MPC-HC tester builds for internal renderer fixes

[RGold]

Get sound loop when I do the following:

1. Player in Windowed mode
2. Drag&drop a video file and automatically start playback.
3. Switch to full screen - no exclusive.

Notes:
1. Using auto change full screen monitor mode. I.E, refresh will be 23.981Hz for 24 Hz video material.
2. Using latest build - didn't see this issue with 4504i
3. Using EVR Custom Pres. With LAV for everything.

[burfadel]

Latest build for me works on Windows 7 and 8. Went back to 7 as I thought there was a memory leak with Windows 8, turns out it was because of Avast 7.0.1442 beta lol. It does seem to change between files quicker now even on Windows 7 (not as fast as on 8 though) but still a nice change over previous versions.

[CruNcher]

I found it funny what recently happened on the Surface Keynote Sinofsky Surface Windows 8 RT freezed *g* i guess that one will go into the list of funny crash, freeze, bug moments in a Microsoft official Demonstration but they are prepared very well these days if such things happen ;)

Already started, hes though the best that could happen to the Windows Team including Russinovich he's inteligent,creative and sporty as well ;)

http://www.youtube.com/watch?v=4z-KXYUP5r8

[sexus]

It's currently being added to the trunk by Bobdynlan, give it a few weeks or so and it'll be there for everyone.

sounds most excellent ;)

[Virtual_ManPL]

Some more bugs:

1. I see no visual difference with Catmull-Rom spline4 and Catmull-Rom spline6. Probably the code is duplicated in both methods. (see test)

2. Nearest Neighbor code is probably swapped with Perlin Smootherstep, because Nearest Neighbor look better than Perlin Smootherstep, which it shouldn't and also Perlin Smootherstep is pixelated which should be Nearest Neighbor. (see test)

3. Lanczos is bugged, I see vertical lines on low quality video (SAMPLE (http://depositfiles.com/files/d9kpmu8qu)) and on zoom I see some kind of 'blocked pixel mask'
Bugged Lanczos (http://i.imgur.com/COzpr.png)
Not Bugged Bicubic (http://i.imgur.com/QIgba.png)

4. Automatic hiding menus don't work in some cases. The best way to reproduce is to do a circle mouse movements near Win start bar hooking also MPC-HC bottom bars and next go to fullscreen. The MPC-HC menus like Seekbar, Control and Status don't want to hide automatically if this steps are performed.



Now some resizer method test (not the same frame in all SS, but I was trying ;p) TESTED FILE (http://depositfiles.com/files/7xhbl0y9j)
Bicubic -0.6 (http://i.imgur.com/dfhUm.jpg)
Bicubic -0.8 (http://i.imgur.com/9ZnCU.jpg)
Bicubic -1 (http://i.imgur.com/yEzBX.jpg)
Bilinear (http://i.imgur.com/8BjnZ.jpg)
B-spline4 (http://i.imgur.com/SxekF.jpg)
B-spline6 (http://i.imgur.com/3USAB.jpg)
Catmull-Rom spline4 (http://i.imgur.com/2sPSM.jpg)
Catmull-Rom spline6 (http://i.imgur.com/Js4TI.jpg)
Lanczos2 (http://i.imgur.com/vRrsS.jpg)
Lanczos3 (http://i.imgur.com/2at7P.jpg)
Lanczos4 (http://i.imgur.com/mS04e.jpg)
Mitchell-Netravali spline4 (http://i.imgur.com/bsaoh.jpg)
Mitchell-Netravali spline6 (http://i.imgur.com/X2F1U.jpg)
Nearest Neighbor (http://i.imgur.com/mmxRE.jpg)
Perlin Smootherstep (http://i.imgur.com/2GdQL.jpg)

In my opinion Bicubic -1.0 still look the best, look on letters on the bottom-left side of the screenshot.
When video is very low quality we can use B-splite4, it greatly masks artifacts. Dunno why B-spline6 look worse, maybe too much sharp effect.
Also I was thinking that Lanczos should be the best, but looking on letters, it isn't. Odd because on normal images, it works the best in all graphic programs.

[burfadel]

I've noticed the OSD stuff doesn't disappear either, particularly the file name at the top of the screen (note this is Direct3D on another monitor). You can hide that by pausing and starting it again, upon which point 'play' may or not be stuck up there!

