madVR - high quality video renderer (GPU assisted) [Archive] - Page 179

madshi

24th July 2011, 19:41

Here's a good starting point: http://www.avsforum.com/avs-vb/showthread.php?t=1038602
I've read that thread and it seems that there are opposing statements from various experts that I trust. Most experts agree that SMPTE C monitors are still wide spread. The thread is 2 years old, though, so things may have changed a little bit since then. But anyway, some of the insiders (e.g. Mr.D and coldmachine) claim that the monitors are profiled and turned into BT.709 monitors by using 3dluts. In contrast to that is what Joe Kane says. So there you have it: The experts are *not* in agreement at all. Actually the majority of the "biggest" experts on AVSForum seems to agree that Blu-Rays are mostly mastered to BT.709. Joe Kane seems to disagree, so do some others. Those on AVSForum who actually have experience in mastering themselves (mostly Mr.D) say that Blu-Rays are mastered to BT.709. All the others are just reporting what "someone else" said.

Without automatic rules for the input gamut, there will never be a near-foolproof solution for automatic gamut mapping I'm afraid.
With automatic rules there won't be, either, if you're right in that USA and EU Blu-Rays have different primaries. Which however I'm still not convinced of.

OK, I've run a comparison again: http://www.mediafire.com/?jab3oh76jbpvxln

using:
-0-255 RGB32 in HR
-0-255 RGB32 in 0.67(same as HR)
-0-255 RGB32 in 0.69(PC levels are automatically detected from ffdshow, but gamma is way too bright)
-0-255 YV12 in 0.69 and setting the PC levels manually(gamma is slightly too dark)

Ideally the two last ones should look exactly like the two first ones :)
The YV12 one looks pretty close now. Not sure if there's naything wrong with it? Anyway, I can see that there's some problem with the RGB handling. Will try to find out what's going on there...

Wile-E-Coyote

24th July 2011, 19:43

Hi madshi, couple of things for you.

1) The option to treat 25p videos as 24p works fine, trouble is it also treats 25i videos as 24p. This is a bit annoying because 25i material is usually TV material and therefore does need to be slowed down.

2) Do your decoders are assisted by the gpu? I use LAV Cuvid specifically for that reason. I'd rather not ask my CPU to decode video frames when my gpu can do it better.

3) For some stupid reason neither my TV nor my graphic card remember when I set them to "no overscan". If I want no overscan I have to set it manually everytime I switch my TV input to HDMI, so I just gave up (overscan isn't so bad anyway). Problem is overscan means the exclusive mode seek bar is barely visible (just the top of it, not even the bar itself), and the OSD is partially hidden.
Any chance you could add the option to move up the seek bar and increase the margin between the border of the frame and the OSD?
The weird thing is even for 2:35 videos the seek bar is at the bottom of the screen, instead of the bottom of the frame... Personally i'd rather have it the other way.

madshi

24th July 2011, 19:44

Would it be possible to use madvr decoding and the sharpening and other postprocessing from ffdshow?
No, that's not possible, unfortunately.

madshi

24th July 2011, 19:49

1) The option to treat 25p videos as 24p works fine, trouble is it also treats 25i videos as 24p. This is a bit annoying because 25i material is usually TV material and therefore does need to be slowed down.
At this point in time you need to deinterlace 25i material before sending it to madVR, because madVR cannot deinterlace itself. If you do that, madVR will not ever get 25i, but instead 50p. Problem solved.

2) Do your decoders are assisted by the gpu? I use LAV Cuvid specifically for that reason. I'd rather not ask my CPU to decode video frames when my gpu can do it better.
The current madVR decoders are all pure software. But what do you mean with "when my gpu can do it better"? It can't do it better. It may be able to do it with less power consumption, but other than that it will at best be identical to software decoding.

3) For some stupid reason neither my TV nor my graphic card remember when I set them to "no overscan". If I want no overscan I have to set it manually everytime I switch my TV input to HDMI, so I just gave up (overscan isn't so bad anyway). Problem is overscan means the exclusive mode seek bar is barely visible (just the top of it, not even the bar itself), and the OSD is partially hidden.
Any chance you could add the option to move up the seek bar and increase the margin between the border of the frame and the OSD?
The weird thing is even for 2:35 videos the seek bar is at the bottom of the screen, instead of the bottom of the frame... Personally i'd rather have it the other way.
These things are on my to do list, but I've a lot of other things to do first.

Okay, here is the log. I've disabled "pause playback until render queue is full" (I had already disabled the other one), waited for more than 30 seconds, and on the second 15 (aprox), the movie audio started, but the screen remained completely black. Also, MPC-HC was unresponsive, so I had to kill the process.
http://www.megaupload.com/?d=CFZZNROG
That's so weird. Same effect again, the log stops after about 150 milliseconds and I can't see why! :( Can anybody else reproduce this problem?

