madVR - high quality video renderer (GPU assisted)

[baii]

In previous version, I got dark image when try 64bit with old setting, it is fixed when I do a setting reset then put back in the old setting. Can probably try the reset bat in madvr folder and see if that helps.

Sent from my 306SH

[Shiandow]

1) It's twice slower than "high" preset (or any other without gradient thing). On laptop with Intel HD 3000 I could use old debanding on 1080p video (downscaled to 768p with CR) but not new one (no/old/new - 30/38/46 ms).
2) I can't get better results on anime sources (destroyed by p̶r̶o̶f̶e̶s̶s̶i̶o̶n̶a̶l̶ ̶e̶n̶c̶o̶d̶e̶r̶s Crunchyroll) like this ones: 1 (https://dl.dropboxusercontent.com/u/16254258/android/viavr/rendering/debanding/akame.y4m), 2 (https://dl.dropboxusercontent.com/u/16254258/android/viavr/rendering/debanding/terraformars.y4m), 3 (https://dl.dropboxusercontent.com/u/16254258/android/viavr/rendering/debanding/wixoss.y4m).
Threshold >= 1.0 causes nasty artifacts near lines (on akame even 0.6 is enough to get them) and with detail level set to 1 there's still more banding.

It seems the part that's causing problems near lines may have become unnecessary after some of the other changes I've made, disabling that part might improve things while being slightly faster. I'll need to look carefully though, since it was also one of the things preventing the colours from shifting.

And I see you share my opinion of Crunchyroll's encodings...

[iDoodle]

Do I need to keep the rendering time under the vsync interval if I use smooth motion? Or is the movie frame interval the limit?

[huhn]

Do I need to keep the rendering time under the vsync interval if I use smooth motion? Or is the movie frame interval the limit?

get 20 % lower than your movie frame rate and you should be save.

and most important only use things you >see< a difference.

[luk008]

I've said that the Shiandow's algorithm looked a little better, but apparently it was just my first impression (need more comparisons to confirm).
I've used the first episode of Akatsuki no Yona from Crunchyroll (1080p) to compare the banding algorithm's. See what I got:

All artifact removal options disabled -> http://i.cubeupload.com/tJMvO1.png
"Old" reduce banding artifacts (high, high) -> http://i.cubeupload.com/tyV3ao.png
Shiandow's (detail = 2, ts = 0.5) -> http://i.cubeupload.com/YznuZV.png
Shiandow's (detail = 3, ts = 0.5) -> http://i.cubeupload.com/PFc4D2.png
Shiandow's (detail = 4, ts = 0.5) -> http://i.cubeupload.com/5r4kAd.png
Shiandow's (detail = 8, ts = 0.5) -> http://i.cubeupload.com/eMZKXo.png

PS: I've used the default madvr settings.
PS2: Higher threshold just make more artifacts to show.

[huhn]

I've said that the Shiandow's algorithm looked a little better, but apparently it was just my first impression (need more comparisons to confirm).
I've used the first episode of Akatsuki no Yona from Crunchyroll (1080p) to compare the banding algorithm's. See what I got:

All artifact removal options disabled -> http://i.cubeupload.com/tJMvO1.png
"Old" reduce banding artifacts (high, high) -> http://i.cubeupload.com/tyV3ao.png
Shiandow's (detail = 2, ts = 0.5) -> http://i.cubeupload.com/YznuZV.png
Shiandow's (detail = 3, ts = 0.5) -> http://i.cubeupload.com/PFc4D2.png
Shiandow's (detail = 4, ts = 0.5) -> http://i.cubeupload.com/5r4kAd.png
Shiandow's (detail = 8, ts = 0.5) -> http://i.cubeupload.com/eMZKXo.png

PS: I've used the default madvr settings.
PS2: Higher threshold just make more artifacts to show.

are you sure uploaded the right screens all of shiadow deband filter look like no debanding at all.

i took the original untouched PNG and applied his debanding filter with 1.0 2 and here is the result: http://abload.de/img/madvrshiandowdeband1.dxsq2.png
madVR medium does a pretty good job on this screen the colors look a little bit better: http://abload.de/img/madvrdebandmediumapq5a.png

i took the 3D LUT out but i forget the custom levels so you can't compare these screen with yours sorry.

[JPulowski]

isn't "refine the image only once after upscaling is complete" post resize?Yes, but it does not apply to downscaled videos and the ones match the monitor's native resolution. (e.g. 1080p -> 720p, 1080p on 1080p (no resize))

[cyberbeing]

Render Queue doesn't fill beyond 6 (or less if present in advance frames <14) when using DX11 Windowed Mode. FSE is normal.

GTX 770
350.12
madVR 0.88.1 (x86 & x64)
Win7 SP1 x64

[aufkrawall]

I recently tested with a news broadcast video, and indeed, madVRs deband filter did better better.
To achieve a similar effect with Shiandow's deband filter, I need to higher the threshold value which leads to artifacts in motion like some kind of flickering.
The vid:
http://download.media.tagesschau.de/video/2015/0508/TV-20150508-2155-5701.webxl.h264.mp4

[luk008]

are you sure uploaded the right screens all of shiadow deband filter look like no debanding at all.

i took the original untouched PNG and applied his debanding filter with 1.0 2 and here is the result: http://abload.de/img/madvrshiandowdeband1.dxsq2.png
madVR medium does a pretty good job on this screen the colors look a little bit better: http://abload.de/img/madvrdebandmediumapq5a.png

i took the 3D LUT out but i forget the custom levels so you can't compare these screen with yours sorry.