[JanWillem32]

@Hera: I've been able to replicate the problem. It's marked in the source code of the trunk (exactly at the point where the exception is made), and has been there since there were only VMR-7 and VMR-9 r.. It's an old problem caused by the quality of service (stats) interface in a DirectShow part of the player. It's not a real renderer issue, and I don't know if I can solve the problem anytime soon.

@RGold: I haven't been able to reproduce this issue yet. Does it occur when you disable the auto change full screen monitor mode feature?

@Virtual_ManPL: It's been a while since I made some screenshots of the resizers in action. Note that to test resizers, you need large magnification factors on lossless images.

The Lanczos filters have a truncation fault, which is native to the windowed sinc filters: http://forum.doom9.org/showthread.php?p=1540386#post1540386 .
Depending on the position of the source pixel(s) relative to the destination pixels(s), such filters will give brightness artifacts. In the case of equal resizing amounts for horizontal and vertical directions, the artifacts will be lines drawing squares on the picture.
Perlin Smootherstep has a square affinity. It simply works like that.
Nearest Neighbor for EVR CP is a placeholder for the EVR-inherited resizer. Its device is created by the display driver.

When the renderer is placed in fullscreen windowed mode, it skips the check for changes in the window bottom parameter. This is to prevent resizing because the bottom bar pushes the renderer window away. The correct solution would be to not use this trick and just let the renderer reset itself because the window resized (how it previously was), but I doubt people would like that. I know the fullscreen windowed mode is not bug-free (it never was), so if anyone wants to re-design the player menu to correct it, they're welcome to try.

@burfadel: I noticed it as well. It mostly happens when you switch from one file to the next in D3D FS mode. I'll take a look at why the OSD doesn't properly clear its status.

Development status: creating the code to allow the alternative scheduler on Vista in Aero desktop mode. (It already works for D3D FS, and on 7 in Aero desktop mode.)

resizers comparison pictures (44.4 MB): http://www.mediafire.com/download.php?5uhh2nha5iaczer

[burfadel]

Interesting enough, now that I wrote there is an issue with the OSD stuff not disappearing... its disappearing like it should! maybe it depends on the source of the file?

Admittedly I changed from a HD6950 to a HD7870 (Gigabyte OC model) Friday night, but I doubt whether that has anything to do with it since I don't use DXVA.

Seems its back again! Don't know why it worked flawlessly for a while...

[RGold]

@RGold: I haven't been able to reproduce this issue yet. Does it occur when you disable the auto change full screen monitor mode feature?



It does not occur when disabling the auto change full screen monitor mode feature.

[Virtual_ManPL]

@ JanWillem32 -
Thanks for resizers comparison test.

I'm curious why other resizers are so poor in my test.
No other method will smooth this 'stairs' on letters so good like the Bicubic -1.

I have an idea. How about making D3DFS default mode for going into fullscreen by binding Enter key etc.
It will prevent us for seeing this annoying artifacts in non D3DFS mode.

[Hera]

Performs even better on W8 with 3xx drivers.

[burfadel]

Just wondering whether there will be a new build any time soon, since its already over 220 builds behind the main branch (r5365 main vs r5142 JW at time of writing)...

Thanks :)

[ryrynz]

I've wondered that myself..

[JanWillem32]

It indeed has been a while.
I re-organized the EVR mixer handler for EVR CP. When running the debug DirectX runtime this is visible because of the lack of additional devices next to the main one.
I re-wrote many parts of the OSD system (the same as usual when I start working on something; it was in a pretty bad state).
I edited the subtitle renderer stats section.

Because the internal filters can't be compiled at this time because of a linker error, only 'lite' builds this time. I won't make a habit of this.
x64: http://www.mediafire.com/download.php?zlwfk87b3dm4bik
x86 SSE2: http://www.mediafire.com/download.php?mqcr33mc5u5xdma
x86 SSE: http://www.mediafire.com/download.php?551s34nyy4irwz1

[HoP]

a newbie question:
what is the work of this option?
http://imgair.net/i/clipboard01-1341568727.jpg

[JanWillem32]

On some systems the actual refresh rate is a bit different than the one reported by the DirectX 9 runtime. That one only reports whole integers, so I programmed it to adapt to the typical NTSC types already: double drr = static_cast<double>(rr);// the standard converter only does a proper job with signed values
m_dDetectedRefreshRate = m_dRefreshRate = m_CurrentRendererSettings.m_RenderSettings.dRefreshRateAdjust * ((rr == 119 || rr == 89 || rr == 71 || rr == 59 || rr == 47 || rr == 29 || rr == 23) ? (drr + 1.0) / 1.001/* NTSC adapted, don't include 95*/ : drr); // exact amounts
The user can refine the refresh rate adjustment factor even further. A good tool for this is PowerStrip. In my case, the refresh rate is reported as 79.939 Hz (at a setting of 80 Hz), so I use an adaptation factor of 0.9992375. Note that the alternative scheduler and constant frame interpolator can also make small adjustments when anomalies are detected.
I might be able to integrate display information functions to process this automatically in the future.

