quote
Old 9th December 2019, 14:58   #7250  |  Link
_kermit
Registered User
 
Join Date: Apr 2017
Posts: 23
Quote:
Originally Posted by froggy1 View Post
@_kermit

I can't see anything in your picture you posted (haven't checked the sample yet). Maybe it's a monitor calibration issue?
it seems what I saw was related to the display used.
_kermit is offline   Reply With Quote
Old 9th December 2019, 23:07   #7251  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 2,984
Quote:
Originally Posted by markiemarcus View Post
I read somewhere on here a while back that 10-bit playback on 8-bit displays can be a little unpredictable.
Yeah. If a display controller truncates instead of dithers, you might get more banding than if the conversion to 8-bit was done pre-encode.

Sent from my SM-T837V using Tapatalk
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 9th December 2019, 23:09   #7252  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 2,984
Quote:
Originally Posted by _kermit View Post
Tried now:



--vbv-maxrate 15000 --deblock=-2:-2 --amp --selective-sao 1 --rskip --output-depth 10 --aq-mode 3 --qg-size 64 --preset slower --crf 20



and



--vbv-maxrate 15000 --deblock=-2:-2 --amp --selective-sao 1 --rskip --output-depth 10 --aq-mode 1 --qg-size 64 --preset slower --crf 18



yes, 3.2.0.18



still shows. It may be less, but still obvious



Interesting part is that it only is that obvious on that region with black/purple texture



it is grainy, but there is worse.



does it help uploading the original and encoded clip?

Onedrive working?
If it is in black/purple, try --aq-mode 3. It doesn't take much blocking in the very low luma range to show up as an artifact with 8-bit SDR.

Sent from my SM-T837V using Tapatalk
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 10th December 2019, 08:01   #7253  |  Link
alexantr
Registered User
 
alexantr's Avatar
 
Join Date: Nov 2019
Location: Belarus
Posts: 1
Quote:
Originally Posted by markiemarcus View Post
...I'm getting more pleasing and accurate results from RDOQ Level 0, maxing out Psy-rd, and using a lower CRF to arrive at the same file size...
rdoq-level=1 can help to retain small noise looking like dither noise. Keep original psy-rd value and try to tune psy-rdoq from 0.5 to 1.0.
alexantr is offline   Reply With Quote
Old 10th December 2019, 10:18   #7254  |  Link
markiemarcus
Registered User
 
Join Date: May 2018
Posts: 16
Quote:
Originally Posted by alexantr View Post
rdoq-level=1 can help to retain small noise looking like dither noise. Keep original psy-rd value and try to tune psy-rdoq from 0.5 to 1.0.
Oh definitely, Level 1 I've had better experience with in the past, particularly at lower values of around 0.5. I'm just scratching my head as to why the default is Level 2; I'm struggling to see any positives. It's also slower IIRC.

However, even with Level 1, RDOQ at 1.0 and Psy-rd at 2.0, I'm able to get better grain motion and more consistent results at the same filesize with Level 0 and massively elevated Psy-rd values instead.

It just doesn't seem right. Surely I'm not the only one who's finding this?

Last edited by markiemarcus; 10th December 2019 at 10:24.
markiemarcus is offline   Reply With Quote
Old Yesterday, 08:36   #7255  |  Link
RainyDog
Registered User
 
Join Date: May 2009
Posts: 170
Quote:
Originally Posted by markiemarcus View Post
Oh definitely, Level 1 I've had better experience with in the past, particularly at lower values of around 0.5. I'm just scratching my head as to why the default is Level 2; I'm struggling to see any positives. It's also slower IIRC.

However, even with Level 1, RDOQ at 1.0 and Psy-rd at 2.0, I'm able to get better grain motion and more consistent results at the same filesize with Level 0 and massively elevated Psy-rd values instead.

It just doesn't seem right. Surely I'm not the only one who's finding this?
Frustratingly there's just no perfect PSY/RDOQ setting with x265 unfortunately

But hours (probably amounting to days...) of testing has led me to conclude that the most balanced for most sources at mid-high bitrates is usually psy-rd 1.0, psy-rdoq 1.0 and rdoq-level 2.