I'm sure that I've used the right options and uploaded them right. Why this? :confused:

In your screenshot Shiandow's looks good.

[huhn]

I'm sure that I've used the right options and uploaded them right. Why this? :confused:

In your screenshot Shiandow's looks good.

how did you took the screens? if i'm not mistaken you have to use control print to get a debanded screen.

[luk008]

how did you took the screens? if i'm not mistaken you have to use control print to get a debanded screen.

In MPC-HC, File > Save Image

I've just tested Shiandow's again and now it looks as nice as the old algorithm. Don't know what happened :scared:

EDIT: Anyway, seems that I've done something wrong. Sorry for the mistake. I'll be more careful next time.

[huhn]

Yes, but it does not apply to downscaled videos and the ones match the monitor's native resolution. (e.g. 1080p -> 720p, 1080p on 1080p (no resize))

so the "image enhancements" is pre resizing. ok i see.
not sure about pre resize sharpening. shouldn't post resize sharpening be better in most cases?

[Anime Viewer]

I would like to start with evaluating the new "DEBAND" algorithm written by Shiandow. Please could you guys try it with your various sources and let me know how it compares to madVR's other debanding algorithms in terms of quality and performance? Based on your feedback I will either remove the new algorithm again, or I will replace one or multiple of the old options with the new one. I want to end up with only one Deband option with max 3 different settings. Your feedback is very welcome - thanks!

Configuration wise I like Shiandow's debanding better. The old debanding feels too limiting with only low, mid, and high options to choose from. While simpler it is not as versatile as Shiandow's.

On an unrelated note to the debanding evaluation it seems enabling the upscaling refinement in scaling algorithms seems to crash the player/video every time I expand the window to full screen. Image enhancements seem to work well on the videos I've watched.
For reference I'm using 0.88.1

[huhn]

Configuration wise I like Shiandow's debanding better. The old debanding feels too limiting with only low, mid, and high options to choose from. While simpler it is not as versatile as Shiandow's.

i agree. i personally like advanced settings more but if i look at the bigger picture i see the issue with that.

maybe a slide bar for debanding straight. but i guess the "old" version is to specialized for this. i guess it is possible with shiandow's versions i don't see any benefits from higher detail levels.

but this is only my observation.

[andybkma]

Since I know you are looking for feedback, on my Win7 64-bit OS but using 32-bit Zoom Player and 32-bit LAV, setting "Use Direct 3D 11 for presentation" immediately created jumping, repeating video.

BTW, thanks for all the new image enhancement features, been using avisynth LSFMod script as a post processor, now to play and see if I can replace that with mVR's internal FineSharp

[huhn]

is it trival to add nedi as a chroma resizer?

i know you removed chroma doubling in this build and using so "hard" resizer for chroma is a waste most of the time. but nedi is kind of cheap or does it perform bad for chroma?

[tobindac]

Nice features madshi. I bet the tv aficionados will mainly love the 10bit feature.

I bet though most are incapable of telling a difference above 6 or 7bit.

I can't even imagine what 16bit monitors are for. Probably for labs.

[tobindac]

My lowly IPS monitor (that almost certainly can't handle 10bit) does seem to get the 10bit signal because the screen goes blank for 1.5secs before going fullscreen with that setting on. So I guess the feature does something to it. The image appears identical, at least in window mode where I can test any obvious differences on the fly.

[FireFreak111]

With D3D11 Presentation path, Interop times pop up when I start the video. ~22ms, apparently spiking to insane figures like 2883597595ms yet not actually causing any effect on the motion of the video. This happens often when I open the video, disappearing 30secs to 2mins in, back to just render/present. Windowed overlay enabled, Windows 10.

I figured out this shows when I enabled SuperRes post-NEDI resize. It can sit at 80ms, with no dropped frames, and no glitches. Doesn't show in the Max Stats either, only in the average stats.

[James Freeman]

* Enabling D3D11 disables Overlay mode.

* D3D11, FSE, no glitches at all (W7 64).

* Is there any indication that 10-bit is used at output?
To enable 10bit I need FSE and D3D11 right?
Should I change "the native display bitdepth is:" to 10bit?

As already suggested by others;
Since madVR is not for beginners in itself, it would be ideal to leave all the options under Advanced tab and settle for 3 (Low,Mid,High) options in the Basic tab.

[cvrkuth]

With Direct3D 11 presentation path enabled I have display switching issue again, now in madVR32 too (There was issue in the 32 bit version, v0.86.10. Madshi then do workaround and issue was gone).
I update my OS (Windows 8.1 Pro), I have same problem again.
I noticed that when playing 1080p24 (mkv, m2ts, ts .. full screen exclusive mode, default settings) my screen goes to 23hz. Because of this, there is a drop frames every 30-40 sec.

If I uncheck 'Present Several frames in advance "(old path), my screen stays in 24hz.
Not any more

[baii]

Windowed any bit, 1080i file, I get presentation glitch at certain interval with either D9 or D11, but D11 will get 1 drop frame at the glitch(unnoticeable for me~) where D9 doesn't.

FSE, any setting works on me. If set too high for present frames in advance, video play fine after enter FSE but will cause crash when exit from full screen, where old version used to be black screen with sound when going into FSE.

Have a native 10 bit and a 8bit+FRC screen, cant really notice a difference tbh~ And FSE 10bit cant be switch on the fly so can't compare ~

[cca]

With Direct3D 11 presentation path enabled I have display switching issue again, now in madVR32 too (There was issue in the 32 bit version, v0.86.10. Madshi then do workaround and issue was gone).