[HoP]

@JanWillem32
many thanks for info
another question
when i choose EVR CP,video is normal:

http://imgair.net/i/clipboard11-1341576353.jpg

but when i choose EVR Sync,video get zoomed.why??

http://imgair.net/i/clipboard12-1341576480.jpg

+
when i press ENTER for full screen,video get freeze for 1sec and then resize(EVR CP).

[gilic]

Don't use EVR-sync it is outdated and unmaintained. The dev left some time ago.

You could test different resizers under options>output, maybe your system isn't powerful enough for the selected resizer and causes the freeze. I think the resizers are sorted from simple at the beginning to complex at the end of the list.

[JanWillem32]

I haven't had the time yet to fully update EVR Sync yet. As you can see, I integrated the new resizer section, but the EVR Sync rending path isn't fully compatible with them.
The video renderer needs a reset like with any other resizing of the video window, and that simply takes a while. For switching the fullscreen borderless mode, the re-arranging of the bottom toolbar takes additional time (albeit much less than the reset). The reset duration of purely the resizer sections can be tested easily: just switch the resizer type on the fly. It hardly distorts playback.

[HoP]

@gilic & JanWillem32
many many thanks for info
goodluck

[burfadel]

Unfortunately the OSD stuff still doesn't disappear in full screen D3D mode for me with r5373.

[JanWillem32]

The OSD timer is still very unreliable, I'm trying to use a different one, but it isn't easy to find such a thing.

[shadewither]

@Virtual_ManPL: It's been a while since I made some screenshots of the resizers in action. Note that to test resizers, you need large magnification factors on lossless images.

The Lanczos filters have a truncation fault, which is native to the windowed sinc filters: http://forum.doom9.org/showthread.php?p=1540386#post1540386 .
Depending on the position of the source pixel(s) relative to the destination pixels(s), such filters will give brightness artifacts. In the case of equal resizing amounts for horizontal and vertical directions, the artifacts will be lines drawing squares on the picture.
Perlin Smootherstep has a square affinity. It simply works like that.
Nearest Neighbor for EVR CP is a placeholder for the EVR-inherited resizer. Its device is created by the display driver.

resizers comparison pictures (44.4 MB): http://www.mediafire.com/download.php?5uhh2nha5iaczer

I'd like to ask a few question about resizers backed on signal-processing:

1. Interpolation Coord:
e.g. For image 100x100 => 200x200, let corner pixel be (0,0)
** Is dst (0,0) = src (-0.25,-0.25) ?
** When interplating out of boundary, is source mirrored? e.g. source(-2,-3)=source(2,3)?

2. Downscaling resizers:
a) Downscaling a lot:
I suppose larger window = better?
b) Downscaling by > 0.5:
What is blurry and sinc-like resizer's characteristics?
Both blurring and ringing are affecting less dst pixels, but I wonder if they are getting worse or alleviated.

3. Is any resizer both intensity and gradient interpolating?

[JanWillem32]

x86 SSE: http://www.mediafire.com/download.php?234wpnwhnso4npo
x86 SSE2: http://www.mediafire.com/download.php?35vcolr2oz95jx3
x64: http://www.mediafire.com/download.php?zopg7v7e81c2cea

I've changed the timer (a type I actually like is used now). The OSD should be okay for now.
I added automatic monitor refresh rate detection to the system. It should activate for Vista and onwards. Note that setting "Refresh Rate Adjustment" to something else than 1 will disable the function. The effect can be seen in the stats screen, and I'd be glad to get some feedback on if it's working for others, too.

