madVR - high quality video renderer (GPU assisted)

[Sunset1982]

@madshi: I noticed that on my nVidia 980 Ti, 1080p frame packed 3D 23Hz translates to really ~23.000fps (so about every 3-4minutes there's a glitch). If I remember correctly, it was mentioned that madVR forces 3D 1080p at 23Hz. Is there a way to change that to 24Hz instead? Both are supported by nVidia at 1080p frame packed 3D, and 24Hz will be a lot closer to 23.976 than 23.

AFAIK, there doesn't seem to be a way to override/create a custom 3D resolution the same way it exists for non-3D resolutions.

That's exactly the same problem I have with my gtx 970. I hope madshi gives us an option to change between 23 and 24 hz in 3d modes.

[Thunderbolt8]

the most sensible thing to do would be to not scale the chromawhy do you recommend not to (down)scale the chroma?

[huhn]

because it has the "correct" resolution already.

[Thunderbolt8]

because it has the "correct" resolution already.I dont understand. when the videos resolution is 4k, but your screen 1080p and the video gets scaled down to this, then the chroma already is 1080p? so why is there then the option in madvr to scale the chroma at all?

edit: actually its chroma upscaling...well Im lost.

[huhn]

read this: https://en.wikipedia.org/wiki/Chroma_subsampling

4:2:0 is used on DVD, BD and UBD. so the chroma resolution is 1/4 the luma resolution. 2160p = 1080p chroma.

[foozoor]

Hello madshi and thanks again for your amazing work.

Could you put back the old version of finesharp?

Crispen edges is sometimes too much for my gpu.
I also still prefer finesharp result than crispen edges one alone.
However I didn't test to combine multiple sharpen shaders at once as my gpu is too limited.

Could you make the display modes generic for any display monitor like smooth motion?

In Kodi there is a option called "Adjust Display Refresh Rate" and this is an on/off button.
I think display modes is a very important thing and it should be simpler for novice people.
It would be great if madVR could automatically get available display frequencies and switch to the appropriated one.

[huhn]

Could you make the display modes generic for any display monitor like smooth motion?

In Kodi there is a option called "Adjust Display Refresh Rate" and this is an on/off button.
I think display modes is a very important thing and it should be simpler for novice people.
It would be great if madVR could automatically get available display frequencies and switch to the appropriated one.

a general just change option is pretty dangerous.

some(if not most) TVs have a broken 23/24p mode some can't deal with 50 hz well.

but he already said he wants to remade this part.

for the moment just add this:

1080p23, 1080p24, 1080p50, 1080p59, 1080p60

it doesn't matter if your screen doesn't support all of them.

[foozoor]

some(if not most) TVs have a broken 23/24p mode some can't deal with 50 hz well.


How to know if it's broken or not?
And what to do if it is? Isn't smooth motion suitable for this special case?

[huhn]

How to know if it's broken or not?
And what to do if it is? Isn't smooth motion suitable for this special case?

yes smoothmotion is suitable in this case but you need custom display modes because 50p may work totally fine.

and if you want to fine out if 23p is broken on your screen either read tests or just look at it.

most TVs with a broken 23p mode do a 23p -> 59p conversation adding a 3:2 judder.

[Thunderbolt8]

read this: https://en.wikipedia.org/wiki/Chroma_subsampling

4:2:0 is used on DVD, BD and UBD. so the chroma resolution is 1/4 the luma resolution. 2160p = 1080p chroma.hm so 4k downscaled to 1080p could look better than native 1080p, just because the chroma needs to be upscaled with native 1080p while it doesnt with 4k --> 1080p?

[huhn]

hm so 4k downscaled to 1080p could look better than native 1080p, just because the chroma needs to be upscaled with native 1080p while it doesnt with 4k --> 1080p?

in theory, yes.

but the is still the chroma subsample position problem.

[chros]

read this: https://en.wikipedia.org/wiki/Chroma_subsampling

4:2:0 is used on DVD, BD and UBD. so the chroma resolution is 1/4 the luma resolution. 2160p = 1080p chroma.
There's an exception: when crop black borders is active. Sometime it crops 1-2 lines and that triggers chroma upscaling even when the 'don't upscale when' option is set. So either it's a bug or ... :) It would be nice if madvr doesn't scale chroma at all in these kind of situations but I don't know whether this can be achieved.

[huhn]

There's an exception: when crop black borders is active. Sometime it crops 1-2 lines and that triggers chroma upscaling even when the 'don't upscale when' option is set. So either it's a bug or ... :) It would be nice if madvr doesn't scale chroma at all in these kind of situations but I don't know whether this can be achieved.

of cause it does. if you crop a picture you crop the chroma channel too.

if you crop by 1 pixel you need to first upscale the chroma.

it is still cropping the image but doesn't scale it anymore.

