madVR - high quality video renderer (GPU assisted)

[nevcairiel]

I concur, the desktop window manager (DWM) is always running on Windows 8, and Desktop Composition is always on. Its required for the Metro interface and some of its animated effects, so it simply cannot be turned off.

[Soukyuu]

Nope, you're wrong. Microsoft is removing more parts of this engine, the guy who is writing an aero glass replacement has to reimplement even more code for 8.1 (see this thread (http://www.msfn.org/board/topic/161788-aero-glass-for-win8-rc3/page__st__2160))

The DWM is there to stay, but compositing isn't.

[nevcairiel]

Just because they're breaking his Glass hack doesn't mean the compositor is gone. Its still there and here to stay.
Sure, they could be cleaning up some unused components, but they will not get rid of the whole thing. It does a lot more than just transparency for glass.

Edit:
Read here:
http://msdn.microsoft.com/en-us/library/windows/desktop/hh848042.aspx

Especially:

In Windows Vista and Windows 7, desktop composition is disabled in a number of scenarios. In Windows 8, DWM desktop composition is a core operating system component and cannot be disabled. With a few exceptions, desktop composition is always on; it’s started before the user logon and remains active for the duration of a session.

[leeperry]

@madshi: as it turns out, 1/2 120Hz BFI could apparently do the trick: Software-based black frame insertion improves motion clarity in MAME on 120 Hz displays (http://www.blurbusters.com/software-based-black-frame-insertion-improves-motion-clarity-in-mame/)

I'll try this modded MAME build on my CRT and I wonder how FRC in mVR would look with 1/2 BFI :confused:

Just for the record, there are more and more 120Hz capable 1080p displays, not just TN LCD, CRT and 720 pj's.....even Panasonic plasma's for that matter: True 120Hz from PC to TV:
Make Your HDTV Accept Real 120Hz Refresh Rate (http://hardforum.com/showthread.php?t=1766384)

[Soukyuu]

@nev: Maybe i'm misunderstanding what compositing means in this case then, but for me it's the act of blending windows with effects such as transparency/blur etc (isn't that what they call compiz etc in linux, compositor?). Since windows8 doesn't use any transparency/blur anymore (which is why desktop composition is usually disabled, to save resources), can we still say it's active? I guess since the MSDN says it cannot be disabled, my definition is wrong/not the same as Microsoft's

[nevcairiel]

@nev: Maybe i'm misunderstanding what compositing means in this case then, but for me it's the act of blending windows with effects such as transparency/blur etc (isn't that what they call compiz etc in linux, compositor?). Since windows8 doesn't use any transparency/blur anymore (which is why desktop composition is usually disabled, to save resources), can we still say it's active? I guess since the MSDN says it cannot be disabled, my definition is wrong/not the same as Microsoft's

Just because it doesn't blur or use transparency anymore, doesn't mean its disabled. It does more things. Its a rather wide term and can mean any effects done to the windows.

The article i linked even mentions a few of those, like Windows Flip (Win+Tab), and Window Previews on the Taskbar (Hover Taskbar Icon, see real-time preview), just to name two. Additionally its used for the Metro Start Menu and its live-tiles and other animations.

In any case, its active and here to stay.

[Keiyakusha]

Soukyuu

transparency and blur is just a visual effect. You can't conclude anything by seeing it missing/disabled. If I'll take some game that uses engine xxxxx and make it show only black&white graphic, you'll say that I disabled the engine? They can set windows on fire if they really want. But that may consume a lot of your GPU resources... so we can only dream of stuff like that.

Edit: BTW making portions of the screen black and white is an effect present in windows too. I forget where I saw it though, it's quite rare
Actually I found it, It happens when you mess with window appearance. It slowly dissolves to b&w during applying changes. But there was some more of it somewhere.
Edit2: besides, on win7 you can disable blur/transparency while keeping desktop composition ON. So I don't see where you got the idea that transparency=desktop composition.

[vivan]

But there was some more of it somewhere.left alt + left shift + prtscr (high contrast mode) maybe?

[Keiyakusha]

left alt + left shift + prtscr (high contrast mode) maybe?