@shadewither: 1. It's (-0.5, -0.5). DirectX 9 sets vertices on pixel boundaries, not centres. http://msdn.microsoft.com/en-us/library/windows/desktop/bb219690%28v=vs.85%29.aspx
There are various states you can pin to a sampled texture stage. For example, you can even set this odd mix of settings:m_pD3DDev->SetSamplerState(2, D3DSAMP_ADDRESSU, D3DTADDRESS_CLAMP);
m_pD3DDev->SetSamplerState(2, D3DSAMP_ADDRESSV, D3DTADDRESS_WRAP);
m_pD3DDev->SetSamplerState(2, D3DSAMP_ADDRESSW, D3DTADDRESS_MIRROR); Note that for mirrorring, the mirroring happens on odd-plane coordinates. From the standard viewport, with mirroring states on U and V: the samles at the interval [(0, 0), (1, 1)] will be oriented the same as [(2, 2), (3, 3)]. [(1, 1), (2, 2)] will be mirrorred horizontally and vertically. Clamp, wrap and mirror once are a bit more straight forward. http://msdn.microsoft.com/en-us/library/bb174456%28v=vs.85%29.aspx

2. For large factor down-scaling, the typical usage is a (large) gaussian kernel. It just averages a lot of pixels and then outputs one blurred pixel.
For exact down-sizing with a factor of 2 or 3, just average the source pixels.
In a 3D renderer mipmaps are often used to scale images down, while keeping the same texture in memory. This way it's easy to re-use, even if the orientation and visible size of the object changes. http://en.wikipedia.org/wiki/Mipmap
"What is blurry and sinc-like resizer's characteristics?" I guess you mean 'Bicubic' and 'windowed sinc'?
Windowed sinc filters have a basis function, such as Lanczos (a sine type), and a window size. The basis function remains the same for all window sizes, you just add more samples with every larger window step (and less weight on the sample, caused by the base function). These functions only add up to 100% when you sample infinite samples.
The segmented filters, which includes bicubic types have a kernel with weights that always add up to 100%. These kernels don't simply expand by adding samples. The kernel for a spline4 and a spline6 are mathemathically different, and these filters often don't react the same way when extended to larger sampling areas.
Blurring is a welcome trait when down-sizing. Ringing is unfortunately pretty much the same for up- and down-scaling. When down-scaling, the artifacts from ringing will become slimmer, but still very present.

3. The interpolators (both resizers and frame time types) interpolate over the individual R, G and B channels. (Over the gamma slope in 8-bit mode, as a performance compromise. The stages are gammaless in the 10-, 16- and 32-bit surface modes.) I didn't use the gradient operators, as these are not supported on pre-ps3.0 adapters. I also don't think it would be useful to use these, either. (These are more a 3D thing, when you have to deal with orientation and size factors.)

[burfadel]

Any idea why the internal filters don't compile with yours but do with the normal builds?

[JanWillem32]

It's a recent development. I get a well-known, but annoying single linker error ("__image_base__" unknown on both x86 and x64). I don't know where it's coming from. __image_base__ is one of the symbols made by the external g++ compiler. I didn't edit any of the third-party code for the internal filters that gets compiled by g++. I'll have to search a bit, I guess.

[burfadel]

OSD in the new version works fine :) thanks!

[HoP]

@JanWillem32
another newbie question(i have many question :D )
what is the difference between Video Renderers?
in this thread (http://forum.doom9.org/showthread.php?t=159614) nevcairiel said:

The Video Renderer is responsible for showing the video frames on your screen, in the proper order and proper timing

now i want to know what makes difference between them?
for example why madVR is a good video renderer?

[vivan]

features (that affect quality) and bugs, stability, resource consumption.
Why madVR is good you can read in madVR thread (http://forum.doom9.org/showthread.php?t=146228).

JanWillem32,
On Intel HD 3000 gpu, when I choose >8 bit in presentation I get no video... With nVidia 540M everything is ok. Is this hardware limitation?

[Hera]

x64 is pdb

[JanWillem32]

Thanks, Hera. I changed the link.

@HoP: Various renderers have various ages. Some renderers are just old, such as Overlay, Old renderer and VMR-7. VMR-9 is reasonably new (it's 11 years old). It supports a DirectX 9 level as renderer to use the video card to transform images and paint it on screen. EVR adds a bit of functionality on top of that, mostly to accommodate the MediaFoundation and new DXVA capabilities of recent hardware.
Haali, MadVR, VMR-9 r. and EVR CP are all renderers based on DirectX 9, but have extensions in the programming to support additional functionality for video playback. The basic added functionality is support for OSDs, stats screens, subtitles and custom resizers. The more advanced features for MadVR can be read in the madVR thread. For the guide on VMR-9 r. and EVR CP, I still need someone to gather, compile and edit the text. I've already written a few parts for usage on the Russian site: http://arestarh.pp.ua/ . There isn't a recent guide in English available. (I still have the original text for the items I already wrote in English, of course.)
The new renderer features in these builds are alpha status code. I will sometimes have to change the name of a certain function later on. Descriptions of these functions are still scattered throughout this thread.