[Ver Greeneyes]

A question about the upcoming downscaler: does it support any (smaller) scale factor, or is it limited to 0.5x scale like the image doubling options? I often watch 1080p videos in ~720p so I can still multitask a bit, but I don't often go all the way down to 50%.

I also agree with the others that it would be nice if downscaling to 50% or below disabled chroma upscaling entirely and simply downscaled the luma channel to 50% instead - seems like that would reduce the overhead by a bunch for large videos. Even for scale factors slightly over 50% (but close to 50%), it might not be a bad trade quality option to downscale luma instead of upscaling chroma.

[huhn]

I also agree with the others that it would be nice if downscaling to 50% or below disabled chroma upscaling entirely and simply downscaled the luma channel to 50% instead - seems like that would reduce the overhead by a bunch for large videos.

if an image is scaled from 2160 to 1080p only luma is scaled.

[Ver Greeneyes]

if an image is scaled from 2160 to 1080p only luma is scaled.
Hmm, but not for smaller resolutions? I don't have any 2160p video, so I just tested scaling 1080p to 540p and the OSD seems to indicate it still scales chroma (then downscales).

[huhn]

i just scaled a 1080p video to 540p and the chroma scaling disappears. if it isn't perfectly scale down by 0.5 it is still scaling chroma.

[Ver Greeneyes]

i just scaled a 1080p video to 540p and the chroma scaling disappears. if it isn't perfectly scale down by 0.5 it is still scaling chroma.
That's odd. I explicitly selected 50% in the latest nightly of mpc-hc on a 1920x1080 video, and checked that the client rect size becomes 960x540, but the madVR OSD still says "chroma > super-xbr" and "image < Catmull-Rom AR". It also stays the same if I further shrink the window to 959x539. I'm using the latest madVR (v0.90.3) in d3d11 new windowed mode, 64-bit, tested on a 4:2:0 video, black bar detection disabled, no trade quality options checked, smooth motion enabled.

[DD51]

I read the most I can to get 3D to work with MadVR.
No.1 thing I needed to do was go from Windows 7 to Windows 8.1 - DONE
After doing a fresh install of Win8.1 I installed:
MPC-HC 32 bit
MadVR 90.3
FFdshow...I used this for sharpening the image

Now I rip the 3D BRD to a MKV file using MakeMKV.
Launch the file and cannot get 3D to work - at all.

Videocard is HD7770
Crimson drivers 16.1 installed

3D works fine if using PowerDVD.

Really don't know what I'm doing wrong. :o

Many thanks in advance

[Della]

a general just change option is pretty dangerous.

some(if not most) TVs have a broken 23/24p mode some can't deal with 50 hz well.

but he already said he wants to remade this part.

for the moment just add this:

1080p23, 1080p24, 1080p50, 1080p59, 1080p60

it doesn't matter if your screen doesn't support all of them.

Pardon the novice question please.

But wouldn't choosing 25,29 and 30 where applicable (vice 50, 59 and 60) allow for more processing time between frames. Thereby allowing more aggressive MadVR settings?

Thanks

[baii]

Pardon the novice question please.

But wouldn't choosing 25,29 and 30 where applicable (vice 50, 59 and 60) allow for more processing time between frames. Thereby allowing more aggressive MadVR settings?

Thanks
Assuming smooth motion is off, the same frame would just get rendered once and shown twice.


Sent from my 306SH

[Della]

Assuming smooth motion is off, the same frame would just get rendered once and shown twice.


Sent from my 306SH

Ah, thanks.

[huhn]

I read the most I can to get 3D to work with MadVR.
No.1 thing I needed to do was go from Windows 7 to Windows 8.1 - DONE
After doing a fresh install of Win8.1 I installed:
MPC-HC 32 bit
MadVR 90.3
FFdshow...I used this for sharpening the image

Now I rip the 3D BRD to a MKV file using MakeMKV.
Launch the file and cannot get 3D to work - at all.

Videocard is HD7770
Crimson drivers 16.1 installed

3D works fine if using PowerDVD.

Really don't know what I'm doing wrong. :o

Many thanks in advance

remove ffdshow and make sure a nightly build from lavfilter is used not the internal lavfilter version from mpc-hc.

Pardon the novice question please.

But wouldn't choosing 25,29 and 30 where applicable (vice 50, 59 and 60) allow for more processing time between frames. Thereby allowing more aggressive MadVR settings?

Thanks

no it doesn't you still have the same time to render the frames.

presenting a frame every vsync with d3d11 is not totally free but can be "ignored".