Oh no, this changes the appearance. On win8 this is qualifies probably (on win7 it should disable desktop composition) but I mean simply effect that takes what on your screen as is and makes it b&w

[pie1394]

I'll try this modded MAME build on my CRT and I wonder how FRC in mVR would look with 1/2 BFI :confused:

Just for the record, there are more and more 120Hz capable 1080p displays, not just TN LCD, CRT and 720 pj's.....even Panasonic plasma's for that matter: True 120Hz from PC to TV:
Make Your HDTV Accept Real 120Hz Refresh Rate (http://hardforum.com/showthread.php?t=1766384)

I am just curious why CRT or PDP needs the black frame insertion from input signal... Isn't BFI the natural characteristic for impulse-type display devices? :rolleyes:

Higher driving rate on CRT / PDP helps to deal with flicking issue in addition to prevent motion judder.

About the motion resolution enhancement of BFI on PDP, Pioneer KURO's trick uses the 1st sub-pixel display timing with mutiple spatial sub-pixel domains to produce the longest BFI with still enough colors. Panasonic's NeoPlasma (since VT50) does not have spatial sub-pixel structure like KURO, but much higher temporal sub-pixel driving count + 3 brightness values per driving time of FFD.

The LED scanning back-light used on high-end LCDs creates better BFI effect than the BFI from 120 Hz input signal. Hitachi ever designed the (special high-quality) CCFL/HCFL flicking back-light on his TV to emulate the Impulse-type displaying.

[leeperry]

Well, FRC in mVR helps a lot on CRT IME and the higher the refresh rate the better.

Also, several ppl in this thread (http://forum.arcadecontrols.com/index.php?topic=130710.msg1339227#msg1339227) seem to enjoy that MAME mod on non-LCD displays:
Inserting a black frame makes the scroll perfectly smooth again for 120 Hz CRT screens too
They did post the mod in this thread and apparently it's done via HLSL, but I guess it wouldn't work with the current build of mVR as this would obviously need to be processed on the windows refresh rate and not the movie frame rate.

You mentioned "high-end LCDs", but these only come in large screens when I'm not interested in anything bigger than 25" so a poor man's BFI could still be useful.

There's only one way to find out how it'd look on any specific display, and it might very well interfere with mVR's FRC anyway....reason why my post was directed to madshi in the first place ;)

[Warlock]

Guys, what is this XySubFilter: https://www.dropbox.com/s/kzxrzz49nhop0n4/xySubFilter_TEST_v3.0.0.470-260613.7z? It is an improved version of xy-VSFilter? How to install it?

[michkrol]

Guys, what is this xySubFilter: https://www.dropbox.com/s/_/xySubFilter_TEST_v3.0.0.470-260613.7z? It is an improved version of xy-VSFilter? How to install it?

xySubFilter will be the new version of xyVSFilter, which uses the madVR's new subtitle interface, however no official beta is available yet, making this file suspicious. Even if it's xySubFilter it's highly likely it's unstable, I would leave it alone.

[kasper93]

@Warlock:

First of all this is not XySubFilter or xy-VSFilter thread! Please don't spam if your question is not related directly to madVR.

You should ask your question in the place where you found this link. XySubFilter is not even in public test yet (or I miss something), so you shouldn't use it unless you know what you're doing and apparently you don't.

It is an improved version of xy-VSFilter?
Sort of, it uses new subtitle interface which is supported by madVR. Point of that it to get HQ subtitles. Basically text subtitle quality is now same as in mpc-hc's ISR.

How to install it?
Exactly the same as xy-VSFilter.

Lastly I would be careful about this .dlls because hash 260613 is not reffered on official XySubFilter repo. Basically you don't know what is it. Or maybe you know, but I don't know where you get that.

[sixor]

Goto 'Devices' "Your device" 'Display modes'
untick 'switch to matching display mode'

that option is not there

you mean on madvr settings?? nop, i can assure you

[cyberbeing]

260613 is not reffered on official XySubFilter repo
It's probably just a date, June 26th 2013.

Anyway, that 3.0.0.470 (assuming it was built from our latest git) introduced a couple new bugs which didn't exist in the prior versions. The change to DVB broke some DVB subtitles. The change to registry setting names broke some of those as well. The fix for YCbCr Matrix didn't work. And so on... XySubFilter 3.0.0.458 from June 24th is much more stable than that one. So if anybody wants to test something now, stick with 3.0.0.458 until we release a beta in within the next few weeks.

I also notice that the build linked is a VS2012 build, which means all the project files are slightly outdated, since we've only been keeping the VS2010 ones up-to-date with XySubFilter changes. xy-VSFilter from the XySubFilter branch is also relatively broken compared to 3.0.0.211, and the consumer is also not really intended for end-users currently and won't be included with our beta release.

[kasper93]

It's probably just a date, June 26th 2013.