I'd suggest trying different MPC-HC builds, because the problem only seems to occur with MPC-HC.

leeperry

24th July 2011, 19:54

I've read that thread and it seems that there are opposing statements from various experts that I trust. Most experts agree that SMPTE C monitors are still wide spread. The thread is 2 years old, though, so things may have changed a little bit since then. But anyway, some of the insiders (e.g. Mr.D and coldmachine) claim that the monitors are profiled and turned into BT.709 monitors by using 3dluts. In contrast to that is what Joe Kane says. So there you have it: The experts are *not* in agreement at all. Actually the majority of the "biggest" experts on AVSForum seems to agree that Blu-Rays are mostly mastered to BT.709. Joe Kane seems to disagree, so do some others. Those on AVSForum who actually have experience in mastering themselves (mostly Mr.D) say that Blu-Rays are mastered to BT.709. All the others are just reporting what "someone else" said.
How about some real world experiments? give it a shot, and let us know what you see? We all have our own points of references, and I trust JK and the ISF more than some hearsay on an internet board..especially when what they say matches what my eyes are seeing(and so do many more ppl on HCFR).

With automatic rules there won't be, either, if you're right in that USA and EU Blu-Rays have different primaries. Which however I'm still not convinced of.
Well, 24p HD is like 25fps DVD's...you cannot automatically detect whether HD is SMPTE-C/EBU/HDTV or whether 25fps SD content is genuine PAL or sped up NTSC. C'est la vie...the only thing we can do is setting up automatic rules for the most common result(based on our very own beliefs) and manually change it on the fly if need be. That's the very reason why automatic refresh rate switching scripts cannot work in Reclock...expect if you start tagging your files and setup automatic rules accordingly. I personally prefer to have some automatic rules that will cover +95% of my content(I hardly ever watch EU BD's) than none at all. Yet, even if you choose to believe that HD=709, then you'll still have a problem w/ DVD(25fps=EBU, 23.976/29.97=SMPTE-C) so there's still room for user based gamut rules IMHO.

The YV12 one looks pretty close now. Not sure if there's naything wrong with it? Anyway, I can see that there's some problem with the RGB handling. Will try to find out what's going on there...
The YV12 looks close but it's not quite there, it should look exactly like HR and 0.67 and it doesn't :o

And I cannot use ddcc() CUDA in RGB32 at all w/ >0.67.

Gleb Egorych

24th July 2011, 20:07

Those of you who have crashes, can you please try this special build to see if it fixes the issue?

http://madshi.net/madVR69test.zip

Yes, it fixed ntdll.dll crashes for me.

I also reported (http://forum.doom9.org/showthread.php?p=1509618#post1509618) a problem with Zoom feature of Zoom Player and madVR exclusive mode in 0.66. The crash was fixed in 0.67 but madVR quits exclusive mode as soon as I try to zoom with "exclusive mode failed" OSD message.

madshi

24th July 2011, 20:09

How about some real world experiments? give it a shot, and let us know what you see? We all have our own points of references, and I trust JK and the ISF more than some hearsay on an internet board..especially when what they say matches what my eyes are seeing(and so do many more ppl on HCFR).
We're talking here about what mastering houses are doing. Who do you think knows that better:

(1) Someone who is working in a mastering house (Mr.D)?
(2) Someone involved with ISF and consumer electronics (Joe Kane)?

Both are knowledgable persons, IMHO, but I think in this specific case Mr.D knows better. Real world experiments? I'm not really the color guy. I can see differences in sharpness, detail, grain etc very quickly. I can usually tell you whether a master is an old 2K master or a new 2K master or a 4K master, just by looking at it. But ask me which colors are correct? I don't know. IMHO the differences between BT.601 and BT.709 primaries are not *that* big. But everyone is different. I care much about detail and lack of edge enhancement. I don't care as much about color. I still want madVR to be as accurate as possible, though, of course.

you'll still have a problem w/ DVD(25fps=EBU, 23.976/29.97=SMPTE-C) so there's still room for user based gamut rules IMHO.
madVR auto switches to EBU primaries for content with a height of 576 pixels and to SMPTE C for all other SD content. That should cover most DVDs. Anyway, I already said that profiles and rules are on my to do list. But we can continue to discuss this for days, it won't change the priority of profiles and rules in my to do list.

Wile-E-Coyote

24th July 2011, 20:35

At this point in time you need to deinterlace 25i material before sending it to madVR, because madVR cannot deinterlace itself. If you do that, madVR will not ever get 25i, but instead 50p. Problem solved.
Well as far as I'm aware the content IS being deinterlace... But is it deinterlaced properly? ffdshow and Nvidia purevideo have options to double the framerate but they don't seem to be doing it. LAVCuvid is the best at deinterlacing that I know, but guess what : it doesn't work with DVDs (yes most of my interlaced content is DVD :p). I'll do some more tinkering with ffdshow to try and find a good deinterlacing method.