I know what you mean with rdoq-level 0 though... In motion, it can look the most consistent and easiest on the eye. But if you compare encode frames to source frames you'll never get transparency with level 0. There seems to be some high-pass filter that sort of dulls grain and high-frequency detail even with psy-rd turned up to 5.0.

With rdoq-level 2 and psy-rd 1.0 plus psy-rdoq 1.0, you can usually see that it at least tries to replicate the source even if that can lead to ugly and sometimes weird artifacts if your bitrate isn't sufficient enough to support it.

As for rdoq-level 1, I find this to just be a rougher, less refined or accurate level 2.

If you can get your hands on any internal 'haich dee bits' x265 encodes for, ahem, research purposes then you'll find that they all use rdoq-level 2 with psy-rd 1.0 and psy-rdoq 1.0. Give or take anyway... I've seen psy-rd 1.2 + psy-rdoq 1.1 and psy-rd 1.4 + psy-rd 1.0 but always within those boundries.

They always consistently use no-cutree and usually qg-size 8 too. AQ mode 1 or 3 and often lots of zones as well. EDIT: and always CTU 32 and no-rskip.

Last edited by RainyDog; Yesterday at 08:40.
RainyDog is offline   Reply With Quote
Old Yesterday, 09:19   #7256  |  Link
markiemarcus
Registered User
 
Join Date: May 2018
Posts: 16
Quote:
Originally Posted by RainyDog View Post
Frustratingly there's just no perfect PSY/RDOQ setting with x265 unfortunately

But hours (probably amounting to days...) of testing has led me to conclude that the most balanced for most sources at mid-high bitrates is usually psy-rd 1.0, psy-rdoq 1.0 and rdoq-level 2.

I know what you mean with rdoq-level 0 though... In motion, it can look the most consistent and easiest on the eye. But if you compare encode frames to source frames you'll never get transparency with level 0. There seems to be some high-pass filter that sort of dulls grain and high-frequency detail even with psy-rd turned up to 5.0.

With rdoq-level 2 and psy-rd 1.0 plus psy-rdoq 1.0, you can usually see that it at least tries to replicate the source even if that can lead to ugly and sometimes weird artifacts if your bitrate isn't sufficient enough to support it.

As for rdoq-level 1, I find this to just be a rougher, less refined or accurate level 2.

If you can get your hands on any internal 'haich dee bits' x265 encodes for, ahem, research purposes then you'll find that they all use rdoq-level 2 with psy-rd 1.0 and psy-rdoq 1.0. Give or take anyway... I've seen psy-rd 1.2 + psy-rdoq 1.1 and psy-rd 1.4 + psy-rd 1.0 but always within those boundries.

They always consistently use no-cutree and usually qg-size 8 too. AQ mode 1 or 3 and often lots of zones as well. EDIT: and always CTU 32 and no-rskip.
There is definitely some sort of lpf noticeable with level 0 and I totally agree that achieving transparency is pretty much impossible that way. But I still think that subjectively it often looks better, almost always so if you're trying to preserve dither. I don't mind the slight dulling of grain assuming what is there is uniform, the pitch fine and the resulting encode looks nice. It's so impressive to see the bitrate reduction you can achieve, even on grainy content, at bitrates where x264 completely falls apart.

I'm just seeing a lot of instances where RDOQ looks pretty nasty, regardless of the bit rate or values used.

Will definitely check out your suggestions though. Appreciate it.

Last edited by markiemarcus; Yesterday at 09:40.
markiemarcus is offline   Reply With Quote
Old Yesterday, 10:38   #7257  |  Link
Yanak
Registered User
 
Join Date: Oct 2011
Posts: 268
Hello,

is anyone having issues trying to build with Intel compiler 2019 integrated in Visual studio 2019 ?

Building the multilib with VS 2019 works,
building the multilib with LLVM Clang integrated inside Visual Studio 2019 works too, ( either the clang 9.0.0.0 version from the VS2019 installer or the last 10.0.0 version of LLVM installed separately and integrated in VS2019 with the LLVM2019 extension from the VS Marketplace. )

But trying to use the Intel compiler brings loads of errors and the build fails, tried also to make the single 10Bits build and it fails too.