Silly me :o

[huhn]

@madshi: as it turns out, 1/2 120Hz BFI could apparently do the trick: Software-based black frame insertion improves motion clarity in MAME on 120 Hz displays (http://www.blurbusters.com/software-based-black-frame-insertion-improves-motion-clarity-in-mame/)

I'll try this modded MAME build on my CRT and I wonder how FRC in mVR would look with 1/2 BFI :confused:

Just for the record, there are more and more 120Hz capable 1080p displays, not just TN LCD, CRT and 720 pj's.....even Panasonic plasma's for that matter: True 120Hz from PC to TV:
Make Your HDTV Accept Real 120Hz Refresh Rate (http://hardforum.com/showthread.php?t=1766384)
why adding black frames have a look at this:
http://www.blurbusters.com/zero-motion-blur/lightboost/

[pie1394]

You mentioned "high-end LCDs", but these only come in large screens when I'm not interested in anything bigger than 25" so a poor man's BFI could still be useful.

Seiki To Unveil $699 39-inch Ultra HD LED LCD TV (http://www.twice.com/articletype/news/seiki-unveil-699-39-inch-ultra-hd-led-lcd-tv/107481)

Any interest to modify this US$699 39" 4K LCD monitor with BFI insertion ? :D

The viewing angle of S-MVA should be just fine if you don't watch it from > 20-degree off-axis angle.

If I am not wrong, it is Innolux native 8-bit 120Hz S-MVA panel, which is at least better than the native 6-bit + FRC 120Hz TN-type one~~~ Unfortunately its input signal module does not provide 4K 60 fps mode. So it needs extra hacks to use different signal processing board to Tcon(panel driving) board...

[clsid]

@cyberbeing

xy-VSFilter and xySubFilter are currently being developed as two separate filters. I see that some code was added to prevent both from loading at the same time. Will the correct one be loaded automatically based on the renderer that is used? It would be annoying if you need to adjust preferred filter settings or whatever when switching renderers. MPC-HC's graph builder could potentially be used to automatically prefer the correct sub filter, based on the selected renderer.

MPC-HC is capable of blocking VSFilter when using the ISR. It could do the same when using MadVR if xySubFilter is available. It's simple and requires just a few lines of code. That 'solution' might be safer than adjusting the loading settings of VSFilter, because such settings may not get restored properly in case of a crash (of the player or one of the filters). Other players that use madVR would need to do the same, but I think that should not be a problem.

[cyberbeing]

xy-VSFilter and xySubFilter are currently being developed as two separate filters. I see that some code was added to prevent both from loading at the same time. Will the correct one be loaded automatically based on the renderer that is used? It would be annoying if you need to adjust preferred filter settings or whatever when switching renderers.
I haven't tested it yet, so I'm not sure how it behaves. Those commits are only experimentation on the idea at this point.

MPC-HC's graph builder could potentially be used to automatically prefer the correct sub filter, based on the selected renderer.

MPC-HC is capable of blocking VSFilter when using the ISR. It could do the same when using MadVR if xySubFilter is available. It's simple and requires just a few lines of code. That 'solution' might be safer than adjusting the loading settings of VSFilter, because such settings may not get restored properly in case of a crash (of the player or one of the filters). Other players that use madVR would need to do the same, but I think that should not be a problem.

Yes, I already suggested exactly this to JEEB yesterday, but as MPC-HC doesn't yet support doing this YuZhouHuang had already went ahead and implemented the other idea I mentioned to him a few days ago (not yet tested). I'd much prefer to not need to mess with VSFilter settings, but both being present in the graph at the same time is indeed a problem.

At first we were just thinking of telling users to uninstall VSFilter when testing the upcoming XySubFilter beta, but this isn't really practical since VSFilter would still be useful outside of madVR, not to mention that most codec packs do not offer a way to disable VSFilter without uninstalling the entire pack. We need a solution of some kind for this, and I do prefer the idea of a player-based blocking method based on subtitle consumer support if at all possible.

[dansrfe]

@madshi

Here is the updated log file related to the backbuffer cycling between empty/full. I've tried all combinations and permutations of flush settings in windowed mode to no avail.

http://www.sendspace.com/file/ururx3

[zero9999]

I haven't tested it yet, so I'm not sure how it behaves.

It seems to work halfway properly here. xy-vsfilter is loaded into the graph even if madVr and xySubFilter are being used, but doesn't render any subtitles onto the video. xySubFilter only loads with madVr and in fact it seems I can't even block xySubFilter in MPC-HC at all, which is kind of inconvenient for testing.