The current madVR decoders are all pure software. But what do you mean with "when my gpu can do it better"? It can't do it better. It may be able to do it with less power consumption, but other than that it will at best be identical to software decoding.
Well yeah that's kinda what I meant. I'm using a laptop, not really a week laptop but I try to save as much calculating power as possible. Not that I wanna play Crysis 2 while watching a video at the same time but I'd rather my movie not lag whenever I'm being contacted on MSN. Which is why I prefer hardware assisted decoders.

nevcairiel

24th July 2011, 21:15

highly knowledgeable industry insiders are willing to share w/ anyone open-minded enough to listen: http://pro.jvc.com/prof/attributes/features.jsp?model_id=MDL101766

One would guess that the pro department of JVC would know what's being "most commonly" used in production environments? :p

You take your "knowledge" from marketing texts? You know that some people write those texts that have probably never touched the device in question, yes?
Marketing is present everywhere, be it in consumer or in professional gear. Don't mistake it for facts.

Either you're grasping for straws with that comment, or you really believe marketing texts .. not sure what is better. :p

cyberlolo

24th July 2011, 21:26

That's so weird. Same effect again, the log stops after about 150 milliseconds and I can't see why! :( Can anybody else reproduce this problem?

I'd suggest trying different MPC-HC builds, because the problem only seems to occur with MPC-HC.

Not only on MPC-HC, but only with the combination MPC-HC + TV output! If I try the same files with MPC-HC + PC Monitor, everything is fine.
The only pattern that I've found for this problem seems to be the application that has been used for coding the movie. I mean, every file that freezes the player has been coded with mkvmerge v4.4.0 or higer. Also, another movie that gives the freeze error has been codec with virtualdubmod.

All the other movies have been codec with versions of mkvmerge lower than 4.4.0, and all of them play perfectly fine. Does this make any sense to you?

leeperry

24th July 2011, 21:30

We're talking here about what mastering houses are doing. Who do you think knows that better:

(1) Someone who is working in a mastering house (Mr.D)?
(2) Someone involved with ISF and consumer electronics (Joe Kane)?
Joe Kane's Sammy projectors are used all over Hollywood and in post-production environments worldwide coz they don't skimp out in the colorimetry department. The french CEO of the ISF claimed that he was supervizing the weekly recalibration of CRT's in mastering houses. Those 2 ppl know a thing or two about what happens in the real world. And again, it matches exactly what me and many other color freaks on the HCFR forum are seeing.

Real world experiments? I'm not really the color guy. I can see differences in sharpness, detail, grain etc very quickly. I can usually tell you whether a master is an old 2K master or a new 2K master or a 4K master, just by looking at it. But ask me which colors are correct? I don't know. IMHO the differences between BT.601 and BT.709 primaries are not *that* big. But everyone is different. I care much about detail and lack of edge enhancement.
Well, it's actually much simpler than you'd think. Using the right gamut ensures that you're seeing the same colors the mastering engineer calibrated. It's like Reclock, ppl who are not used to it will call it a tool for OCD'ed videophiles...but watch movies in butter smooth 24p for a while, then go back to 60Hz if you dare. Same story w/ colorimetry, roll the LUT's on US and EU BD, take your time...there's a good chance that you'll find the colorimetry far more natural and the subjective contrast ratio higher because colors won't be overlapping each other anymore...you get 1:1 colorimetry w/ what was shot and mastered, goodness gracious :cool:

I don't care as much about color. I still want madVR to be as accurate as possible, though, of course.
I understood that when you told me "The YV12 one looks pretty close now. Not sure if there's anything wrong with it?". You regularly nitpick about virtually invisible placebo-ridden changes in your VR, then you ask me if a slightly too dark picture is OK'ish? :p

There are many naysayers on audiophile forums that like to call bs on anything they haven't tried, just for fun...all I've read so far about this gamut matter here on doom9 is ppl who haven't done any real world experiments. Someone said "talking about music is like dancing about architecture"...that's the exact same story as w/ this gamut mumbo jumbo, depending on who you ask you'll be told that most/some mastering houses do their job properly, and that some/most others don't. How about trying it for yourself, and see what your brain has to say about it?

madVR auto switches to EBU primaries for content with a height of 576 pixels and to SMPTE C for all other SD content. That should cover most DVDs. Anyway, I already said that profiles and rules are on my to do list. But we can continue to discuss this for days, it won't change the priority of profiles and rules in my to do list.
I'm as much bored of selling the SMPTE-C argument as you are of listening to it, honestly. At this point, it's just a troll bait calling for ppl who are too lazy to run real world experiments to give their "I heard it through the gravepine" opinion, nothing more. I gave it another shot as you were asking for it once more.