Not any more

Same issue noticed here, when switching to FSE the refresh rate is not consistent. Using an Radeon 5850, Win 8.1.

Another issue I noticed, in FSE mode (both DX9 and DX11) the seek bar on the bottom is not clickable, I cannot seek using it, if I try to click it the video just pauses like if I had clicked anywhere else on the image. Player used is MPC-HC 1.7.8 32 bit.

EDIT: the seek bar issue seems to have been fixed by updating to the latest MPC-HC nightly.

Also to add, I have not encountered any of the issues reported by other so far in my setup.

[XMonarchY]

no don't do that.
use a rec 709 madVR does the rest. of cause you can add the 601 with the bt 709 but don't put a 601 in the 709 slot!

Hmm... To be certain, I place only 1 Rec.601 3DLUT in Rec.709 line and it works fine, even for SMPTE C content that is actually labeled Rec.601/SMPTE C. The Rec.601 line stays empty... Kind of strange that madVR uses Rec.709 line even with Rec.601 3DLUT to render EITHER Rec.709 or Rec.601 content. What is the point of 3DLUT line labels then???

If this feature was properly integrated, madVR would not be using any 3DLUT from Rec.709 line to render content with Rec.601 label... Then again, not all content is labeled properly. I think there should be no 3DLUT line labels, just generic lines. Users should manually select the right 3DLUT to go into those generic lines based on content they plan on watching. Leaving it to madVR creates more confusion, considering that madVR will use whichever 3DLUT, Rec.601 or 709, in Rec.709 line to render both Rec.709 and 601 content, with or without labels. Its a complicated MESS IMHO.

[omarank]

Wow, so many things to try out in the new version. Thanks!

I was wondering you didn’t provide 12 bit and 16 bit support, just wondering not that I need them. If a display processes internally in say 12 bit and then dither the output to 8 bit, it may be better to feed it 12 bit than 10 bit.

In my testing I found madVR’s old debanding algorithm producing overall better looking images than the new debanding algorithm, irrespective of the parameters and the amount of banding removed.

Seekbar doesn’t appear in FSE mode D3D11 presentation after a couple of seeks. I am using Windows 8.1 and it happens with x86 madVR build only.

[nevcairiel]

Yes, but it does not apply to downscaled videos and the ones match the monitor's native resolution. (e.g. 1080p -> 720p, 1080p on 1080p (no resize))

processing -> image enhancements apply all the time, no matter of down/up scaling, or no scaling.

[XMonarchY]

Wow! Quality is a matter of preference, but those who agree that the best quality for previous madVR versions was:
Chroma Upscaling - NNEDI3
Image Doubling/Quadrupling (if needed for lower-resolution content) - NNEDI3
Image Upscaling - Jinc 3tap + AR
Image Downscaling - Catmull-Rom + AR + LL
Dithering - Error Diffusion,

What do you recommend as new setting for:
1. HQ content that uses same resolution as display's native resolution
2. LQ content that uses same resolution as display's native resolution
3. HQ content that uses resolution below that of display's native resolution (ASIDE from NNEDI3 Image Doubling/Quadrupling)
4. LQ content that uses resolution below that of display's native resolution (ASIDE from NNEDI3 Image Doubling/Quadrupling)

?????????

[XMonarchY]

Wooowzers! I had to go back to previous version. This new one, even with all the new features disabled, using the same exact values for settings I used in the older madVR, produces MASSIVE framerate drop. MASSIVE! I used the settings I described in the post above without adding the new Sharpen and De-band and refinement stuff. I do not know what is wrong... I know for a fact that I PROPERLY uninstalled the old version and reset this new one to DEFAULTs using the .bat as Admin. Using Exclusive mode or not using it makes 0 difference.

OK, more detail. This test was for Mid/Low-ish Quality content that is 720x480 on my 1080p TV, using NNEDI3 64n for Chroma Upscaling, and NNEDI3 32n for Image Doubling/Quadrupling effect - both set to Always. I turned off Quadrupling and framerate went back to normal. HOWEVER, in previous version, I was able to use NNEDI3 32n for Image Luma AND Chroma Doubling AND Quadrupling running this exact 720x480 content. Why can't I use NNEDI3 32n Quadrupling in this build without massive framerate drop? After disabling Quadrupling, I enabled Upscaling Refinement - LumaSharpen ONLY with default values, Shiandow's Deband Algorithm with default values, and Image Enhancement - LumaSharpen ONLY with default values. New features do not seem to make the image as nice as using Chroma and Luma Image Quadrupling in the older version of madVR.


In the older version, I was even able to use NNEDI3 128n for Chroma Upscaling + NNEDI3 32n for Chroma+Luma Doubling/Quadrupling. Now I get framerate drop if I use NNEDI3 128n for Chroma Upscaling + NNEDI3 32n for Image Doubling. I have to use NNEDI3 64n for Chroma Upscaling if I want no framerate drops.

[Qaq]

Nice features madshi. I bet the tv aficionados will mainly love the 10bit feature.
I can't even imagine what 16bit monitors are for. Probably for labs.Panel may be 8 or 10 bit, but picture always processed in digital video processor first. In what bitdepth? Say, 10-16 bits.Thus its best to transport picture in highest bitdepth DVP can accept to get less roundings.

BTW. Whats the point to use dithering in madVR if TV dithers picture at the last stage again anyway?

@madshi
Tried 088. Windows 7 x86, 7750, Catalyst 13.12.