Otherwise r470 is working great here, the performance issues appear to be fixed. Great job!

[6ari8]

It is really weird. I tried going back to V0.86.1 and the problem didn't go away. Still is in v0.86.6

Do you have any idea what could be causing this? It happened suddenly because I'm 100% sure it wasn't there a little while before updating to v0.86.3. I did update my Nvidia driver a while ago though.

I updated the Nvidia driver, removed all codecs, madvr, and mpc-hc then reinstalled but nothing has changed.

I found that the only combination that produces this output is the following:
1. Lanczos 8 taps.
2. Activate anti-ringing filter.
3. Use full screen (either windowed or exclusive mode)
4. Use on external monitor.


Gotta ask, why are you using linear light with upscaling? I've found it does nasty things to lines.

Can you give some examples? Because I don't see anything bad caused by it.

[madshi]

It works like it always worked, it outputs over HDMI whatever the HDMI device requests. Sadly most TVs only request/offer limited RGB in their HDMI handshake, even if they allow and are configured for full RGB.
It can output full RGB just fine, for example when i connect my PC screen over HDMI, it just has no toggle to override the HDMI handshake (or well, it has one, but it doesn't work)
I'm not sure EDID really says anything about which RGB levels the display wants? I did read the EDID spec some time ago and I don't remember having seen anything like this? But maybe I missed it. If this info is really in there, shouldn't an EDID override help?

I tried increasing the backbuffers for the old path FSE but the OSD is still showing the default 3 being used.
Am I missing something or is this a bug?
The old path uses (has to use) Direct3D9 instead of Direct3D9Ex, and Direct3D9 is limited to only 3 backbuffers.

when playing back an old NTSC video, 29.97fps (progressive) it says I get 1 frame repeat in 2 hours, but the frame repeat count goes up about 12 per second...
Is deinterlacing active? Forced film mode? What is your refresh rate?

I wonder do you still remember about issue that results 720x480 video with 32:27 sar (and possibly other cases, I don't remember well myself) being displayed as, for example 0,0,1919,1080 (target rectangle)?
I do remember it. It's really the fault/responsibility of the media player, but fixing this inside of madVR might be possible. But I consider it a cosmetical issue, so it's not a high priority for me at the moment. But feel free to create a bug entry in the madVR bug tracker for this. Although it's not really a bug, I would accept it.

Only "problem" i have is that in exlusive fullscreen mode any other video on a second monitor stops playing
That's a Direct3D limitation. Nothing I can do about it. The same thing should also happen if you start e.g. a game on the first monitor in fullscreen exclusive mode.

maybe there's something I'm missing here, but it seems that it's impossible to fast forward when bitstreaming audio to the avr. Is there anyway around that except for not bitstreaming audio to the avr?
Is this in any way madVR related? I don't think so, because madVR doesn't really have anything to do with audio.

Yes, and on my GTX 660, GPU-Z shows only 10% usage when a video is playing with highest madVR settings, so there's room for even more demanding algorithms in the future.
Well, for 60i or 60p content, GPU usage would probably be higher than 10%. Of course 24p playback is much easier on the GPU.

I decided to try fullscreen windowed mode instead of fullscreen exclusive mode since I started using DisplayFusion Pro and in exclusive mode all focus gets taken by madVR. So, interesting observation. I have 59Hz, 50Hz, and 48Hz in display mode changer. These are all custom resolutions set within NVidia control panel. Now when going into fullscreen windowed mode the composition rate stays at 59Hz (and stays at that for exclusive as well) and the refresh rate is changed to 48Hz for this particular video.
That's a problem. In Vista and win7 I solved this by switching Aero/Desktop Composition off and on again. Doing so made sure that the composition rate is updated. Unfortunately win8 doesn't allow me to do that, anymore. I don't know any way to make the composition rate update in win8, unfortunately. You may have to close and restart the media player. Do you then get the proper composition rate?

is it normal to get 1000+ repeated frames in the first 2 minutes of a video with madvr deinterlace on?
There are situations when deinterlacing is active which can confuse the "repeated frames" counter. If you want, you can create a bug entry in the madVR bug tracker for this.

as it turns out, 1/2 120Hz BFI could apparently do the trick
Normally you say "something does the trick" in the context of a discussion about a specific problem. Now you use this phrase without any context.