[huhn]

That's odd. I explicitly selected 50% in the latest nightly of mpc-hc on a 1920x1080 video, and checked that the client rect size becomes 960x540, but the madVR OSD still says "chroma > super-xbr" and "image < Catmull-Rom AR". It also stays the same if I further shrink the window to 959x539. I'm using the latest madVR (v0.90.3) in d3d11 new windowed mode, 64-bit, tested on a 4:2:0 video, black bar detection disabled, no trade quality options checked, smooth motion enabled.

this looks fine or not? http://abload.de/img/50zoomtest8qkju.png

[omarank]

A question about the upcoming downscaler: does it support any (smaller) scale factor, or is it limited to 0.5x scale like the image doubling options? I often watch 1080p videos in ~720p so I can still multitask a bit, but I don't often go all the way down to 50%.

I also agree with the others that it would be nice if downscaling to 50% or below disabled chroma upscaling entirely and simply downscaled the luma channel to 50% instead - seems like that would reduce the overhead by a bunch for large videos. Even for scale factors slightly over 50% (but close to 50%), it might not be a bad trade quality option to downscale luma instead of upscaling chroma.

That's odd. I explicitly selected 50% in the latest nightly of mpc-hc on a 1920x1080 video, and checked that the client rect size becomes 960x540, but the madVR OSD still says "chroma > super-xbr" and "image < Catmull-Rom AR". It also stays the same if I further shrink the window to 959x539. I'm using the latest madVR (v0.90.3) in d3d11 new windowed mode, 64-bit, tested on a 4:2:0 video, black bar detection disabled, no trade quality options checked, smooth motion enabled.
If you use Linear Light downscaling, chroma will always be upscaled. Untick Linear Light, then madVR will not upscale chroma when video is downscaled by a large scaling factor, like 50%.

[Ver Greeneyes]

this looks fine or not? http://abload.de/img/50zoomtest8qkju.png
That looks fine, but it's not what I see :) There must be a bug somewhere.

Here's what I get, playing back a 1080p video at 50% scale: http://i.imgur.com/YcCRswk.png

[nevcairiel]

That looks fine, but it's not what I see :) There must be a bug somewhere.

Here's what I get, playing back a 1080p video at 50% scale: http://i.imgur.com/YcCRswk.png

madVR needs to upscale chroma when you do linear light downscaling, thats just a requirement for linear light to function (linear light requires RGB, which requires chroma to be at luma resolution first). If you turn that off, you avoid the chroma upscaling in this combination.

[Ver Greeneyes]

madVR needs to upscale chroma when you do linear light downscaling, thats just a requirement for linear light to function (linear light requires RGB, which requires chroma to be at luma resolution first). If you turn that off, you avoid the chroma upscaling in this combination.
Aha, I see! Yes, now it works. Thanks for clearing that up!

Though, I do have a question: couldn't you technically apply the same gamma curve to the luma channel to make it linear light? I mean, it is the brightness channel. Perhaps a trade quality option: Downscale luma channel at 50% scale or below even when linear light is enabled.

[nevcairiel]

Though, I do have a question: couldn't you technically apply the same gamma curve to the luma channel to make it linear light?

Gamma is a RGB function, can't apply it to Luma alone I'm afraid.

[Ver Greeneyes]

Gamma is a RGB function, can't apply it to Luma alone I'm afraid.
I don't really understand why you couldn't, though it wouldn't necessarily be correct - I mean, it's just a 1D function applied separately to 3 channels. I suppose the higher the saturation of a particular pixel, the less accurate the conversion becomes, since 1 or 2 of the color channels would be dark and the others bright. But doesn't linear light mostly help for darker shades anyway? Seems to me you'd still be closer to the right ballpark than with gamma light.

[huhn]

you can even convert the luma channel to RGB and back lossless but not sure about the result when scaled this way.

[Asmodian]

I don't really understand why you couldn't, though it wouldn't necessarily be correct - I mean, it's just a 1D function applied separately to 3 channels. I suppose the higher the saturation of a particular pixel, the less accurate the conversion becomes, since 1 or 2 of the color channels would be dark and the others bright. But doesn't linear light mostly help for darker shades anyway? Seems to me you'd still be closer to the right ballpark than with gamma light.

Doing it wrong to same some performance is not the madVR way. :D

It think that would be wrong enough that it would be better to simply downscale in gamma light.

[Ver Greeneyes]

Doing it wrong to same some performance is not the madVR way. :DTrue, but it's a "trade quality for performance" option I'd happily enable ;) I'm not so happy about having to disable Linear Light scaling to bypass chroma upscaling right now though (I guess I could avoid it with profiles, but still) - at the very least it seems unintuitive.