Either way, I'm glad to hear that some automatic gamut mapping rules will be in the works at some point, as that will make a lot of colorimetry freaks happy fosho!

madVR auto switches to EBU primaries for content with a height of 576 pixels and to SMPTE C for all other SD content
How about upscaled EU DVD then? :p

e-t172

24th July 2011, 21:58

The problem with these "real-world experiments" that you seem to value above all is that you don't have the actual source to compare to. The only way this could work is if you bring your system to the mastering house, then play the BD side-by-side with the mastering monitors. Then the comparison would be valid. But of course this isn't gonna happen.

Right now all you're doing is switching gamuts on some BDs and declaring "oh, that's much better". Which is purely subjective. And no, you can't say "oh, that's much closer to the source", because you don't have the source to compare it with, so you honestly can't make that statement. One thing I learned when calibrating audio equipment (I'm developing audio calibration software) is that you can't trust people's judgments when it comes to fidelity. Including my own.

Remember that the goal here is fidelity. I'm perfectly fine seeing purple lemons on my screen as long as I'm sure that they were purple on the screen of the mastering engineer.

Blight

24th July 2011, 22:02

madshi:
Good news, tried the problematic video and reloaded it 50 times with resume last position and no crash on load.

Bad news, switching MadVR between fullscreen on monitor #1 and windowed mode on monitor #2 causes several issues:
The video will freeze with audio playing requiring a pause+play (graph stop+play) for the image to continue playing.
On the second attempt, the video blacks out, requiring a pause+play for the audio to play and several seconds for the video to return, but when the video does return, the entire player interface becomes unresponsive, forcing a task kill.
This happens even with fullscreen exclusive mode disabled...
I can't create a debug log, as the latest test version didn't include a debug build.

madshi

24th July 2011, 22:03

The french CEO of the ISF claimed that he was supervizing the weekly recalibration of CRT's in mastering houses.
So? What happens afterwards with 3dluts is a separate story. Mr.D says he has different luts to switch his mastering monitor to different standards, so he can get whatever he needs.

Fact is, the experts do *not* agree on this one. And subjective impressions of what might look more correct to your eyes are just that: subjective impressions. I'm not saying you're wrong. But I'm also not saying you're right. I consider this matter undecided. Maybe some Blu-Rays are encoded to SMPTE C primaries, maybe some are encoded to BT.709 primaries, maybe some are encoded to EBU primaries. Maybe things will change fluidly, as new mastering monitors are bought (studio CRTs are out of production). I believe the best solution for madVR is to stick to the standards.

How about upscaled EU DVD then? :p
Already works, too, if you upscale in realtime somewhere in the DirectShow processing chain. madVR will detect the dimensions of the original video before upscaling and use that for primaries auto switching. Give it a try!

Not only on MPC-HC, but only with the combination MPC-HC + TV output! If I try the same files with MPC-HC + PC Monitor, everything is fine.
The only pattern that I've found for this problem seems to be the application that has been used for coding the movie. I mean, every file that freezes the player has been coded with mkvmerge v4.4.0 or higer. Also, another movie that gives the freeze error has been codec with virtualdubmod.

All the other movies have been codec with versions of mkvmerge lower than 4.4.0, and all of them play perfectly fine. Does this make any sense to you?
Not really. Can I have a sample? Maybe I can reproduce it then on my secondary TV.

madshi

24th July 2011, 22:04

Good news, tried the problematic video and reloaded it 50 times with resume last position and no crash on load.
Great!

Bad news, switching MadVR between fullscreen on monitor #1 and windowed mode on monitor #2 causes several issues:
The video will freeze with audio playing requiring a pause+play (graph stop+play) for the image to continue playing.
On the second attempt, the video blacks out, requiring a pause+play for the audio to play and several seconds for the video to return, but when the video does return, the entire player interface becomes unresponsive, forcing a task kill.
This happens even with fullscreen exclusive mode disabled...
I can't create a debug log, as the latest test version didn't include a debug build.
I'll see if I can reproduce that here. I guess it occurs with every video file, correct? FWIW, does the option "delay playback until render queue is full" have anything to do with this problem?

leeperry

24th July 2011, 22:08

Right now all you're doing is switching gamuts on some BDs and declaring "oh, that's much better". Which is purely subjective. [..]
Remember that the goal here is fidelity.
I'm well aware of everything you're saying and I'm as much of an audiophile as a videophile(two nasty diseases :p), but the french CEO of the ISF said:
here's what happens when using a display with sRGB/HDTV gamut :

-for telecines that were made on an EBU monitor, only the green primary will be off.