Backbuffer emptying out was a problem on my end. Everything is working well with v0.86.6.
Here is the updated log file related to the backbuffer cycling between empty/full. I've tried all combinations and permutations of flush settings in windowed mode to no avail.
You gotta make up your mind. Was it a problem on your end or not? If you have this problem again: Have you tried overlay mode? Do you have desktop composition turned on or off?

I updated the Nvidia driver, removed all codecs, madvr, and mpc-hc then reinstalled but nothing has changed.

I found that the only combination that produces this output is the following:
1. Lanczos 8 taps.
2. Activate anti-ringing filter.
3. Use full screen (either windowed or exclusive mode)
4. Use on external monitor.
So it does not occur on the primary monitor? I would like to look into this, but I can't seem to reproduce the problem on my PC. Does it occur with all zoom factors and with all videos?

[leeperry]

Normally you say "something does the trick" in the context of a discussion about a specific problem. Now you use this phrase without any context.
context (http://forum.doom9.org/showpost.php?p=1623007&postcount=18227) to refresh your memory :p
good displays seem to use even higher refresh rates for BFI/DFI than 120Hz, and they do this internally. So I wonder if this even makes sense to build into madVR.
I realize that feature requests might take a while before being even considered, I just wanted to point out that 1/2 120Hz BFI seems to work fine but I kinda suspect it would mess up with mVR's FRC.

why adding black frames have a look at this:
http://www.blurbusters.com/zero-motion-blur/lightboost/
It looks amazing and apparently would measure as well as a CRT(from what my last link says anyway) but it only works with über-pricey TN's.

[dansrfe]

@ madshi

I was trying to figure out if it was a LAV or madVR problem before but I've now realized that this problem kind of just started happening on its own and not really related to upgrades. I have been having some other subtle issues with explorer.exe in the past couple of days but strangely enough I end up having to restart explorer shortly after or before using MPC-HC but other times it's at random. This may be affecting the composition rate problem as well.

EDIT: Overlay mode works flawlessly and the backbuffer remains filled up but the problem with it is that it doesn't apply the ICC profile. Later on I intend to calibrate the monitors and use 3DLUTs but until that time I'll have to use ICCs. Overlay mode has a lot more slicker feel to it for some reason. And I don't have the explorer.exe issues/slowdowns when using Overlay either.

Also, what does "use a a separate device for presentation" mean and does it have an effect on windowed mode presentation with respect to the issue I'm having? It seems to stabilize windowed mode (non-overlay) a little bit.

[cyberbeing]

It seems to work halfway properly here. xy-vsfilter is loaded into the graph even if madVr and xySubFilter are being used, but doesn't render any subtitles onto the video. xySubFilter only loads with madVr and in fact it seems I can't even block xySubFilter in MPC-HC at all, which is kind of inconvenient for testing.
Yes, this was the behavior up to this point. The problem is that xy-VSFilter being connected a the video output pin at all has a performance overhead, and it restrict some of the color formats which can connect to madVR.

Otherwise r470 is working great here, the performance issues appear to be fixed. Great job!

FWIW, the changes which added blocking were in r476 but unfortunately it doesn't work as expected at all and ends up blocking XySubFilter entirely from entering the graph, rather than the other way around. Back to the drawing board... Hopefully the enhancement to MPC-HC graph building which was talked about ends up panning out.

[Edit: On 6-27-2013 some commits were reverted and re-pushed to fix a couple bugs, but some issues with these latest commits still exist, so this will likely happen again soon]

[6ari8]

So it does not occur on the primary monitor? I would like to look into this, but I can't seem to reproduce the problem on my PC. Does it occur with all zoom factors and with all videos?

No, it only occurs on an external monitor and with any video. I'm sorry but I'm not sure what you mean by zoom factor.

Also, my laptop has two relevant outputs: MDP, hardwired to the discreet GPU (GT 650M) which I use with an adapter for HDMI, and a HDMI port, hardwired to the integrated GPU.

I primarily use the MDP to produce a correct 23.976HZ successfully with custom resolution because I can't change the integrated intel GPU refresh rates. Thinking the problem may be caused because of the Nvidia driver, I tried the HDMI port today and it produced the same output. Hope this helps.

Thanks!

[madshi]

URL="http://forum.doom9.org/showpost.php?p=1623007&postcount=18227"]context[/URL] to refresh your memory :p
Well, my reply back then is still mostly true today.

