madVR - high quality video renderer (GPU assisted)

[4h4h270]

Originally Posted by madshi

How does "now" define exactly? Starting with which madVR build? And which previous build did not have the problem?

Can you provide me with screenshots of the banding, compared to how a previous working madVR build didn't have banding in the same frame? Please also provide a screenshot of the original unscaled frame, so that I can try to reproduce the problem on my own PC. Which downscaling algo are you using?

Do you have any non-default "trade quality for performance" options selected, like e.g. "use 10bit image/chroma buffer" or something like that?

It happens from 0.90.6, I've tried 0.90.4 nothing wrong. I turn off all
"trade quality for performance" options. This bug can not reproduce by scrshot pic, only when playing video. Any downscale with linear light on and chroma upscale with reconstruction & jinc.

RAW:
http://abload.de/img/rawc1prg.png
0.90.6:
http://abload.de/img/rawc1prg.png
0.90.4:
http://abload.de/img/rawc1prg.png

[leeperry]

thanks for the new builds, just ran a few tests:

-first time I updated from .4 to .6 I was greeted with major posterizing but I don't seem to be able to reproduce anymore
-I don't like the 1.6 gamma thingie as it kills picture depth, I believe I prefer LL enabled but I would need to run more tests
-I can't get either SSIM options in .6 to look anywhere near as natural as in .4, I regularly get that stunning "look through a window" feeling with the latter but both options in .6 look way oversharp & aliased(I do run a combination of full SR + NNEDI for luma & RS for chroma so I'm on a tightrope so to speak), one being even worse than the other

time for some lighthouse ground truth comparisons I guess? should I use CC AR LL for the original 1080p>720p downscale?

[madshi]

Originally Posted by GCRaistlin

now it works OK. Thank you!

Good to hear.

Originally Posted by QBhd

I just watched a 1080p film downscaled to my 1024x768 plasma and NNEDI3 was not being used for chroma. madVR seems to be falling back to JINC AR for some reason. This was not an issue two builds ago since I watched two movies Saturday night with v0.90.4

I suppose you're using linear light downscaling? Earlier madVR builds first upscaled chroma in that situation and then converted the image to RGB and then downscaled. The latest build may opt to scale luma and chroma separately for performance reasons. In that situation madVR only uses NNEDI3 for chroma upscaling if the upscaling factor needed for chroma is at least 1.5x. In your case it's 768/540 = 1.42, so madVR decided not to use NNEDI3 because it probably wouldn't bring any visual benefit.

I guess that's probably not something you like to hear? It does make sense to limit NNEDI3 to be used only for certain upscaling factors, though, especially for chroma (where upscaling quality is much less important than for luma, anyway). E.g. if you upscale chroma by only 5%, using NNEDI3 instead of e.g. Jinc or Lanczos wouldn't really bring any visible benefits. So I gotta define some upscaling factor at which NNEDI3 starts to be used.

Originally Posted by leeperry

-first time I updated from .4 to .6 I was greeted with major posterizing but I don't seem to be able to reproduce anymore

See last couple of posts. Should be fixed in the next build.

Originally Posted by leeperry

-I don't like the 1.6 gamma thingie as it kills picture depth, I believe I prefer LL enabled but I would need to run more tests

Ok, thanks.

Originally Posted by leeperry

-I can't get either SSIM options in .6 to look anywhere as natural as in .4, I regularly get that stunning "look through a window" feeling with the latter but both options in .6 look way oversharp & aliased(I do run a combination of full SR + NNEDI for luma & RS for chroma so I'm on a tightrope so to speak), one being even worse than the other

SSIM in .4 was partially broken, at least when using gamma downscaling. It works as intended in .6 now.

Originally Posted by leeperry

time for some lighthouse ground truth comparisons I guess?

The problem with that is that for downscaling we don't really have an objective groundtruth. You could make comparison screenshots, but with downscaling we're back at subjective "what looks better to your vs my eyes" judgement calls. The only way to judge this objectively would be to display the image in the original resolution on display A, then move display A back until the image has the same size as the downscaled image on display B. Then you can compare whether the downscaled image on display B looks near to the original res image on display A or not. But it will still be difficult to compare because it will be two different focus depths.

Originally Posted by 4h4h270

It happens from 0.90.6, I've tried 0.90.4 nothing wrong.

