madVR - high quality video renderer (GPU assisted)

[leeperry]

Originally Posted by Sunset1982

I would like to know is, what settings the other guys here prefer when using NNEDI or superXBR. Are you using SR with both? Any other suggestions?

SR for both, reconstruction soft for chroma & sxbr75/jinc for luma + SSIM 2D 100%.

Originally Posted by Shiandow

Perhaps it's interesting to point out that Anti-Bloating simply seems to make the result closer to what you'd get with the latest version of SuperRes. Not entirely sure what differences there are with MadVR's SuperRes, but among other things this version uses Linear Light, and no anti-ringing of any kind.

Ouh, touché!

[Iznogûd]

Originally Posted by QBhd

madshi,

What changes happened between 0.90.20 and 0.90.22 that would hurt performance? As I have said in the past, I like pushing my GPU to the max and my 720p24 profile is right on the edge (so much so that even a chroma superres of just 1 pushes it over the edge). When I tried 0.90.22 the first file I watched at 720p was actually 718 and I thought the performance hit was because of that... however later I watched one that was 720 and I could no longer maintain render queue. I rolled back to 0.90.20 and all is well again.


QB

I'm experiencing the same issues as well, running a GTX750ti against a 1440x900 display.

[jerryleungwh]

I'm trying to use the debug mode to find out what's preventing me from using D3D11 exclusive mode but I don't quite understand what the log says. These two lines seem significant though

00006286.019 Window   blocking WM_NCACTIVATE
00006289.356 Render   prevented Direct3D from setting display mode 0x0, 0 Hz, progressive

Could this be the problem and how can I fix it? If no what should I look for in the log for such problem?

[whitestar999]

Originally Posted by leeperry

Anyway, it's not like anyone is forcing you to run AB on SR and I kinda rest my case that the top left corner with sxbr75+SR1 looks better with AB25 than without: http://screenshotcomparison.com/comparison/178293

This is what AB does, it cleans edges for smoother movements. Either way there is no free lunch to be expected from upscaling, it's all about compromises and I find the AB75% compromise more than worth it as picture doesn't look harsh and oversharpened anymore.

I have to admit I didn't saw any change in top-left corner but I did saw a black spot diminishing(on mouse over with AB25) & then enlarging on mouse out on bottom right corner of small egg like structure below the antenna/rod at the top of building structure.If only ppl make their voting choices in general elections as diligently.

[burfadel]

The adaptive sharpen shader has been updated, thought I'd mention it since it's used under the upscaling refinement and image enhancements options.

http://forum.doom9.org/showthread.php?t=172131

Changelog:
2016-07-10 - Replaced lerp used for under/overshoot compression with a weighted power mean, reduces some ringing.
- Corrected minor error in the custom smoothstep function.
- Slightly improved weighting for nmax & nmin.
- Tweaked some parameters.

[Sunset1982]

is there any difference in motion clarity when using NNEDI instead of SuperXBR?

[aufkrawall]

That's impossible, there can be no difference between scaling one frame seperated or many in a row with the same algorithm.
If you want best motion clarity, let madVR feed a display that supports black frame insertion with the correct refreshrate without smooth motion.

[ryrynz]

Originally Posted by Shiandow

Not entirely sure what differences there are with MadVR's SuperRes, but among other things this version uses Linear Light, and no anti-ringing of any kind.

Looks good, a bit too much sharpness and aliasing but likely you were using a different luma upscaler. I never got the impression SuperRes really helped much at all and have never enabled it. Might actually look at it when this updated version makes it way into madVR.

[pirlouy]

Originally Posted by aufkrawall

That's impossible, there can be no difference between scaling one frame seperated or many in a row with the same algorithm.

Well it looks possible to me. I guess lesser quality images might allow a better motion clarity than well defined images.

[Anima123]

Originally Posted by Shiandow

Perhaps it's interesting to point out that Anti-Bloating simply seems to make the result closer to what you'd get with the latest version of SuperRes. Not entirely sure what differences there are with MadVR's SuperRes, but among other things this version uses Linear Light, and no anti-ringing of any kind.

Was this version of SuperRes already available to public?

[Shiandow]

Originally Posted by Anima123

Was this version of SuperRes already available to public?

It should be the same as the one here.

[SpoCk0nd0pe]

Hey Madshi,

I've got a rather special request:

Could you implement changing 3d formats, allowing a change from frame packed 3d to frame sequential 3d?

or

Could you implement support for frame sequential 3d input?

[digitech]

does anyone know if i use the nvidia edge enhancement option with "use nvidia setting" selected on control panel if madvr really apply this setting or just ignore it?

[huhn]

Originally Posted by digitech

does anyone know if i use the nvidia edge enhancement option with "use nvidia setting" selected on control panel if madvr really apply this setting or just ignore it?

should be ignored except with DXVA processing. it may work with deinterlancing or DXVA scaling.

[Ironclad]

Originally Posted by nevcairiel

Having to copy the image in high-bit-depth from the GPU to system memory and back after processing will take a significant part of the performance, so don't expect too much free magic.

Oh so more so than the GPU needing to process an additional 36 or more frames per second, it's the lag caused by storing in system memory that's slowing everything down. I'm probably oversimplifying it but 640x480 frames->1440x1080 frames should increase the traffic through system memory by like 5 both ways compared to the current method.

Even if the GPU is doing 2.5-6 times as much work, that's not comparable to the difference in speed between RAM and VRAM.
https://en.wikipedia.org/wiki/List_o...-access_memory

Oh of course that's why already high framerate videos are no issue! The only solution would be if the interpolation is done entirely by a GPU. I guess using DDR4 will help somewhat. I'll see what the SVP developers think but I'm guessing it would require completely rewriting the filter.

[ShiftyFella]

Looks like HDR processing in .22 is kinda broken or maybe it's just me? LG Chess and Sony Camp videos play absolutely fine with hdr processing turned on and picture looks like it should but this Samsung demo and files from this post look totally wrong. HDR processing option turned on produces color shifted image. Playing with any of the options doesn't help and only turning it off helps. Using .20 or older produced no such issues, I skipped .21 and updated straight to .22 from .20 tho


samsung hdr demo:
hdr processing turned on:


hdr processing off:



sony camp:
hdr on:

hdr off:

[pose]

madshi, i was always curious as of what settings and scaling algorithms you prefer yourself. Can you tell me please? Also thank you for the madVR! My chinese cartoons look amazing!

[zoyd]

Originally Posted by ShiftyFella

Looks like HDR processing in .22 is kinda broken or maybe it's just me?

I'm not seeing any color shifting of that image here. Note that the text and image overlay is encoded with peak white at 255 (10,000 nits).

350 nit display (madVR set to 400 nits HDR on, preserve hue on)

[mastrboy]

Is it possible for MadVR to have more controls than the seek bar in exclusive mode? (like next/previous chapter buttons for example)

[ryrynz]

Originally Posted by ShiftyFella

HDR processing option turned on produces color shifted image.
sony camp:
hdr on:

I've watched this HDR test video quite a lot at work on HDR & non HDR TV's, and that's exactly how it looks.