EDIT: Overlay mode works flawlessly and the backbuffer remains filled up but the problem with it is that it doesn't apply the ICC profile. Later on I intend to calibrate the monitors and use 3DLUTs but until that time I'll have to use ICCs. Overlay mode has a lot more slicker feel to it for some reason. And I don't have the explorer.exe issues/slowdowns when using Overlay either.
Ok, then you have a solution for the future.

Also, what does "use a a separate device for presentation" mean and does it have an effect on windowed mode presentation with respect to the issue I'm having? It seems to stabilize windowed mode (non-overlay) a little bit.
It helps some people. If it works for you, then use it.

No, it only occurs on an external monitor and with any video. I'm sorry but I'm not sure what you mean by zoom factor.
You've told madVR to use Lanczos8 for upscaling. But madVR only upscales if it's necessary. E.g. if your display is 1920x1080 and if the original video is also 1920x1080, then madVR does not have to scale at all. If you have a 4K display, though, then suddenly madVR has to upscale the original video to 2x size, so you have a 2x zoom factor.

In MPC-HC you can zoom by using the num keypad numbers 9 and 1. Try playing with these to find out if the problem only affects certain zoom factors. Also, in MPC-HC you can set the media player to different "Video Frame" modes which will also directly affect the zoom factor. E.g. if you use "Video Frame -> Normal Size", madVR doesn't have to do any zooming/scaling at all (unless the video needs aspect ratio correction). If you use "Video Frame -> Double Size" you get a scaling/zoom factor of 2x etc...

[huhn]

context (http://forum.doom9.org/showpost.php?p=1623007&postcount=18227)
I realize that feature requests might take a while before being even considered, I just wanted to point out that 1/2 120Hz BFI seems to work fine but I kinda suspect it would mess up with mVR's FRC.

i made a 120 hz vid from a 30p vid with an simple

[f1 b1 f1 b1] [f2 b2 f2 b2]

as long as no frames drop the picture is watchable it's super smooth and i can't see any motion blur but! the hole picture is very dark (ofc if i hit pause on an not black frame the frame looks normal) and you can see overdrive artefact's anyware.
the picture looks like its blended the hole time with an black frame.
the picture quality is so low is a waste of time nothing more.

i may make an vid with 60 fps and

[p1 b1][p2 b2] and use frc always on and 120 hz

i used an asus vg248 for that test my that's the problem but ...

[madshi]

i made a 120 hz vid from a 30p vid with an simple

[f1 b1 f1 b1] [f2 b2 f2 b2]

as long as no frames drop the picture is watchable it's super smooth and i can't see any motion blur but! the hole picture is very dark (ofc if i hit pause on an not black frame the frame looks normal) and you can see overdrive artefact's anyware.
the picture looks like its blended the hole time with an black frame.
Yeah, but try this:

[clip(f1*2), f1*2 - clip(f1*2)]

Of course you'll need to use linear light. And overdrive artifacts should still be visible...

[leeperry]

i made a 120 hz vid from a 30p vid with an simple

[f1 b1 f1 b1] [f2 b2 f2 b2]

as long as no frames drop the picture is watchable it's super smooth and i can't see any motion blur but! the whole picture is very dark
Yep, same goes for BFI on Sammy TV's whom user manual is very clear about the fact that you need to increase the backlight brightness in order to compensate.

Many computer monitors such as BenQ really don't seem to do anything more than 1/2 BFI at 120Hz so the concept works, as proven by that MAME custom build.

Anyway, I'll try that MAME build everyone raves about and will report back. I still think that this would be an interesting addition to mVR as we could always try 1/2, 1/4 or 1/5 BFI at 120/144Hz and it wouldn't appear to be rocket science via HLSL, but madshi doesn't seem too convinced anyway so let's call it a day then and it's not the right time for feature requests either way. I guess mVR's FRC smoothness would suffer in the process, I don't see how it wouldn't tbh.

[huhn]

sry can't follow you i'm not good with avisynth.

i can do the basics nothing more.

that's how i did it:

a = FFVideoSource("G:\New folder\30p.mkv")
c = FFVideoSource("C:\black.mkv")
b = FFVideoSource("C:\black.mkv")

GScript("""
for(i=0, 3599, 1){
b = b + a.trim(i, i) + c.trim(0, 0) + a.trim(i, i) + c.trim(0, 0)
}
""")

return b


encoding remuxing ready to go.
the black.mkv is a 1 black frame mkv with 30 fps.

i checked it frame by frame so it work out after i cut out the first part which was 2 time the original video X-).