DX11 works fine, except picture blinks on 1080 videos in windowed (downscaled) mode. Not a big deal for me though.

10 bit worked fine with 2-3 attempts. Can't say I saw any difference with 8 bit picture. On last attempt I switched video decoder SW > DXVA2 native, started 1080 video (in windowed mode as always) and got static pattern with vertical lines. Couldn't get rid of that no matter what. Player was working in background, I heard the sound from a movie. Had to reboot the system, switched madVR to 8 bit on 720 video in windowed mode and didn't try 10 bits anymore. Dithering was ED1, if that matters.

[ikakun]

My experience with the new madVR version(0.88.1):

With D3D11: No problem with any queues. No tearing. Only gets presentation glitches when opening a video, switching between fullscreen/window, & on seeking.

D3D11 + SM + FSE + ED 1 or 2 (with or without options ticked) = Plays fine but gets into a problem when the video is paused. It shows only black screen (when Ctrl-J OSD is on, it disappears too) while the sound continue playing. It gets fine when exiting FSE.

D3D11 + SM + FS windowed mode (new path) + any dithering option = anythings fine.

9 to 10 bit = Like others reported, video gets dark in non-FSE mode with ED1/2.

EDIT:

with D3D11:
Turning on D3D11 and playing in Windowed Mode causes any video to repeat only a few frames (like a really short GIF, probably whatever frames is set to "pressented in advance") while the audio continues in window mode.

Happens to me too, but it rarely does.

[tyto]

I dont ever know if to update madVR I just need to delete all the files and then copy the new ones, or also click install.bat, or also uninstall before.

So far I delete everything, copy and click install.bat, but Im not sure if that's how you are supposed to do.

EDIT:
I wanted to do some comparisons, but Im not sure that if I take a screenshot with mpc-hc what I get its the "final" image (since its not even the right resolution)

EDIT 2: grey-dark problems
here's my feedback about that.
If I set 10bit, the only way to avoid "greyness" is to use D3D11 (and even with that, it only works in fullscreen)

EDIT 3: D3D11 PROBLEMS
with d3d11 I got presentation glitches like no tomorrow

[yok833]

Just a quick question... Do we still need to use error diffusion or any dithering when we are in 10 bit output??
Same question with debanding?? ... Is it still necessary?

[nevcairiel]

Just a quick question... Do we still need to use error diffusion or any dithering when we are in 10 bit output??
Same question with debanding?? ... Is it still necessary?

Use your eyes, decide what looks best. It greatly depends on what your TV does with the image!

[tyto]

if my display support 4:4:4, should I set lav video to only output to 4:4:4? or just leave all selected?

[ashlar42]

madVR v0.88.0 released
5) Please note that currently there are tons and TONS of options. This definitely is overkill and needs to be cleaned. This is a task we have to work on together. For each algorithm I'd like to define max 3 options (e.g. low, mid, high) and that must be sufficient. Otherwise the sheer amount of options will shy new users away.
First of all: thank you for the continued development on madVR. It's been such an exciting run so far. :)

Regarding point 5, I am writing to humbly suggest to keep the options available, maybe as an xml file to be configured manually, for those that really like to tinker with their configurations (similar to what Kodi does with advancedsettings.xml). My two cents, nothing more.

[feisty2]

I can't even imagine what 16bit monitors are for. Probably for labs.

U sure such monitor does exist and its out there in some lab?

[James Freeman]

can anyone please answer this:

* Is there any indication that 10-bit is used at output?
To enable 10bit I need FSE and D3D11 right?
Should I change "the native display bitdepth is:" to 10bit?

[huhn]

* Enabling D3D11 disables Overlay mode.

* D3D11, FSE, no glitches at all (W7 64).

* Is there any indication that 10-bit is used at output?
To enable 10bit I need FSE and D3D11 right?
Should I change "the native display bitdepth is:" to 10bit?

As already suggested by others;
Since madVR is not for beginners in itself, it would be ideal to leave all the options under Advanced tab and settle for 3 (Low,Mid,High) options in the Basic tab.
can anyone please answer this:

* Is there any indication that 10-bit is used at output?
To enable 10bit I need FSE and D3D11 right?
Should I change "the native display bitdepth is:" to 10bit?

10 bit output needs FSE d11.
you have set it under your device properties.
in FSE d11 you will see a line in the OSD like ~ fullscreen exclusive mode, 10 bit

Hmm... To be certain, I place only 1 Rec.601 3DLUT in Rec.709 line and it works fine, even for SMPTE C content that is actually labeled Rec.601/SMPTE C. The Rec.601 line stays empty... Kind of strange that madVR uses Rec.709 line even with Rec.601 3DLUT to render EITHER Rec.709 or Rec.601 content. What is the point of 3DLUT line labels then???

If this feature was properly integrated, madVR would not be using any 3DLUT from Rec.709 line to render content with Rec.601 label... Then again, not all content is labeled properly. I think there should be no 3DLUT line labels, just generic lines. Users should manually select the right 3DLUT to go into those generic lines based on content they plan on watching. Leaving it to madVR creates more confusion, considering that madVR will use whichever 3DLUT, Rec.601 or 709, in Rec.709 line to render both Rec.709 and 601 content, with or without labels. Its a complicated MESS IMHO.

ok i try it again.
don't use a smpte c 3d LUT in bt 709 line you will get wrong colors.
use a smpte c 3d LUT in smpte C line. i know that is broken right now so you can't do that. are you creating a bug entry in the bug tracker? i can do that just say so.
so in the meanwhile you have to use a bt 709 3D LUT. just take your measuring and create a bt 709 3D LUT and add this in bt 709 so everything should work fine. you can addionally add a smpte c 3D LUT now. but this is not needed madVR will make sure the rest is ok.
if my display support 4:4:4, should I set lav video to only output to 4:4:4? or just leave all selected?

leave lavfilter at defaults.

madVR is outputting RGB and RGB is kind of 4:4:4.