-for telecines that were made on an SMPTE-C monitor, colors will be oversaturated.
What matters here is the end-user satisfaction, and using the wrong gamut gives *wrong* colors...use SMPTE-C on a EU BD, and it'll look completely funky, the same way SMPTE-C in HDTV will look oversaturated. Use the right gamut, and all of a sudden colors look natural and the contrast ratio seems increased.

But let's agree to disagree, as noone is willing to change his mind anyway..based on his own real world experience, or lack thereof.

leeperry

24th July 2011, 22:28

Fact is, the experts do *not* agree on this one. And subjective impressions of what might look more correct to your eyes are just that: subjective impressions.
Last time I checked, motion pictures and audio equipment were mere illusions fooling the brain. We are here for the sheer subjective impressions, that's the fuel of it all.

Many audio equipment manufacters will throw amazing SNR/THD/IMD specs at you, but that doesn't prove anything whatsoever. It can measure like crazy and sound like crap...the subjective judgment and the end-user satisfaction are what truly matter here, and the only way to find out who's right or wrong on a case to case basis is to run real world experiments. There's a good chance that you'll be seeing the light, or maybe not...but at least you will have tried. Your real world experience will give weight to your opinion.

Already works, too, if you upscale in realtime somewhere in the DirectShow processing chain. madVR will detect the dimensions of the original video before upscaling and use that for primaries auto switching. Give it a try!
Will do! but I'm currently forced to use ddcc() CUDA to get proper gamut mapping, and mVR >0.67 can't output untouched RGB32. And there's no way in hell I'll be watching my BD's in HDTV gamut anyway :D

But a proper 0-255 YV12 pipeline would be great, for when I don't have enough CPU horse power to run ddcc().

madshi

24th July 2011, 22:31

There's another problem with real world experience: Not all Blu-Rays are equal. Actually far from it. Just look at how FOTR suddenly got a teal tint. The difference between the TE and EE editions of FOTR alone are much *much* bigger than the difference between SMPTE C and BT.709. And many newly remastered movies (e.g. Aliens) seem to have that teal tint. Also many movies were intentionally regraded to look unnatural. Probably that applies more to fantasy, science fiction, action etc movies, and less to dramas. But well, the former is the movie type I prefer. Anyway, what I'm trying to say is that I doubt that subjective impressions can have any say in which is the "right" gamut. Furthermore: Our eyes/brain adjust very quickly to different white points (that's why FOTR EE doesn't look nearly as tinted during actual movie watching compared to looking at screenshots). That's another big problem in subjective impressions...

cyberlolo

24th July 2011, 22:34

Not really. Can I have a sample? Maybe I can reproduce it then on my secondary TV.

Ok, will try to get some short one. BTW, two more things:

1) Is it possible to change the OSD shortcut from Ctrl+J to some other combination?
2) Using KMPlayer, madVR's seekbar never shows. If you move the cursor at the bottom, it shows the KMPlayer's seekbar, so switching to windowed mode.

Blight

24th July 2011, 22:36

madshi:
I disabled 'delay playback...', disabled 'use a sep device for present', enabled both 'use a sep' and 'use d3d11', none of which helped.
This is strange, it even happens when going fullscreen on the same monitor! I'm going to investigate if it's a specific code I use in the fullscreen switch that messing with MadVR in some way.
The first report was for an FLV with H264/AAC.
It's actually even worse with some other sample files, I get crashes in 'fault module StackHash_0a9e' (AVI with XVID/MP3).
I get the main monitor completely blanked out while the second monitor flashes the video for a second and then remain frozens with the player UI frozen as well. (MKV with H264/AC3).

So it seems to be unstable in a variety of ways (reminder, exclusive mode is disabled), so it's not the switch event.

Xaurus

24th July 2011, 22:38

edit: posted below

leeperry

24th July 2011, 22:40

There's another problem with real world experience: Not all Blu-Rays are equal. Actually far from it. Just look at how FOTR suddenly got a teal tint.
Indeed, the human brain can "adjust" to any white point very quickly...the same way it won't let "untouched" audio pass through for more than a fistful of seconds. A good rule of thumb IME is to go US=SMPTE-C/EU=EBU...and in the rare cases where I'd find the colors artificial looking, I'd be using the gamut mapping hotkey and look for a better choice. Indeed, nothing's written in stone and if the movie was meant to look green on purpose(like Amelie or Revolver) then good luck finding the right gamut on sight. There's a standard, but some/most ppl don't care about it, because the average customer uses an uncalibrated display in 60Hz anyway. It's all dumbed down on purpose.

madshi

24th July 2011, 22:41

There's a good chance that you'll be seeing the light, or maybe not...
I see the light every day, thank you very much.

1) Is it possible to change the OSD shortcut from Ctrl+J to some other combination?
Changing key combinations is on my to do list, like many other things.