[f1 b1 f1 b1] [f2 b2 f2 b2]
so this was an very bad explanation sry

a = FFVideoSource("G:\New folder\30p.mkv")
c = FFVideoSource("C:\black.mkv")
b = FFVideoSource("C:\black.mkv")

GScript("""
for(i=0, 3599, 1){
b = b + a.trim(i, i) + c.trim(0, 0)
}
""")

return b
3599 is the frame count of the video -1

this should return a 60 fps video with every second frame black.

[nevcairiel]

My TV has a built-in BFI mode as an alternative/addition to the interpolation modes, but the loss of brightness and the slight "strobing" effect doesn't make it look all that great, IMHO.

[leeperry]

Yep, apparently most/all Sammy TV's dont synchronise their BFI with 24 and 50Hz, duh.....I didn't have the patience of trying 60Hz, but obviously desynced BFI will look utterly hiccupy.

[leeperry]

this should return a 60 fps video with every second frame black.
This AVS call on a 60fps source file should give a nice and smooth 120fps output: Black (or dark) frame insertion ? (http://forum.doom9.org/showthread.php?p=1237611#post1237611)
Original = Avisource("Yourmovie.avi")
Black=Original.Blankclip()
Interleave(Original,Black)
Selectevery(2,0,1,0,1)

[huhn]

i got already a film black film black film black 120 fps from 30 fps.

but i try that out should be a lot faster than the random script i created.

[pie1394]

The A-FRC + Overdrive on 120Hz LCD panel could cause some unwanted effects. But it definitely improves the LCD G-2-G response if black color value is inserted every 1/120 of 1/60 second. If I don't remember it wrong, the crystal cell needs to be turned to black color first before it is changed to another new color.

Hope the design by scanning backlight may give you guys some ideas how to do BFI with the method from 120 Hz input signal.

The scanning backlight of Sony MotionFlow "Clear" on HX850/HX920/HX950 looks like this picture, but without interpolated picture. It still repeats the same 60fps content's picture 3 times every 1/240 second.

http://www.sony.jp/products/picture/y_KDL-55HX920_049.jpg

With "Clear+" option, I think this TV inserts more backlight off time since its brightness is lower than the "Clear" mode's. I feel that on my 65HX920 and it yields better motion resolution even on 24 fps contents.


On HX850 / HX950, there is a new MotionFlow option called "Impulse". It turns off the backlight at 75% time of every 1/60 second. The backlight is turned on at the last 1/4 part of every 60fps frame time, which guarantees the LCD has been stable on desired colors. It provides the best motion resolution for 60Hz FPS games --- but with CRT-alike flicking effect.

http://www.sony.jp/products/picture/y_KDL-55HX950_013.jpg

HX750 actually has the same "Impulse" option. It has the same effect.

http://www.sony.jp/products/picture/y_KDL-55HX750_009.jpg


I think HX750's "Clear+" option works like this since its backlight control can only work as top-down half blocks. Backlight is turned off on 50% time of every 1/60 second:

http://www.sony.jp/products/picture/y_KDL-55HX750_008.jpg

[dansrfe]

Isn't a part of motion blur due to lower camera shutter speeds?

[Asmodian]

That is one kind of motion blur and obviously not influenced by what the TV does.

There are other sources too though. One is due to the to the time it takes a pixel to change color so turning off the backlight during the change helps a lot. There is also motion blur from your retina taking time to change, turning off the backlight helps this too.

For a 24fps film I, personally, don't think all this backlight flickering helps much. For gaming or sports (faster shutter speeds) the benefit goes up.

It would be interesting to see how much MadVR could do with BFI as part of the FRC on a 120+ Hz display.

[Q-the-STORM]

Is deinterlacing active? Forced film mode? What is your refresh rate?

I can't believe I didn't test deactivating deinterlacing... seems like this was this "confused counter problem" you talked about in another post.. I deactivated deinterlacing now, so the counter stays at 0 now..



Just for the record, there are more and more 120Hz capable 1080p displays, not just TN LCD, CRT and 720 pj's.....even Panasonic plasma's for that matter: True 120Hz from PC to TV:
Make Your HDTV Accept Real 120Hz Refresh Rate (http://hardforum.com/showthread.php?t=1766384)
well, the only Panasonic that "apparently" can do 1080p@120Hz is the VT50, but this was never confirmed... the post only says
Some friend of mine told me his in law is running the Panasonic VT50 at a forced 120hz refresh rate at 1080p without problems. Just more heat output. I cannot confirm what he says but I guess that gives us hope. which does not sound credible...