Should be fixed in the next build.

[madshi]

madVR v0.90.7 released

http://madshi.net/madVR.zip

* fixed: banding when using linear light downscaling (introduced in v0.90.5)
* fixed: some problems when moving madVR to a different monitor
* fixed: OSD display bug when using different scaling algos for X and Y

[4h4h270]

no more banding now

[leeperry]

Originally Posted by madshi

SSIM in .4 was partially broken, at least when using gamma downscaling. It works as intended in .6 now.

I far prefer SSIM LL AR in .4 than in .6 as either of the two new options in the latter are too sharp and aliased for my other settings(both SR/RS/NNEDI3/FRC/etc). I understand that you want to remove as many options as possible but they are all very much dependent on each other, I mean I personally already find strength 1 of both SR's too sharp so in the end a very sharp picture through an über-sharp downscaler ends up aliased and oversharpened

Originally Posted by madshi

The problem with that is that for downscaling we don't really have an objective groundtruth. You could make comparison screenshots, but with downscaling we're back at subjective "what looks better to your vs my eyes" judgement calls. The only way to judge this objectively would be to display the image in the original resolution on display A, then move display A back until the image has the same size as the downscaled image. Then you can compare whether the downscaled image looks near to that or not. But it will still be different to compare because it will be two different focus depths.

Right and for instance a videoprojector with 3 discrete panels will always suffer from misconvergence and strongly benefit from being fed a very sharp picture, but a LCD screen is plenty sharp to begin with so at the end of the day mVR is going for extreme sharpness but real life and cinema reels don't come with this "surgical" and harsh digital look, it should be possible for end-users to choose how sharp they want their picture and it's not quite possible atm in mVR if you want to enable all its bells & whistles

You made clear that soon there will only be one single SSIM button, either of them in .6 are useless to me when .4 looks so great <sigh>. You're the boss and I fullly respect that but how about a "soft" SSIM setting that wouldn't be any sharper than it was in .4 and another one called "sharp"? SSIM is meant to be sharper than CC AR LL but a sharpening downscaler without a single strength knob and when all other options(SR/RS/NNEDI3) are already extremely sharp all together end up next to unusable

My point is that bicubic downscaling now comes with as many as six sharpness options so you are effectively allowing us to subjectively pick its sharpness depending on the rest of our post-processing, personal taste and video display but SSIM is meant to go the other way around, with you deciding what looks subjectively best™ to you on your gear and us trying to adapt the rest of our settings to it.....but the only knobs I have access to are SR's and 1 is already VERY sharp for both of them =/

Oh well, talk about a missed opportunity bleh, I'll just bite the bullet and play around with the bicubic sharpness then. A sharper CC AR LL would be right up my alley I guess, hopefully I'll get it to deliver.

[NicolasRobidoux]

Originally Posted by madshi

There should be no negative RGB data. If there is, luma/luminance would be negative, too, in the same situation. I'm currently scaling in RGB. In all my tests I could rarely see a difference when scaling in RGB vs YCbCr. I'm currently doing sigmoidization in linear light RGB, and it does seem to help in some specific situations, and hurt in others, same as last time we tested this. I'm pretty sure that converting to YCbCr and only applying sigmoidization to Y wouldn't change anything at all. All that would do is to limit the effect of sigmoidization to the brightness information, while keeping color information unaffected. But brightness is where all the positive and negative sigmoidization changes are happening already right now, when doing all this in RGB. There's no effect on colors that I can see. So I don't think doing this in YCbCr instead of RGB would make any difference.

Or did I misunderstand your suggestion?

I indeed was suggesting that you only use sigmoidization with Y (or Y').

Here is something I don't understand. How do you manage to make sure that >>without clamping<< your RGB (or R'G'B') values are (for example) never negative when they are obtained by a conversion from something like >>destructively compressed<< YCbCr?

[Warner306]

Originally Posted by madshi

Hi Nicolas - long time no see!
First of all, atm all 3D related functionality in madVR is limited to 3D content coming from a compatible decoder with left and right eye images being sent separately. The only compatible decoder atm is nightly LAV and the LAV splitter currently only supports 3D MVC MKVs. This will hopefully be improved by nevcairiel soon. If you play movies which have the left & right images interweaved somehow (e.g. hard encoded as side-by-side), madVR treats them as 2D.