[madshi]

I get loads of presentation glitches in D3D11 FSE mode, on Windows 8.1, NVIDIA Titan X (350.12). On a multi-monitor setup if that maybe matters.
The number sky-rockets, almost like every frame (or every second) produces a glitch counter.
Does your refresh rate match the movie framerate? Can you test with a matching video, too? E.g. if your display can only do 60Hz, does the presentation glitch problem go away if you play a 60fps video (or a deinterlaced 60i video)?

W8.1 x64
It wasn't that bad (by the time I finished watching ~20min movie I had like 14-15 presentation glitches. Usually it's around 2-4). Could be because I stiched in and out of FSE more than usual.
Sounds ok. Presentation glitches can occur when switching between windowed <-> exclusive mode.

Also, just had a video playing on the background in D3D11 mode while browsing and it hanged after a few seconds. Pausing\unpausing resumed the sound, but the picture remained frozen.
Can you reproduce this reliably? If so, please provide me with a step-by-step guide to reproduce the issue. Thanks. Also, maybe a freeze report from this situation (Ctrl+Alt+Shift+Break/Pause) might help.

Pausing an XVID on DX11 made the video repeat a fraction of a second over and over again like one of those old VCRs. Resuming play didn't work either.
Can you reproduce this? Does it only occur with the XVID? Or also with other files?

I enabled D3D11 mode and get presentation glitches and dropped frames after these steps:

I switch from window to fullscreen exclusiv mode. Now when I jump in movie (by clicking on the bottom line) = that problem.

When I am in window mode and jump in movie = no problems. When I switch to fullscreen exclusive mode and don't jump in that mode, also no problems.
Do these glitches/drops only occur while seeking and then stop when the seek is complete? Or do these glitches/drops continue to add up all the time during playback once you've seeked in FSE mode? I mean do the glitches/drops increase with every second of video playback?

I doubt this, since D3D11 works well with MPDN in windowed mode.
That is no proof. Different software calls APIs in different ways. Just because there's no problem in games, or with a different renderer, doesn't mean that madVR must be at fault. It can still be the OS' or the GPU driver's fault. I know how MPDN calls the APIs but I don't like to call them like that. I believe my method is better. Maybe not in your specific case, though. However, I see zero tearing on any of my 3 PCs I've tested, and on none of the 3 Intel/NVidia/AMD GPUs I have in my development PC.

The "old" debanding algorithm seems to be better than the new one. To maintain the same level of details as the low preset, you need to use "treshold : 0.3" and "detail level :5" but in that case the debanding is weaker than the old one.
Thanks for the feedback. Ok, so the "new" algorithm can't compete with the old "low" algo, if you aim for max detail preservation. But how does the new algo compare to the old "high" algo? From what Shiandow has written so far I think that's the likeliest situation where the new algo could complete (and maybe beat) with the old one.

I'm also getting a tonne of presentation glitches in the DX11 path and my render times are fine.
Not enough information. Windowed mode or fullscreen windowed mode or FSE mode? All the time or only after doing something specific? Which GPU? Which OS? Come on, you can do better than that... :)

with D3D11:
Turning on D3D11 and playing in Windowed Mode causes any video to repeat only a few frames (like a really short GIF, probably whatever frames is set to "pressented in advance") while the audio continues in window mode. This applies to any bit depth (7-10 bit).
Switching to FSE, the video plays normally without repeated frame (short gif).
Was able to reproduce this on one of my PCs. Will try to find the cause of the issue.

Smooth Motion:
SM + any bit + D3D11 + Ordered Dithering:
Presentation Glitches increase by 22-26 per second if time skipped to a different point in video and continues to increase.
Initially playing video with SM+D3D11 doesn't cause any PG and will continue play smoothly. Until time is skipped.
I suppose this is in FSE mode? Or also in windowed mode?

Render Queue doesn't fill beyond 6 (or less if present in advance frames <14) when using DX11 Windowed Mode. FSE is normal.

GTX 770
350.12
madVR 0.88.1 (x86 & x64)
Win7 SP1 x64
Strange. You do have the Windows 7 Platform Update installed, right?

I recently tested with a news broadcast video, and indeed, madVRs deband filter did better better.
To achieve a similar effect with Shiandow's deband filter, I need to higher the threshold value which leads to artifacts in motion like some kind of flickering.
Ok, thanks for the feedback. I suppose you compared to the old "high" setting, right?

it seems enabling the upscaling refinement in scaling algorithms seems to crash the player/video every time I expand the window to full screen. Image enhancements seem to work well on the videos I've watched.
For reference I'm using 0.88.1
Don't have that problem here. Nobody else reported it, either. Can you reproduce it? If so, is it any of the refinement options? Or one specific?

Configuration wise I like Shiandow's debanding better. The old debanding feels too limiting with only low, mid, and high options to choose from. While simpler it is not as versatile as Shiandow's.
Ok, but I'm much *much* more interested in image quality and performance comparisons. How much configuration control you get is a totally separate issue. If I keep Shiandow's deband algorithm, it will probably replace one of the 3 old presets. The current controls for the Shiandow deband algorithm will go away. So please compare primarily image quality and with secondary priority performance. Thanks!