I actually tested a bit on my European VT50 and failed... but I only tested PowerStrip on 1920x1080 @120Hz, @96Hz and @72Hz... When I got time I will test the other methods.. I will also test 1280x720@120Hz, which will be useless for my purposes but will definitely be interesting for some folks...
the European and US VT50 are very different though (why Panasonic, why...) so there might be very different results on the US model...

[leeperry]

It might require custom timings(the same way 72/75Hz over 165Mhz SL-DVI does (http://hardforum.com/showthread.php?t=1670817)) and of course you'll need a 340Mhz TMDS connection(but I'm sure you didn't overlook that point).

Anyway, I already raised the 120Hz 1/2 BFI subject a while ago and simply wanted to elaborate on that matter, madshi is the sole decider on what features to implement and he has made clear that this is not the right time for features requests.

By the time mVR 1.0 will be out and we can whine for BFI again, we'll prolly have more 120Hz displays to play around with....so until then I'll keep drooling at how smooth mVR's FRC looks on a 96Hz CRT :cool:

[rurifan]

That is one kind of motion blur and obviously not influenced by what the TV does.

There are other sources too though. One is due to the to the time it takes a pixel to change color so turning off the backlight during the change helps a lot. There is also motion blur from your retina taking time to change, turning off the backlight helps this too.

For a 24fps film I, personally, don't think all this backlight flickering helps much. For gaming or sports (faster shutter speeds) the benefit goes up.

It would be interesting to see how much MadVR could do with BFI as part of the FRC on a 120+ Hz display.

With modern LCD displays, the most major source of motion blur is caused by the human visual system interpolating smooth motion while the image is actually staying in place for long periods of time. You end up perceiving multiple images.

This problem exists with all low frame rate sources that have to show the same frame for extended durations, but it's particularly bad on LCDs because they display each frame for the entire duration (Sample And Hold).

I played with a Sony W802A display which has an Impulse backlight mode -- it strobes without doing any frame interpolation. It's flickery and very dim, but the resulting motion clarity is shocking. The car test in PerPixAn (http://www.prad.de/en/monitore/testsoftware/pixperan.html) was crystal clear. (It also revealed the actual LCD response time trail, which I'd never seen before.)

Also check out this summary of various motion artifacts. (http://www.blurbusters.com/faq/lcd-motion-artifacts/)

(I've been obsessing over poor motion reproduction lately... would love it if madvr could help. :) )

[rurifan]

I just discovered that my HDTV is capable of displaying RGB 4:4:4 color when the HDMI input is named "PC". Apparently I've been seeing 4:2:2 forever. (Check this link (http://forums.guru3d.com/showpost.php?p=4557704&postcount=25) if you're curious.)

Now I'm wondering how madvr fits into this situation. Do I need to tell it my output colorspace is different? Or did madvr assumes RGB 4:4:4 all along? The desktop colors look a bit different.

(Note: this is a different issue than the 16-235 vs 0-255 levels.)

[Asmodian]

MadVR was 4:4:4 all along, the colors might look different due to video card or TV differences in the different mode.

[madshi]

Now I'm wondering how madvr fits into this situation. Do I need to tell it my output colorspace is different? Or did madvr assumes RGB 4:4:4 all along? The desktop colors look a bit different.
madVR always outputs RGB 4:4:4. However, the GPU might convert that to YCbCr 4:2:2 or 4:4:4 behind madVR's back, depending on your GPU control panel settings, and eventually on your display's EDID. Hopefully you can force the GPU control panel to output RGB 4:4:4.

If you get different colors depending on your display's settings, than that is probably not madVR related, but probably your display runs through different processing chains.

[mindz]

Im in the market replace my 5770. It needs to be future proof, in that it has to be able to output @ 4K with Jinc 4 AR for chroma and image. Which would you guys advise me? Thanks!

[huhn]

wait until you got an 4k display don't buy an card yet.

[rurifan]

madVR always outputs RGB 4:4:4. However, the GPU might convert that to YCbCr 4:2:2 or 4:4:4 behind madVR's back, depending on your GPU control panel settings, and eventually on your display's EDID. Hopefully you can force the GPU control panel to output RGB 4:4:4.

If you get different colors depending on your display's settings, than that is probably not madVR related, but probably your display runs through different processing chains.

The nvidia control panel Digital color format only has "RGB" and "YCbCr444" options. Does this suggest that many people with HTPCs are outputting broken color from madvr?

I made a very simple test pattern (http://i.imgur.com/gTsKnxk.png).