Your wiki might be a bit confusing to an end user, I think. Your "auto" section is probably meant to reach from "auto" until "Force 3D format below"? But in the midst of this "auto" section there are two more bold lines starting with "Active/Passive 3D". It's not totally clear if these still belong to the "auto" section or if they are meant to be separate sections. I'd optimize the font size and styles to make it clearer which parts of the text are meant to be one section and has which title etc.

The "Active/Passive 3D" lines are not clear. Since they stand under "auto", the end user might understand it as if madVR would automatically detect active/passive 3D displays and treat them differently. That's not the case.

"auto": madVR sends frame packed, if HDMI 1.4+ is available, *and* if the display supports this format. This can be an active or passive 3D display. If frame packing is not supported by either the GPU, OS or display, madVR instead renders the 3D video into "Half SBS".

"side-by-side": madVR renders Half SBS.

"top-and-bottom": madVR renders Half Top-and-bottom.

"line alternative" is described incorrectly. I'm not sure, maybe there's an official "line alternative" communication protocol that works as you describe. But that's not how madVR renders. This mode is for passive 3D displays which have polarizing filters which change polarization for each pixel row. The polarization is fixed, though. So for each frame sent by madVR, the odd lines are always for the same eye, and the even lines for the other eye. Odd/even does NOT change the eye from frame to frame.

"column alternative": same as "line alternative", just columns instead of rows/lines.

"swap left/right eye": This option is (in the long run) *not* meant to be used to fix sources which have the eyes in the wrong order. The problem with such sources should be fixed automatically by reading this information from the video file, and then communicating it to madVR. This will be added to future LAV + madVR builds. For now you can misuse this option to fix problems with such sources, but in the future this option has a different purpose: Some 3D displays *always* draw the images in the wrong order. So this option fixes that. Also for line/column alternative passive 3D displays it's not clearly defined which line/column is for which eye. So this option allows you to set that up correctly for each display.

Thanks for the clarification. I think I am (mostly) up to speed on how 3D tech works.

[Warner306]

Could someone tell me the advantages of turning off 3D in the operating system during playback of 2D or 3D videos?

[Patrik G]

there is a dude that still thinks his PS4 is the ultimate player for blu ray.
i need to convince him that madvr passes every blu ray player out there for PQ if you starting to make use of the chroma upscaling settings.
but of course you need a fast graphics card to get the most out of it.

have anyone compared the picture quality from madvr with an PS4?
when and with what settings will madvr pass it?
it cant be that hard right ?

all i know is that i sold my Pioneer BDP-LX58 blu ray player last year after comparing it to madvr.
the PS4 must be way behind even that player right?

[Codex0nz]

I've been looking around for a while now but can't find a good answer in regards to madvr, I get quite bad jitter at times with playback even with smoothing enabled, tried reclock and all sorts.

Thinking it might be because GPU limitation, system is running a Intel Pentium g3258 at 4ghz, I do have a gt220 laying around but don't know if it would help.

If it's definitely the intel hd graphics limiting my rendering performance then was thinking of getting a gt730 passive (for that sweet silence)

Thoughts on this?

[ryrynz]

Originally Posted by Patrik G

there is a dude that still thinks his PS4 is the ultimate player for blu ray.

There's are words for people that blindly believe in things with no real evidence..

Originally Posted by Patrik G

all i know is that i sold my Pioneer BDP-LX58 blu ray player last year after comparing it to madvr.
the PS4 must be way behind even that player right?

I can't imagine that it would be better. There isn't a consumer hardware device available that can best madVR in picture quality or options. For some people a PS4 may be good enough. I wouldn't waste time trying to convince anyone of anything they're not already open to.

[Warner306]

Originally Posted by Codex0nz

I've been looking around for a while now but can't find a good answer in regards to madvr, I get quite bad jitter at times with playback even with smoothing enabled, tried reclock and all sorts.

Thinking it might be because GPU limitation, system is running a Intel Pentium g3258 at 4ghz, I do have a gt220 laying around but don't know if it would help.

If it's definitely the intel hd graphics limiting my rendering performance then was thinking of getting a gt730 passive (for that sweet silence)

Thoughts on this?

Are you using refresh rate switching (e.g. 23.976 fps played back at 1080p23)?