For discussion about configuration topics, see below:

Actually, I would suggest using the popular Basic/Advanced options paradigm. In other words, have few options shown by default ("Basic") and then have an Advanced button to click to show all options.
Second this, as madVR's flexibility is one of its main advantages. The ability to fine-tune almost everything is crucial for such a feature-rich renderer that is used in many different environments and scenarios.
Regarding point 5, I am writing to humbly suggest to keep the options available, maybe as an xml file to be configured manually, for those that really like to tinker with their configurations (similar to what Kodi does with advancedsettings.xml). My two cents, nothing more.
As already suggested by others;
Since madVR is not for beginners in itself, it would be ideal to leave all the options under Advanced tab and settle for 3 (Low,Mid,High) options in the Basic tab.
I hear you. What you're asking for is not so easy to implement, though. For most users (I'd say 99% or more) it's much better to have a "low/mid/high" switch for every important algorithm then to have a dozen different knobs and dials. Most users wouldn't know what to do with so many options and would be overwhelmed. I understand that some advanced users like you would like to have more control. But let's be honest here:

How many of you are using the "debanding custom settings - toggle" keyboard shortcut which madVR currently offers, in order to fine tune those 5 or 6 different debanding parameters? Seriously, does *anybody* use it? I doubt it's more than 5 out of all madVR users.

One thing I could imagine is adding an advanced settings page somewhere where all the hundreds of different options can be fine tuned. However, it's not really easy, because the main dialog will continue to have low/mid/high settings. So would the advanced options have to allow you to redefine what low/mid/high means exactly? That would increase the number of options by a factor of 3, because every option would have to be available for low, mid and high presets. <sigh>

Anyway, something like this is not coming any time soon. It's too time consuming. For now I want to find proper simple defaults like low,mid,high and I *will* remove the other options. Sorry. I have to set priorities on the benefits of the majority of users. And the majority wants an easy to understand configuration dialog.

is it trival to add nedi as a chroma resizer?

i know you removed chroma doubling in this build and using so "hard" resizer for chroma is a waste most of the time. but nedi is kind of cheap or does it perform bad for chroma?
Are you talking about chroma upsampling (4:2:0 -> 4:4:4), or are you using about chroma upsampling, when using NNEDI3 to double luma? In the first case, yes, NEDI might be an option to consider. But in the latter case I believe the benefit is too small to consider the complications that would mean for my code.

When talking about 4:2:0 -> 4:4:4 chroma upscaling, I think the SuperRes filter has more potential than using NEDI. I'm not happy with the current SuperRes default for chroma upsampling, though. But this is too early to discuss that. I want to concentrate on debanding for now.

With D3D11 Presentation path, Interop times pop up when I start the video. ~22ms, apparently spiking to insane figures like 2883597595ms yet not actually causing any effect on the motion of the video. This happens often when I open the video, disappearing 30secs to 2mins in, back to just render/present. Windowed overlay enabled, Windows 10.

I figured out this shows when I enabled SuperRes post-NEDI resize. It can sit at 80ms, with no dropped frames, and no glitches. Doesn't show in the Max Stats either, only in the average stats.
Can't seem to reproduce that here.

I bet though most are incapable of telling a difference above 6 or 7bit.
I guess so, too. But it's hard to judge, anyway, because it will differ from display to display.

[vivan]

Thanks. So you prefer the "old" algorithm?In the current state - yes.

Configuration wise I like Shiandow's debanding better. The old debanding feels too limiting with only low, mid, and high options to choose from. While simpler it is not as versatile as Shiandow's.i agree. i personally like advanced settings more but if i look at the bigger picture i see the issue with that.There're several options that you could tweak in custom mode (ctrl+alt+shift+d) - avgDif/midDif/maxDif, angleBoost/maxAngle. Old shortcuts were ctrl+alt+1..0 but know they don't work anymore and I don't know what new ones are.

[madshi]

* Is there any indication that 10-bit is used at output?
To enable 10bit I need FSE and D3D11 right?
Should I change "the native display bitdepth is:" to 10bit?
When using FSE, the debug OSD should report the bitdepth.

What you should do or not I can't say. Depends on your display. Let your eyes be the judge.

I was wondering you didn’t provide 12 bit and 16 bit support, just wondering not that I need them. If a display processes internally in say 12 bit and then dither the output to 8 bit, it may be better to feed it 12 bit than 10 bit.
Direct3D doesn't support 12bit output. 16bit is supported, but in a weird way and different GPUs produce different results. Rather weird results, sometimes. So I decided not to support 16bit output for now.

Feeding the highest possible bitdepth to the display may make sense, but only if the HTPC/GPU output is really proper (and I have my doubts about that with 16bit at the moment), and only if the display handles high bitdepth input in a good way, and only if the display has a decent dithering algorithm itself. So many ifs and maybes. So once again, nobody can tell you how to configure bitdepth, because it all depends on your display, and nobody sees it but you.

In my testing I found madVR’s old debanding algorithm producing overall better looking images than the new debanding algorithm, irrespective of the parameters and the amount of banding removed.
Ok, thanks for the feedback.

Seekbar doesn’t appear in FSE mode D3D11 presentation after a couple of seeks. I am using Windows 8.1 and it happens with x86 madVR build only.
Can't reproduce that here. It just doesn't show? And this problem did not occur with madVR v0.87.21?

