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. |
28th February 2014, 07:28 | #23981 | Link | ||||
Registered User
Join Date: Sep 2013
Posts: 919
|
@Asmodian
I don't understand your post, but I really want to because its in my interest also. Do you meant you want to leave the VideoLUT that's loaded in the video card untouched, and have the 3DLUT do color correction? Quote:
Leaving VideoLUT untouched, while applying the 3DLUT. Quote:
Effect on what? VideoLUT loaded in the GPU? Quote:
Quote:
Its not very beneficial to anyone if you're the only one who can understand it... As I understand it: The 3DLUT should have an embedded Gamma/Calibration ramp (Linear, if you do NOT select "Apply Calibration" in the 3DLUT creator). "disable GPU gamma ramps" should disable the VideoLUT ramp. If Not selecting "disable GPU gamma ramps" and activating 3DLUT, you'll get two calibration ramps one on top of the other. You can always just Profile in ArgyllCMS without Calibrating. This will embed a totally flat linear calibration file into the resulting profile/3DLUT.
__________________
System: i7 3770K, GTX660, Win7 64bit, Panasonic ST60, Dell U2410. Last edited by James Freeman; 28th February 2014 at 07:47. |
||||
28th February 2014, 07:56 | #23982 | Link | |
Audiophile
Join Date: Oct 2006
Posts: 353
|
Quote:
w/e. My eyes hurt from staring at the monitor too close to compare these images. |
|
28th February 2014, 08:06 | #23983 | Link |
QB the Slayer
Join Date: Feb 2011
Location: Toronto
Posts: 697
|
I have a quick question... and staring at the screen for numerous settings has my eyes all fuzzy now... so I thought I would throw it out there for the experts to chime in. I push my R9 270x to max if possible and I have come to a fine hair's edge and can't decide the best of two scenarios:
ED (option 2) + Dynamic and use NNEDI3 x16 for Chroma upscaling OR Ordered Dithering and use NNEDI3 x32 for Chroma upscaling the difference between the two is so small, but it's the difference between a few dropped frames to none. Is the added level of NNEDI3 for Chroma upscaling worth the down-grade in dithering? QB
__________________
|
28th February 2014, 08:39 | #23984 | Link | |
Registered User
Join Date: Mar 2009
Posts: 3,650
|
Quote:
Last edited by ryrynz; 28th February 2014 at 08:51. |
|
28th February 2014, 09:00 | #23985 | Link | ||||
Registered User
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,407
|
Quote:
Quote:
Quote:
-a file.cal Apply calibration curves to link output and append linear -H file.cal Append calibration curves to 3dlut Both of these situations are correct but if you want to use madVR's 3DLUT instead of the video cards LUT you have to have Windows also use linear gamma ramps when madVR is running. I want to be able to use "-a" but still have Windows calibrated all the time. edit: I can do this if I un-append the linear curves (hexedit) and use madVR's "disable GPU gamma ramps". Quote:
I hope I made more sense this time. Last edited by Asmodian; 28th February 2014 at 09:36. |
||||
28th February 2014, 09:21 | #23986 | Link | |
Registered User
Join Date: Dec 2013
Posts: 753
|
Quote:
|
|
28th February 2014, 10:42 | #23987 | Link | ||||||||||||||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Not right now. Let's keep all the options for a while and then after a couple of weeks or so we can revisit who likes which algorithm the most and which algorithm might not be used at all, anymore, and then maybe drop some options based on that...