Was able to use ICC 2019 inside VS2017 some months ago and it was building correctly so I'm not sure if it's a problem on my side only or more general issue on the code, but after dozen of tries i can't find out what is wrong with Intel compiler so i ask here in case anyone have an idea.

Thank you .
Yanak is offline   Reply With Quote
Old Yesterday, 18:24   #7258  |  Link
_kermit
Registered User
 
Join Date: Apr 2017
Posts: 23
Quote:
Originally Posted by RainyDog View Post
Frustratingly there's just no perfect PSY/RDOQ setting with x265 unfortunately

But hours (probably amounting to days...) of testing has led me to conclude that the most balanced for most sources at mid-high bitrates is usually psy-rd 1.0, psy-rdoq 1.0 and rdoq-level 2.

I know what you mean with rdoq-level 0 though... In motion, it can look the most consistent and easiest on the eye. But if you compare encode frames to source frames you'll never get transparency with level 0. There seems to be some high-pass filter that sort of dulls grain and high-frequency detail even with psy-rd turned up to 5.0.

With rdoq-level 2 and psy-rd 1.0 plus psy-rdoq 1.0, you can usually see that it at least tries to replicate the source even if that can lead to ugly and sometimes weird artifacts if your bitrate isn't sufficient enough to support it.

As for rdoq-level 1, I find this to just be a rougher, less refined or accurate level 2.

If you can get your hands on any internal 'haich dee bits' x265 encodes for, ahem, research purposes then you'll find that they all use rdoq-level 2 with psy-rd 1.0 and psy-rdoq 1.0. Give or take anyway... I've seen psy-rd 1.2 + psy-rdoq 1.1 and psy-rd 1.4 + psy-rd 1.0 but always within those boundries.

They always consistently use no-cutree and usually qg-size 8 too. AQ mode 1 or 3 and often lots of zones as well. EDIT: and always CTU 32 and no-rskip.
can you provide the complete command line you prefer?
_kermit is offline   Reply With Quote
Old Today, 14:29   #7259  |  Link
RainyDog
Registered User
 
Join Date: May 2009
Posts: 170
Quote:
Originally Posted by _kermit View Post
can you provide the complete command line you prefer?
Sure, the below is always my starting point then I tweak from there if necessary with a dial or two down on --aq-strength and/or a notch or two up on --psy-rd. This is for a broad range of film sources at 1080p resolution.

Quote:
--preset slow --output-depth 10 --rd 3 --ctu 32 --limit-refs 1 --limit-tu 4 --tu-intra-depth 4 --tu-inter-depth 4 --no-rect --b-intra --aq-mode 1 --aq-strength 0.9 --ipratio 1.3 --pbratio 1.2 --no-cutree --subme 4 --merange 40 --max-merge 4 --weightb --bframes 10 --rc-lookahead 60 --lookahead-slices 0 --deblock -3:-3 --selective-sao 0 --no-sao --psy-rd 1.0
RainyDog is offline   Reply With Quote
Old Today, 15:41   #7260  |  Link
redbtn
Registered User
 
redbtn's Avatar
 
Join Date: Jan 2019
Location: Russia
Posts: 120
Quote:
Originally Posted by RainyDog View Post
They always consistently use no-cutree and usually qg-size 8 too. AQ mode 1 or 3 and often lots of zones as well. EDIT: and always CTU 32 and no-rskip.
I often see this also. But I don't understand why they use it. I did tests and noticed that CTU 32 and qg-size 8 is not better than CTU 64 qg-size 32 (and sometimes even worse). And also it seems to me that aq-mode 2 is better than 1. I'm not a professional, it's only my humble opinion.
Ps: I looked at your settings, do you sure that ctu 32 merange 40 is better for 1080p than ctu 64 merange 57?

Last edited by redbtn; Today at 15:44.
redbtn is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 16:13.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2019, vBulletin Solutions Inc.