Wooowzers! I had to go back to previous version. This new one, even with all the new features disabled, using the same exact values for settings I used in the older madVR, produces MASSIVE framerate drop. MASSIVE! I used the settings I described in the post above without adding the new Sharpen and De-band and refinement stuff. I do not know what is wrong... I know for a fact that I PROPERLY uninstalled the old version and reset this new one to DEFAULTs using the .bat as Admin. Using Exclusive mode or not using it makes 0 difference.

OK, more detail. This test was for Mid/Low-ish Quality content that is 720x480 on my 1080p TV, using NNEDI3 64n for Chroma Upscaling, and NNEDI3 32n for Image Doubling/Quadrupling effect - both set to Always. I turned off Quadrupling and framerate went back to normal. HOWEVER, in previous version, I was able to use NNEDI3 32n for Image Luma AND Chroma Doubling AND Quadrupling running this exact 720x480 content. Why can't I use NNEDI3 32n Quadrupling in this build without massive framerate drop? After disabling Quadrupling, I enabled Upscaling Refinement - LumaSharpen ONLY with default values, Shiandow's Deband Algorithm with default values, and Image Enhancement - LumaSharpen ONLY with default values. New features do not seem to make the image as nice as using Chroma and Luma Image Quadrupling in the older version of madVR.

In the older version, I was even able to use NNEDI3 128n for Chroma Upscaling + NNEDI3 32n for Chroma+Luma Doubling/Quadrupling. Now I get framerate drop if I use NNEDI3 128n for Chroma Upscaling + NNEDI3 32n for Image Doubling. I have to use NNEDI3 64n for Chroma Upscaling if I want no framerate drops.
Which image upscaling algorithm do you have selected? Is it something crazy like Jinc8 AR maybe? Try lowering that to something reasonable, like e.g. Lanczos3 AR. That algorithm doesn't need to be so fancy if you're using image doubling, anyway. And doing that will greatly reduce GPU consumption.

Panel may be 8 or 10 bit, but picture always processed in digital video processor first. In what bitdepth? Say, 10-16 bits.Thus its best to transport picture in highest bitdepth DVP can accept to get less roundings.
In theory yes. Provided that the display actually does high-bitdepth processing and dithers itself down with a decent algorithms. Which is not guaranteed. So in the end every user will have to test and judge based on what his eyes tell him.

BTW. Whats the point to use dithering in madVR if TV dithers picture at the last stage again anyway?
Dithering is mathematically necessary to avoid quantization noise, whenever bitdepth is reduced. Doing dithering in the display does not replace dithering in madVR, unless madVR outputs a very bitdepth to the display, *and* if the display properly handles that and dithers it down later properly. As written above, that's a lot of its and maybes.

@madshi
Tried 088. Windows 7 x86, 7750, Catalyst 13.12.

DX11 works fine, except picture blinks on 1080 videos in windowed (downscaled) mode. Not a big deal for me though.
Strange.

Just a quick question... Do we still need to use error diffusion or any dithering when we are in 10 bit output??
Same question with debanding?? ... Is it still necessary?
IHMO yes and yes. Although when using 10bit output, random dithering or ordered dithering should do.

yes d11 FSE. and i tried both 8 and 10 bit both have this issue.
not sure if it matters at all i'm on windows 10 x64 10074.
With Direct3D 11 presentation path enabled I have display switching issue again, now in madVR32 too (There was issue in the 32 bit version, v0.86.10. Madshi then do workaround and issue was gone).
<sigh> Will have to investigate that. Which is difficult because my computer monitors don't really support switching between different refresh rates... :(

Thanks for including LumaSharpen. I did some testing and here is some feedback:
[LIST]
Looks like currently you are using the old version of LumaSharpen, which is 1.4.1. Current version is 1.5.0, which is included with Reshade + SweetFX bundle. I have already modified it for MPC-HC use, you can find it here (http://pastebin.com/rxDjqmGg).
Current madVR implementation uses Pattern 2 (9 tap gaussian, 4+1 texture fetches), although it looks nice on video games it doesn't look that good when it comes to videos (http://forum.doom9.org/showthread.php?p=1673897#post1673897). In my opinion, it would be better if you'd include the other patterns as options.
I didn't know there was a version 1.5.0. In which way is it different to 1.4.1? I see that the experimental limiter is missing which 1.4.1 had.

I'm using pattern 3. I can't see any pattern in 1.5.0 which looks more suited to video than pattern 3? Which pattern would you use instead?

This is more of a personal request: I am normally used to MPC-HC's shader management. For instance I do always apply LumaSharpen, no matter if there is resizing or not. With madVR even though I can use profiles it gets complicated in this case. Because sometimes I watch videos downsized in windowed mode, which LumaSharpen does not apply unless it is pre-resize. And pre-resize and post-resize(upscale) applications of LumaSharpen is quite different from each other and requires additional fiddling. It would be nice if there would be an option to apply LumaSharpen always as post-resize.
I don't like the concept of always applying sharpening, regardless of the situation, maybe even with the same sharpening settings. That *cannot* produce good quality for all sources and situations.

There are soft sources, and sharp soures. Some sources need upscaling, some don't. Some are downscaled, some not. Are you saying that you really want to apply the same sharpening to all sources, after scaling, regardless of whether they're soft to start with, and regardless of whether they were upscaled or downscaled? Seriously?? Makes zero sense to me, to be honest.

I tried to use a different approach: If we presume that the source is nicely sharp, then we don't need to sharpen it any further, as long as we render the source in its original resolution, or as long as we downscale it. We do still need to sharpen the image after using upscaling, though, because if the upscaling factor is large enough, even a nicely sharp source becomes blurry, even when using NNEDI3. So if we presume that the source is nicely sharp, we only need sharpening after upscaling, and the amount of sharpening should depend on the upscaling factor, don't you agree?