@vivan: The EVR mixer can black out without a warning for various reasons with more-than-8-bit outputs. Try disabling the additional video playback filter options in the Intel video control panel. You can also try VMR-9 r., which has worse timing and lacks some functions, but is generally more lenient with its output surfaces option. CruNcher has a very recent Intel CPU with IGP, but I think he also never got that one to work with more-than-8-bit outputs on EVR either.

[GTPVHD]

Jan, can you see if you can start implementing D3D11.1 support?

http://msdn.microsoft.com/en-us/library/windows/desktop/hh404562

"Process video resources with shaders" section seems interesting, but I don't know if it's feasible or not.

[JanWillem32]

I'm indeed looking forward to DirectX 11.1, but the DirectX 11 renderer I already started on isn't quite working (the EVR chokes on it, even with the DirectX 9 compatibility layer). DirectX 11.1 adds many video playback-specific upgrades, which I'd love to try in a while. (I should first get the current renderer in proper working order.)

[vivan]

Try disabling the additional video playback filter options in the Intel video control panelThey are all disabled.

You can also try VMR-9 r.Yes, with VMR-9 renderless it works... Unless I choose shader "YV12 Chroma Upsampling" (since Intel uses NN for chroma upsampling, like ati) =/ It works only with 8-bit surfaces.

[Hera]

I tried DX10 renderer for Unreal Tournament GOTYE and my CPU told me to GTFO. Is that what you mean by "EVR chocked on it"?

[shadewither]

Thanks for the reply.


@shadewither: 1. It's (-0.5, -0.5). DirectX 9 sets vertices on pixel boundaries, not centres. http://msdn.microsoft.com/en-us/library/windows/desktop/bb219690%28v=vs.85%29.aspx
There are various states you can pin to a sampled texture stage. For example, you can even set this odd mix of settings:m_pD3DDev->SetSamplerState(2, D3DSAMP_ADDRESSU, D3DTADDRESS_CLAMP);
m_pD3DDev->SetSamplerState(2, D3DSAMP_ADDRESSV, D3DTADDRESS_WRAP);
m_pD3DDev->SetSamplerState(2, D3DSAMP_ADDRESSW, D3DTADDRESS_MIRROR); Note that for mirrorring, the mirroring happens on odd-plane coordinates. From the standard viewport, with mirroring states on U and V: the samles at the interval [(0, 0), (1, 1)] will be oriented the same as [(2, 2), (3, 3)]. [(1, 1), (2, 2)] will be mirrorred horizontally and vertically. Clamp, wrap and mirror once are a bit more straight forward. http://msdn.microsoft.com/en-us/library/bb174456%28v=vs.85%29.aspx

I meant for a theoretical pure image resizer (e.g. in plain C, so 1=1 pixel/texel; x,y=0,1,...,100, not uv coord).
I thought that pixel=middle of pixel, so after 100x100=>200x200 resize, NEW pixel(0,0)=NEW position(0,0), should map to a OLD position(-0.25,-0.25) ?


2. For large factor down-scaling, the typical usage is a (large) gaussian kernel. It just averages a lot of pixels and then outputs one blurred pixel.
For exact down-sizing with a factor of 2 or 3, just average the source pixels.
In a 3D renderer mipmaps are often used to scale images down, while keeping the same texture in memory. This way it's easy to re-use, even if the orientation and visible size of the object changes. http://en.wikipedia.org/wiki/Mipmap

So Mipmaps are typically generated by averaging 4 pixels?

[JanWillem32]

@vivan: good to know that Intel chips also need the extra chroma up-sampling stuff. I'll add it to the list of auto-detection for this item, and change the name of the chroma up-sampling fix.

@Hera: EVR just refused to output on the DirectX 11 surfaces I gave it. It never started to output at all. Later on, I found out that DirectX 11.1 support was added to the EVR.dll/DXVA2.dll version in Windows 8. I don't know about the status of upgrading the DirectX base in Vista and 7 to 11.1 yet. They upgraded Vista to DirectX 11, let's hope 11.1 is pushed in the same way in the near future.