2) Using KMPlayer, madVR's seekbar never shows. If you move the cursor at the bottom, it shows the KMPlayer's seekbar, so switching to windowed mode.
KMPlayer is faster than madVR in showing the seekbar. Nothing I can do about it.

madshi:
I disabled 'delay playback...', disabled 'use a sep device for present', enabled both 'use a sep' and 'use d3d11', none of which helped.
This is strange, it even happens when going fullscreen on the same monitor! I'm going to investigate if it's a specific code I use in the fullscreen switch that messing with MadVR in some way.
The first report was for an FLV with H264/AAC.
It's actually even worse with some other sample files, I get crashes in 'fault module StackHash_0a9e' (AVI with XVID/MP3).
I get the main monitor completely blanked out while the second monitor flashes the video for a second and then remain frozens with the player UI frozen as well. (MKV with H264/AC3).

So it seems to be unstable in a variety of ways (reminder, exclusive mode is disabled), so it's not the switch event.
Is this a new problem? It sounds really bad. Can you say which exact madVR version introduced this? Does it occur on your PC with ZP7, too?

Blight

24th July 2011, 22:54

Madshi:
I partially figured it out. You're capturing the Alt+Enter key and doing something (switching to exclusive mode even if it's disabled in the settings?).
If I use another key to go fullscreen, it works well on the same monitor.

There is a separate issue when switching to a second monitor, the video appears on the second monitor in the wrong AR, then the AR is adjusted after maybe 100-200ms, then the whole thing freezes, requiring a task kill.
The strange thing, it doesn't freeze if I drag the player window back and forth between monitors. I'm still investigating.

madshi

24th July 2011, 22:56

Let me know what you find out. FWIW, I'm not looking for Alt+Enter. I'm not looking for any specific key combination for fullscreen switching. The only keys I'm looking for are those I need for madVR functionality (like Ctrl+J to toggle the madVR OSD). Fullscreen detection is done by enumerating windows.

Xaurus

24th July 2011, 23:01

Is it only me or does the "delay exclusive mode for 3 seconds" not work anymore?

Blight

24th July 2011, 23:03

madshi:
There's a clear difference between pressing 'Alt+Enter' and using another keyboard event or even a mouse click on the GUI fullscreen button.
Are you sure there isn't any debug hooks that's maybe left in there? I tried changing the other components in the graph to make sure they didn't hook the Alt+Enter key, but it didn't make a difference.
The thing is, when you hit alt-enter, the whole desktop blinks, as if switching resolution or entering exclusive mode.

pankov

24th July 2011, 23:05

Blight

24th July 2011, 23:52

Madshi:
I verified one thing, if I switch the video to fullscreen by calling WinAPI's SetWindowPos and giving it the fullscreen coordinates of the second monitor and then set the player window state to 'maximized', MadVR gets unstable (weird crashes).
I wrote a work-around that first moves the window to the new monitor without resizing it, followed by enabling the maximized window state and finally switching the window resolution to fullscreen, but it's not the most elegant solution.

Still, there's something else going on as well, as the playback sometimes (not always) freezes (no audio or video), requiring a pause/play to get it going again.

6233638

25th July 2011, 00:35

0.69 seems to be a good improvement over 0.67. I can play Blu-ray again, and the internal MPEG-2 decoder works great for most DVDs and allows for subtitles to be displayed properly. (using ffdshow as an external filter causes macrovision errors, LAV did not display subtitles)

I am unfortunately still having DVDs be indefinitely paused when "delay playback start until render queue is full" is enabled.

I don't get a seek bar in fullscreen with DVDs either (this has been around for ages) and some titles show significant amounts of combing - this is film, not video. With most titles, using the internal MPEG-2 decoder works fine and extracts the full resolution from the DVD without artefacts though.

I also tried the resolution changer again tonight after leaving it for a few versions, and it still doesn't work for me.

With DVDs, it generates a macrovision error so playback cannot start, and I am still getting a black screen with MPC-HC crashing when playing anything else. This is switching from 1080p60 to 1080p24 with Blu-ray.

My filter setup for DVDs is even simpler now thanks to this update: LAV Splitter LAV Audio Decoder ffdshow Audio ProcessormadVR does the MPEG-2 Decoding, and "Microsoft DTV-DVD Video Decoder" & "DVD Navigator" get added to the filter list automatically in MPC-HC when playing back DVDs.

Wile-E-Coyote

25th July 2011, 01:51

My filter setup for DVDs is even simpler now thanks to this update: LAV Splitter LAV Audio Decoder ffdshow Audio ProcessormadVR does the MPEG-2 Decoding, and "Microsoft DTV-DVD Video Decoder" & "DVD Navigator" get added to the filter list automatically in MPC-HC when playing back DVDs.
Are you sure madVR does the decoding? From what you're saying it sounds more like Microsoft's Decoder is doing it, otherwise why would it be in the list?