Are the CPU/GPU queues full in madVR's rendering stats (Cntrl + J)?

Have you tried a simple audio renderer that is not ReClock?

Are the dropped frames and presentation glitches counters staying at zero during playback?

If you can answer yes to all of these, it may be your display is at fault or the low frame rate of the content.

The GT 730 would allow you to get a lot more out of madVR than integrated graphics. If you can afford it, it would be a wise upgrade.

[ryrynz]

Originally Posted by madshi

Can't seem to reproduce this here. Can you please make a screenshot which shows the problem, with the Ctrl+J OSD turned on? Please also create an empty file named "ShowRenderSteps" in the madVR folder before creating the screenshot. Also, could you please upload your settings.bin file?

This occurs when I have SSIM downscale detailed only with quadrupling and when there's either nothing ticked in upscaling refinement or when a refinement is ticked but only applied when image scaling is complete, no issues when every 2x upscaling step is ticked (applies to any enhancement/combination)

I reset to defaults and only changed downscale to SSIM detailed and image doubling to quadrupling without touching anything else. It doesn't matter how many neurons are selected, so probably little point in uploading the settings.dat, it can't be related to that.

Almost looks like you run out of OSD space when the render steps are shown.

http://i.imgur.com/xyzjGlS.jpg

Originally Posted by aufkrawall

I think "detailed" looks too aliased.

I prefer the detail level.. but yes edges do seem a touch over aliased although clean isn't *that* much better.
madshi is it possible to slightly lower the sharpness in some way of SSIM so that detailed could be preferred?
I like the extra detail throughout the picture but I'm not a huge fan of the extra edge sharpness that comes with it.

[FreeFall]

Originally Posted by StinDaWg

Well, this is sort of fixed for me, but not really. Instead of getting a black screen when switching out of FSE to windowed in D3D11, I just get a still image from the last part of the video. When I go back to FSE it resumes fine.

In other words, I can't watch anything in windowed mode after leaving FSE D3D11 because it just shows a still image while audio keeps playing. D3D9 works fine as usual.

Using AMD 7850.

Also, when going from FSE to windowed and back to FSE in D3D11 sometimes the present queue will drop down to 1-3 and it will never recover, while the preset stats also increase ~10ms and don't drop again.

This usually also happens after the windows screen saver comes on while a video is paused. Doesn't happen in D3D9. To reproduce set windows screen saver to 1 minute, pause video, let screen saver kick in then play video. Check OSD. Stats are increased and present queue drops to 1 and can't recover.

I have windows screen saver set to 1 minute because I have a plasma and want to minimize IR/burn in.

madshi, I'm having a similar problem.

Windows7 x64 Pro (GeForce GTX 980Ti, 356.04), Zoom Player Max 11.1, latest LAV nightly build and madVR 0.90.7.

Using the default d3d9 FSE mode, to reproduce play a DVD and then stop playback (s key), the last frame stays on screen until you exit the player (alt + x). If I enable the option use a seperate device for presentation (Vista and newer) playback works without any problems.

Thanks.

[MokrySedeS]

Hi madshi, I have two very similar issues.
1. When I drag mpc-hc window from one display to another the playback pauses and I have to press Play/Pause button twice to resume. It was introduced in the latest .7 version. It happens every time in both D3D9 and D3D11.
2. When the player picks up another file from the playlist and madVR switches to different refresh rate - the playback pauses. What's interesting is that in D3D9 I only need to press Play/Pause once to resume, twice in D3D11. Also, it happens only from time to time, more often in D3D11 than D3D9. This was introduced earlier but I'm not sure in which version exactly.

I'm on Windows 10, NVidia GTS 450, driver v361.75.

Originally Posted by Codex0nz

I've been looking around for a while now but can't find a good answer in regards to madvr, I get quite bad jitter at times with playback even with smoothing enabled, tried reclock and all sorts.

Thinking it might be because GPU limitation, system is running a Intel Pentium g3258 at 4ghz, I do have a gt220 laying around but don't know if it would help.

If it's definitely the intel hd graphics limiting my rendering performance then was thinking of getting a gt730 passive (for that sweet silence)

Thoughts on this?