@shadewither: Then you are correct. The pixel center is .25 pixel size top left of the old pixel center for the top left pixel when doing a plain 2× resize. Resizing is done using clamping on the borders, by the way. Mirrored sampling gives ringing artifacts.
Mipmaps are half-size-rounded-up for every step into the mipmap, for every dimension. That way, the lowest mipmap is always a single pixel. Sampling a mipmap is easy. It's presented in the sampling intrinsic as an extra dimension. For 2D textures you sample 8 pixels: 2 (width) × 2 (height) × 2 (mipmap). For 3D textures it's 16 pixels for the depth factor. The output sample is typically linearly interpolated (I'll skip anisotropy for now).

[janos666]

I guess older Windows versions will get the new DX with their next service packs. Win7 will likely have it but I am not sure about Vista.

I use the Windows 8 RP right now and I would gladly test an mpchc build with a DX11 renderer. :)

By the way... Any news about a real YCC 4:2:0 HDMI output?

[ryrynz]

Win7 will likely have it but I am not sure about Vista.

7 is guaranteed. Vista won't get it.

[nevcairiel]

By the way... Any news about a real YCC 4:2:0 HDMI output?

HDMI doesn't support 4:2:0, you need to upsample to 4:2:2 at least.

[burfadel]

Agreed, Vista won't get Directx 11.1 but Windows 7 should.

[gilic]

This is how my stats screen looks like with the new version. I don't know if 6Hz is supposed to be correct.
pic (http://postimage.org/image/nn7154snr/full/)

[janos666]

HDMI doesn't support 4:2:0, you need to upsample to 4:2:2 at least.

Ah, yes, sorry. So, any news about a real YCC 4:2:2 output?:o

[Hera]

Latest build seems to work well in as so far.

[gilic]

I've found some sort of regression with your build JanWillem32. The subtitles of this really low res video aren't displayed, normal nightlies and xy-filter both work. Other subtitled videos with a more sane resolution also work correct. The video in question is hyouko 11.5 from this page here (http://mazuisubs.com/212). Settings are subpictures 6, screen resolution and both checkboxes ticked.

edit: Not sure if this is caused by lav splitter. I backtracked to your oldest revision @ mediafire, which has the same issue in combination with lav splitter, the internal splitter works.
edit2: tested so far:

dfr5403li with lav 0.51 -> doesn't work
dfr4739 with lav 0.51 -> doesn't work
dfr4739 with internal splitter -> works
1.6.3.5417.x86.Lite nightly with lav 0.51 -> works

[Gelatinous]

I've found some sort of regression with your build JanWillem32. The subtitles of this really low res video aren't displayed, normal nightlies and xy-filter both work. Other subtitled videos with a more sane resolution also work correct. The video in question is hyouko 11.5 from this page here (http://mazuisubs.com/212). Settings are subpictures 6, screen resolution and both checkboxes ticked.

edit: Not sure if this is caused by lav splitter. I backtracked to your oldest revision @ mediafire, which has the same issue in combination with lav splitter, the internal splitter works.
edit2: tested so far:

dfr5403li with lav 0.51 -> doesn't work
dfr4739 with lav 0.51 -> doesn't work
dfr4739 with internal splitter -> works
1.6.3.5417.x86.Lite nightly with lav 0.51 -> works

I had the exact same issue with this file. I'm using AV Splitter instead of LAV though.

[sexus]

It's currently being added to the trunk by Bobdynlan, give it a few weeks or so and it'll be there for everyone.

has it already been added?

p.s: this D3D11.1 support should be interesting with madvr dont tell me even more epic video quality perhaps? OMG!!!!!!!

[ryrynz]

It hasn't. I figured a slight overhaul of the existing code wouldn't take very long, looks like he's pacing himself.

[sexus]

darn was hoping i could enjoy myself some gui goodness again , well i sure hope it gets added soon

[Keiyakusha]

p.s: this D3D11.1 support should be interesting with madvr dont tell me even more epic video quality perhaps? OMG!!!!!!!

Video have the same quality regardless. The only things that somewhat affect it is correct dithering, good upsample algorythm (or better so say wide choice of them for pleasure of different people), correct colorimetry handling... madvr does everything in "correct" way, it delivers maximum quality but it no way improves quality in a way you seems to be thinking of. In other words you won't see more quality untill someone will invent a way to produce details from air and apply to the video.