Luv

25th July 2011, 08:04

[About the madVR mpeg2 decoder]
Are you sure that the internal decoder is used? I wouldn't expect it to output 4:2:2. What does the madVR OSD say exactly? Please check whether there's still a decoder filter in the chain.

Don't bother,Madshi,it probably works as you said.I'm just stuck with the Cyberlink one for the moment:it seems it's the default for dvds when present on the system and I don't know how to switch to another one right now.
Other than this,no problems.

SoupRKnowva

25th July 2011, 09:15

so all these problems with input levels and such are completely eliminated by using the internal filters correct? cause i assume that they know exactly what to do since they are built in?

Also, since this is my first post ever, i want to thank you madshi for building madVR, loving it! and i love the internal decoders since they do 10 bits, so i dont have to worry about using goofy versions of ffdshow to accomplish it.

6233638

25th July 2011, 15:16

Are you sure madVR does the decoding? From what you're saying it sounds more like Microsoft's Decoder is doing it, otherwise why would it be in the list?Hah, you could be right there....

It seems to be the only decoder that reports the DVD framerate to madVR correctly for some reason.

I wonder why madVR isn't doing the decoding then.

madshi

25th July 2011, 16:11

madshi,
a while ago you asked us (users) what refresh rate do we prefer for 23.976 content 24Hz or 59.940Hz.
If I remember correctly the answer was definitely 24.
Tonight I've decided to go that route and stop fighting with custom resolutions and never exactly getting 23.976 but go for 24.000 which is pretty stable/exact on my PC but I was surprised to see that even the latest version of madVR still switches to 1080p59 when I feed it 23.976 while having the following display modes 1080p24, 1080p50, 1080p59 in madVR's settings.
Did I do something wrong or you simply didn't have the time to change the automatic display mode switching?
Forgot about it.

so all these problems with input levels and such are completely eliminated by using the internal filters correct?
Yes, there are no level problems when using the internal decoders. There should be no level problems when using external decoders, either, though, unless you have a very complicated setup.

madshi:
There's a clear difference between pressing 'Alt+Enter' and using another keyboard event or even a mouse click on the GUI fullscreen button.
Are you sure there isn't any debug hooks that's maybe left in there? I tried changing the other components in the graph to make sure they didn't hook the Alt+Enter key, but it didn't make a difference.
The thing is, when you hit alt-enter, the whole desktop blinks, as if switching resolution or entering exclusive mode.
madVR has a keyboard hook installed for its own key shortcuts, but I *definitely* do not look for Alt+Enter. I'm 100% sure on that. I don't know why Alt+Enter behaves differently for you than other key combinations.

Madshi:
I verified one thing, if I switch the video to fullscreen by calling WinAPI's SetWindowPos and giving it the fullscreen coordinates of the second monitor and then set the player window state to 'maximized', MadVR gets unstable (weird crashes).
I wrote a work-around that first moves the window to the new monitor without resizing it, followed by enabling the maximized window state and finally switching the window resolution to fullscreen, but it's not the most elegant solution.
That's really not elegant. Is this problem limited to ZP8? Or does it occur with ZP7, too?

Still, there's something else going on as well, as the playback sometimes (not always) freezes (no audio or video), requiring a pause/play to get it going again.
Yeah, when you move madVR to another monitor, madVR has to recreate all Direct3D stuff. Sometimes the timing of that doesn't work out properly, so that graph ends up in paused state. I need to improve that, of course.

cyberbeing

25th July 2011, 21:41

madshi, is there a full list of all the madVR keyboard shortcuts somewhere? Could it be added to the readme.txt in the next release?

Blight

25th July 2011, 21:47

madshi:
There's something definitely going on with the Alt+Enter issue. Is it possible for you to put a hook into your 'exclusive mode switch function' and press alt+enter in ZP and see if it somehow reaches that code, or triggers any other code that it shouldn't.
It may not even be your keyboard hook, but maybe some other hook in one of the libraries you're using.

The problem with the switch to fullscreen should happen with all versions of Zoom Player, but it may not affect all versions of windows the same. I'm experiencing the issue using Win7 32bit.

madshi

25th July 2011, 21:55

madshi, is there a full list of all the madVR keyboard shortcuts somewhere? Could it be added to the readme.txt in the next release?
You can search in the changelog for "Ctrl".

There's something definitely going on with the Alt+Enter issue. Is it possible for you to put a hook into your 'exclusive mode switch function' and press alt+enter in ZP and see if it somehow reaches that code, or triggers any other code that it shouldn't.
It may not even be your keyboard hook, but maybe some other hook in one of the libraries you're using.