Quote:
Quote:
Quote:
Quote:
Quote:
Quote:
Quote:
Quote:
Quote:
http://madshi.net/madVRfinalDither4.rar Quote:
http://forum.doom9.org/showpost.php?p=1670818&postcount=23921 Quote:
Scientifically, your eyes should be able to differentiate about 11-12 bits of brightness information in a smooth gray ramp. So if you set madVR to 7 bit without dithering, you're throwing away TONS of potential brightness steps for no good reason. Maybe current content is often not good enough to show a clear benefit, but as I said, that depends on the exact movie/scene. And we'll hopefully soon get 4K content with higher native bitdepths. Quote:
Haha! Quote:
And if you perform the ArgyllCMS calibration with linear ramps loaded into the GPU, then the GPU ramps also have to be linear when playing the video, otherwise you'll get incorrect colors. Am I missing something? Quote:
So to answer your question, I fear you will need to try both configurations you suggested yourself and check which one your eyes/brain prefer. Which dithering algorithm looks best might also be dependent on your display. And whether more neurons help will also depend on the movie/scene. So I really wouldn't know what to recommend to you... ------- So here's another "final" build which fixes the green tint when using "colored" ordered dithering. The fix may also modify/improve mono colored ordered dithering ever so slightly compared to the previous "final" builds: http://madshi.net/madVRfinalDither4.rar |
||||||||||||||
28th February 2014, 11:03 | #23988 | Link |
Registered User
Join Date: Sep 2010
Posts: 35
|
madshi,
Hello, I think I have an audio noise lag related to MadVR - https://trac.mpc-hc.org/ticket/4076 MadVR somehow forces audio to be in perfect sync with the video and so makes it create a noise when it can't be perfectly synced due to CPU overuse (it starts at player using 50-60% CPU), I tried old madVR versions - 09 and 020 and those don't seem to have this (because as I lowered priority of the MPC HC player and made video lag forcefully (cuz it didn't want to - used way less resources than madvr does now) - audio desynced but didn't make any noise, (from version madVR040.zip and above the noise is there as I tested) so could you please (please please please) make an option to turn off perfect audio syncing to video so it could stop making that noise? |
28th February 2014, 11:14 | #23990 | Link | |
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
From a technical point of view there is no way to "disable the GPU gamma ramps". Basically they are always used. So disabling them practically means filling them with linear values. So disable=linear. |
|
28th February 2014, 11:23 | #23991 | Link | |
Registered User
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,407
|
Quote:
If I use a linear ramp attached to the 3DLUT the white point in Windows changes too. The video in MadVR looks the same either way (global linear or just "disable the GPU gamma ramps" checked). Is this not what you are expecting with that option? It doesn't load linear ramps into Windows for me (Win 8.1). Last edited by Asmodian; 28th February 2014 at 11:29. |
|
28th February 2014, 11:27 | #23992 | Link | ||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Quote:
My best guess is that x264 uses buggy math to convert the video from RGB to (fullrange) YCbCr. But of course this is only a wild guess, nothing else. I did see (and report) bugs in x264's color conversion code in the past already, though. So this being x264's fault would not come as a surprise to me. The only way this could possibly be madVR's fault would be if madVR generally would handle all YCbCr fullrange input incorrectly. I think somebody would have noticed that already, if that were the case? Anyway, anybody willing to put this to a test? I've no experience with encoding stuff, so I'm probably not the right person to do the testing/experimentation... We're just trying to play the encode with the same colors as the original file. Since the encoding process is very unlikely to change the primaries, I don't think we have to play with different primaries in madVR. The color encoding/decoding matrix and the levels (limited range vs full range) is probably where it's at. |
||
28th February 2014, 11:33 | #23993 | Link | |
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
This is an extremely complex topic because so many variables are involved. Right now I'm pretty confused about what you were/are doing, and why you manually modifed the 3dlut file etc etc. Makes all no sense to me right now. Have you read and understood what I wrote about the ramps having to be linear=disabled while doing ArgyllCMS calibration? Is that the way you were doing the measurements/calibration? |
|
28th February 2014, 11:38 | #23995 | Link | |
Registered User
Join Date: Sep 2010
Posts: 35
|
Quote:
so it's the LAV video decoder \ splitter which does this? LAV has some if\else conditions for MadVR version then to know when to make the noise and when not to? Last edited by Owyn; 28th February 2014 at 11:57. |
|
28th February 2014, 11:54 | #23996 | Link | ||
Registered User
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,407
|
Quote:
I do start calibration with linear ramps, dispwin -c. I then profile with linear ramps loaded but using dispread -K: -K file.cal Apply calibration file to test values while reading I then create the 3DLUT with collink -a: -a file.cal Apply calibration curves to link output and append linear I then have a calibration which looks great with linear ramps loaded and very odd with my "normal" ramps from dispcal loaded. I want linear ramps for madVR and my calibrated ramps in Windows. I had thought this was impossible too... but it seems to work. Quote:
If it only works in this exact situation I will be happy with my hack and you can ignore me. |
||
28th February 2014, 11:55 | #23997 | Link | |
Registered User
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,407
|
Quote:
|
|
28th February 2014, 11:57 | #23998 | Link | ||
QB the Slayer
Join Date: Feb 2011
Location: Toronto
Posts: 697
|
Quote:
Quote:
BTW... ED option 2 + Mono Dynamic is awesome! QB
__________________
Last edited by QBhd; 28th February 2014 at 12:02. |
||
28th February 2014, 12:01 | #24000 | Link | |||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
http://forum.doom9.org/showthread.php?t=161915 I hope that after reading that thread you'll have more trust in madVR than in swscale... Quote:
Quote:
One thing in favor of ED is that *IF* your eyes can see a difference at all with ED, then you should see that in most movies and most scenes. While improved chroma upscaling will only be visible in specific scenes. If your eyes do generally prefer ED over ordered dithering, then I'd recommend going that way. |
|||
Tags |
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling |
|
|