I had an awful jitter with 25fps files played at 50Hz on my tv but the queues were full, no dropped frames etc. I had to turn D3D11 presentation off. Now it's buttery smooth. Your issue might be totally different though, too little detail so I'm just guessing :P

[Anima123]

Madshi, with 0.90.7, if double a 720p video with NNEDI3 32 then using SSIM for downscaler, the video will be 'greenized' if I didn't use 'double chroma resolution'.

The 'greenized' screen shot:
https://www.dropbox.com/s/w4nhags6wh...edPic.png?dl=0

The normal screen shot (not the same frame but it's obvious):
https://www.dropbox.com/s/q6o84zhtdl...alPic.png?dl=0

Edit: There's no SuperRes used for both. Super-xBR image doubler does not suffer this issue.

[ryrynz]

Originally Posted by Anima123

Madshi, with 0.90.7, if double a 720p video with NNEDI3 32 then using SSIM for downscaler, the video will be 'greenized' if I didn't use 'double chroma resolution'.

Could be related to my issue. Look a few posts above and perhaps do what he requested I do.

*Fixed in his beta build*

[omarank]

Originally Posted by madshi

I need your feedback about the following two things:
1) Do you prefer SSIM "clean" or "detailed" downscaling? The "detailed" option should have a bit more aliasing and sometimes maybe some minor artifacts, but looks a bit more detailed. The "clean" option is a little bit slower, and should have a bit less problems with aliasing and artifacts, but might smooth things over ever so slightly, compared to "detailed". So which one do you prefer? Ideally I'd get rid of one option, so please help me decide which one to keep and which one to drop.

2) You can (for downscaling) now decide between using normal (gamma) light downscaling, linear light downscaling which uses a 2.2 gamma value) and the new "use 1.6 gamma instead of 2.2" option, which does linear light downscaling with a 1.6 gamma value. So the 1.6 option pretty much is exactly in the middle between gamma and linear light downscaling. I would *really* like to get rid of two options and always do downscaling in one specific light (either gamma or 1.6 linear or 2.2 linear). So which of those 3 options do you prefer and can I get rid of the other two?

1. I prefer “detailed” over “clean”. The former may have more aliasing but from a distance it looks more pleasing than the latter. The difference is subtle though.

2. 1.6 linear looks more like Gamma light to me. I prefer the look of 2.2 linear. For the past few years, Catmull Rom AR LL (with 2.2 gamma) has been considered as the best downscaling setting. I would suggest to not replace the 2.2 gamma with 1.6 gamma LL, so that one can always go back to CR AR 2.2LL.

Originally Posted by madshi

Probably I used a too low bitdepth texture format in order to save performance...

I think you were using 32 bit float texture format specifically for the downscaling steps when Linear Light downscaling was used. Have you changed that to even less than 16 bit texture format?

Originally Posted by madshi

I suppose you're using linear light downscaling? Earlier madVR builds first upscaled chroma in that situation and then converted the image to RGB and then downscaled. The latest build may opt to scale luma and chroma separately for performance reasons.

As you explained it earlier, Linear Light downscaling is done on RGB image so chroma upscaling must be done first to get the RGB image. Now in the latest build you are scaling luma and chroma separately (perhaps for large downscaling factors?). How is it done now?

Well, I seem to like the previous downscaling method especially for 4k to FHD when the chroma was upscaled to 4k and then image was downscaled to FHD. I think the supersampling effect on chroma affects the perception of depth in the images. I am not sure if this can be shown through screenshots. Anyway, I would prefer if you could add a trade quality option to restore the previous downscaling method.

[huhn]

Originally Posted by 6233638

I'm not saying that you're wrong, but where do you get that 2.2 gamma is "mathematically correct" for film?
If you're talking about the BT.709 transfer function, that's not 2.2 gamma. (more like 1.96 for the best-fit gamma)

film is not really mastered for gamma 2.4 or at gamma 2.2.
the trick is to assume the gamma is 2.2 and to change it to 2.4 for a bat cave to get the "correct" gamma of 2.4.

because the mastering lightning and the viewing lightning is totally different.

so i have no clue why video should be 2.2 while film should be 2.4.
at this moment it is assumed the source is gamma 2.2 and altered to 2.4 so if the source was 2.4 in the first place we wouldn't get gamma 2.4 with our gamma 2.4 3D LUTs.

the same reason "color & gamma" is increasing the gamma when set to 2.4. because it doesn't know the 3d LUT is already 2.4.