The problem with the switch to fullscreen should happen with all versions of Zoom Player, but it may not affect all versions of windows the same. I'm experiencing the issue using Win7 32bit.
I know for a fact that I'm not doing anything special for alt+enter. I'm doing my main development testing with MPC-HC, and that is setup to go fullscreen with F11 (and not alt+enter), so on my PC alt+enter does not have any special meaning at all.

FWIW, I've today tested ZP8 (the latest RC build you PMed me) with switching between fullscreen and windowed and multiple monitors with exclusive mode on and off, and with the latest v0.70 I can't reproduce any problems, anymore. So maybe that version takes care of at least some of the problems. If not, we'll have to dig, but I can say right now that I can't reproduce any more problems here, I've really tried today.

pirlouy

25th July 2011, 22:06

I have a noob question (I'd love a guide for madVR to be honest, so if there's someone good in english, and which wants to share his knowledge, you'll help a lot of people):

How to know what our display can accept ? 16-235 or 0-255 ? 8 or 10 bits ? Is there a site which can help me ? Thanks in advance. For my TV, I let the default, 8 bits, 0-255, but maybe there's better... :/

madshi

25th July 2011, 22:07

madVR v0.70 released

http://madshi.net/madVR.zip

* delay times improved for option "pause playback until render queue is full"
* option "pause playback until render queue is full" is now disabled by default
* exclusive mode tweak options regrouped and branded as "not recommended"
* internal MPEG2/h264 decoders are enabled by default now, VC-1 stays disabled
* display mode switcher now "likes" 24.000 Hz mode for 23.976 fps movies better
* improved timestamps of internal libav/ffmpeg MPEG2 decoder
* improved seeking with internal libav/ffmpeg h264 decoder
* fixed: moving from one monitor to another sometimes caused graph to pause
* fixed: display mode changer caused internal decoders to stop working
* fixed: gamma processing with RGB input produced a black screen
* fixed: h264 mediatype parsing corrupted RAM, eventually resulting in a crash
* fixed: color processing accurateness problems introduced in v0.68
* fixed: exclusive mode seekbar didn't work during DVD playback
* fixed: tracks with a "|" in the name confused the tray icon context menu
* fixed: at playback start, and after unpausing there was sometimes judder
* fixed: OSD sometimes flickered at playback start, and after unpausing
This is mainly a bugfix release, with some minor improvements. I hope that the biggest bugs are all fixed now. If not, let me know. FWIW, I might have to slow down madVR development a bit in the next couple of weeks. So don't expect any major improvements in the next 1-3 months.

cyberbeing

25th July 2011, 22:18

So is this all of them?

Could you add keyboard shortcuts for enabling/disabling 3DLUT, enabling/disabling Gamma Correction and enabling/disabling Automatic Fullscreen Exclusive mode?

Current
Enable OSD - Ctrl+J
Reset OSD Statistics - Ctrl+R
Gamma Curve type - Ctrl+Alt+Shift+C
Increase Gamma - Ctrl+Alt+Shift+G
Decrease Gamma - Ctrl+Alt+Shift+F
Source Primaries - Ctrl+Alt+Shift+P
Decoding Matrix - Ctrl+Alt+Shift+M
Input Video Levels - Ctrl+Alt+Shift+I
Output Display Levels Ctrl+Alt+Shift+Y

Depreciated
Decoding Matrix (Color Format) - Ctrl+Alt+Shift+C

madshi

25th July 2011, 22:23

"Ctrl+Y" is a typo in the changelog, should read "Ctrl+Alt+Shift+Y", and that key combination is still current for changing the output/display levels. You can enable gamma correction by pressing one of the gamma related keyboard shortcuts twice. Why would you need to switch fullscreen exclusive mode on/off via a keyboard shortcut? A keyboard shortcut for 3dlut is a bit difficult because I'd practically have to "disable all calibration controls" to disable the 3dlut. Is that really the desired effect?

Edit: Ctrl+R resets the statistics in the OSD.

cyberbeing

25th July 2011, 22:35

Fixed: It looks like 0.70 fixed the frozen black screen w/ refresh rate changer and the internal decoders bug. :)

I guess you could make it enable/disable the Calibration page. What I want is the ability to switch on-the-fly between 'Calibrate this dispay by using...' and 'Disable Calibration Controls for this Display'.

As for the Fullscreen exclusive toggle, why not? It would be useful for testing playback differences between Windowed & Fullscreen, as well as being one less thing you need to enter the settings dialog for if you picked the wrong mode (like on my 7800GTX where it doesn't work well). If you can override the default MPC-HC Fullscreen Exclusive shortcut of CTRL+F (especially since it conflicts with madVR settings), that would be great.

You can enable gamma correction by pressing one of the gamma related keyboard shortcuts twice.
But how do you disable it once enabled?