Now of course not all sources are sharp. So if we have a soft source, we may want to sharpen it. That's what "image enhancements" are for. If the source is already very soft, it makes more sense to sharpen it right there, before scaling it, IMHO.

so the "image enhancements" is pre resizing. ok i see.
not sure about pre resize sharpening. shouldn't post resize sharpening be better in most cases?
Let me try to explain how sharpening algorithms work: They look at each pixel separately and modify it, based on the values of neighbor pixels. Now if you upscale an image with a very large upscaling factor, the upscaled image will be very soft, and for each source pixel there will be lots and lots of target pixels. If you run a sharpening algorithm on such an image, the sharpening algo will look at the neighbor pixels, but they all originate from the same source pixel in this situation! So sharpening an image after using a very large upscaling factor is not very effective.

Sharpening has the biggest effect if you apply is before upscaling. However, from a quality point of view, sharpening looks better if you apply it after upscaling, because then the sharpening can be done in a higher resolution. So when should we sharpen? We need to find the right place and time in the processing pipeline. That's why the latest madVR build now allows sharpening as an upscaling "refinement" step: The upscaling is split into several 2.0x upscaling steps, and sharpening is applied after each step. This is IMHO a good compromise between sharpening effectiveness and quality. Sharpening before upscaling looks ugly. Sharpening after upscaling with a large upscaling factor has barely any effect. So we sharpen after upscaling with a max 2.0x upscaling factor to get the best results.

So why do I offer image enhancements which is sharpening applied to the source? Well, if the source is very soft, it is a comparable situation to upscaling with a very large factor: The neighbor pixels in a very soft image are already very similar. So if you take such a soft image and upscale it further, after that a sharpening algorithm won't have much effect at all. So for very soft sources it's more useful to sharpen them before upscaling, otherwise sharpening will almost be without effect.

Well, at least that's my thoughts about sharpening. I'm willing to be convinced that my sharpening concept is not good, but I'll need good arguments and ideally screenshots etc for that.

Anyway, one again, let's not discuss everything at the same time, that will only become a big mess. Let's all first concentrate on debanding. Once we have the debanding topic covered, we can move on to the next topic.

[madshi]

There're several options that you could tweak in custom mode (ctrl+alt+shift+d) - avgDif/midDif/maxDif, angleBoost/maxAngle. Old shortcuts were ctrl+alt+1..0 but know they don't work anymore and I don't know what new ones are.
The new ones by default are empty, but you can assign any keys you like in the madVR settings.

[vivan]

The new ones by default are empty, but you can assign any keys you like in the madVR settings.Hm, I don't see any options for it there - only "debanding - toggle" and "debanding custom settings - toggle".
But by mashing the keyboard I figured out that new shortcuts are ctrl+shift+arrows on numpad :)

[huhn]

Are you talking about chroma upsampling (4:2:0 -> 4:4:4), or are you using about chroma upsampling, when using NNEDI3 to double luma? In the first case, yes, NEDI might be an option to consider. But in the latter case I believe the benefit is too small to consider the complications that would mean for my code.

When talking about 4:2:0 -> 4:4:4 chroma upscaling, I think the SuperRes filter has more potential than using NEDI. I'm not happy with the current SuperRes default for chroma upsampling, though. But this is too early to discuss that. I want to concentrate on debanding for now.

i mean 4:2:0 -> 4:4:4.

isn't SuperRes a lot slower?

i come back to this after debanding.

[James Freeman]

<sigh> Will have to investigate that. Which is difficult because my computer monitors don't really support switching between different refresh rates...
Man, you really need a good European TV with 23.976, 24, 25, 29.97, 30, 50, 59.94 and 60 Hz.

[nevcairiel]

Does your refresh rate match the movie framerate? Can you test with a matching video, too? E.g. if your display can only do 60Hz, does the presentation glitch problem go away if you play a 60fps video (or a deinterlaced 60i video)?

Matching framerate seems to be fine (60 @ 60, anyway), I produced the problem playing 24p on a 60 Hz screen.

[MS-DOS]

Can you reproduce this reliably? If so, please provide me with a step-by-step guide to reproduce the issue. Thanks. Also, maybe a freeze report from this situation (Ctrl+Alt+Shift+Break/Pause) might help.

Yes. MPC-HC's window needs to be fully covered by another window or minimized, then it freezes after ~30 seconds being like that. Sometimes it doesn't appear frozen after getting back to\unminimizing it and instead very quickly shows all the frames it supposed to show while being stuck, OSD reports 900+ dropped frames and the playback continues. Pretty weird. I PM'd you a freeze report.

[aufkrawall]

Ok, thanks for the feedback. I suppose you compared to the old "high" setting, right?

Yes.
madVR high:
http://abload.de/thumb/madvr3iuuh.png (http://abload.de/image.php?img=madvr3iuuh.png)

Shiandow 0,7; 2 (MPDN default):
http://abload.de/thumb/shiandowlwuqm.png (http://abload.de/image.php?img=shiandowlwuqm.png)

However, Shiandow's deband can be very interesting with low quality sources, since it can be used as some kind of deblocking filter.

[James Freeman]

For SD content I definitely like sharpening before scaling; I find sharpening SD after scaling does very little.
When sharpening before upscaling one needs to be very careful because a little more can be too much.
To summarize, sharpening before scaling has a sweet spot, overdoing that sweet spot can destroy the image; What more, is that this sweet spot is personal...