It think that would be wrong enough that it would be better to simply downscale in gamma light.
I don't know about that. I mean, linear light downscaling mostly helps against ringing on sharp transitions in brightness, right? So I feel like there would still be some benefit. We don't use it for upscaling, so I don't think it's the most scientific approach in the world anyway. Still, maybe it's already been tried and people found that it doesn't look good. And who knows, maybe linear light doesn't even work that well with Shiandow's new algorithm. I'm sure madshi will have an opinion on this either way - I'll happily defer to his judgment (though I do think the current behavior is confusing).

[Shiandow]

I don't know about that. I mean, linear light downscaling mostly helps against ringing on sharp transitions in brightness, right? So I feel like there would still be some benefit. We don't use it for upscaling, so I don't think it's the most scientific approach in the world anyway. Still, maybe it's already been tried and people found that it doesn't look good. And who knows, maybe linear light doesn't even work that well with Shiandow's new algorithm. I'm sure madshi will have an opinion on this either way - I'll happily defer to his judgment (though I do think the current behavior is confusing).

Linear light also helps when downscaling saturated colours (for more info see this (https://www.youtube.com/watch?v=LKnqECcg6Gw) nice explanation). That said I'm not 100% sure if it would be that bad to only use gamma correction for the luma channel, because you only have the downscaled chroma to begin with, so if you can find the correct luma then it should work fine.

Only one way to find out though. There's only so much you can justify theoretically, and especially with gamma light vs linear light there is not always an obvious 'correct' choice.

Oh, and I think linear light would work well with the new downscaling algorithm, but I haven't had a chance to try it yet.

[Portioli]

hi guys..
i need a little help. I used finesharp with the old way i.e. avisynth

this is the script i had:

# Small and relatively fast realtime-sharpening function,
# for 1080p, or after scaling 720p -> 1080p during playback (to make 720p look more like being 1080p)
# It's a generic sharpener. Only for good quality sources!
# (If the source is crap, FineSharp will happily sharpen the crap.) ;)
# Noise/grain will be enhanced, too. The method is GENERIC.
#
# Modus operandi: A basic nonlinear sharpening method is performed, then the *blurred* sharp-difference gets subtracted again.
#
# - Didee




function FineSharp(clip c, int "mode", float "sstr", float "cstr", float "xstr", float "lstr", float "pstr", float "ldmp")
{

mode = default(mode, 1 ) # 1 to 3, weakest to strongest. When negative -1 to -3, a broader kernel for equalisation is used.
sstr = default(sstr, 2.0 ) # strength of sharpening, 0.0 up to ??

_cstr = spline(sstr, 0,0, 0.5,0.1, 1.0,0.6, 2.0,0.9, 2.5,1.00, 3.0,1.09, 3.5,1.15, 4.0,1.19, 8.0,1.249, 255.0,1.5)
_cstr = (mode>0) ? _cstr : pow(_cstr,1./1.25)




cstr = default(cstr, _cstr) # strength of equalisation, 0.0 to ? 2.0 ? (recomm. 0.5 to 1.25, default AUTO)
xstr = default(xstr, 0.19 ) # strength of XSharpen-style final sharpening, 0.0 to 1.0 (but, better don't go beyond 0.249 ...)




# Viscera parameters
lstr = default(lstr, 1.49 ) # modifier for non-linear sharpening
pstr = default(pstr, 1.272) # exponent for non-linear sharpening
ldmp = default(ldmp, sstr+0.1) # "low damp", to not overenhance very small differences (noise coming out of flat areas)




str1 = sstr
str2 = cstr
SSTR = string(sstr)
CSTR = string(cstr)
LSTR = string(lstr)
PSTR = string(pstr)
LDMP = string(ldmp)
rg=mode>0?11:20

b = (abs(mode)==1) ? c.removegrain(11,-1).removegrain(4,-1)
\ : (abs(mode)==2) ? c.removegrain(4,-1).removegrain(11,-1)
\ : (abs(mode)==3) ? c.removegrain(4,-1).removegrain(11,-1).removegrain(4,-1) : c

shrpD = mt_lutxy(c,b,"x y - abs "+LSTR+" / 1 "+PSTR+" / ^ "+SSTR+" * x y - x y - abs 0.001 + / * x y - 2 ^ x y - 2 ^ "+LDMP+" + / * 128 +")

shrp = (str1<0.01) ? c : c.mt_adddiff(shrpD,U=2,V=2)




shrp = (str2<0.01) ? shrp : shrp.mt_makediff(shrpD.mt_lut("x 128 - "+CSTR+" * 128 +").removegrain(rg,-1),U=2,V=2)




shrp = (xstr<0.01) ? shrp
\ : mt_lutxy(shrp,shrp.removegrain(20,-1),"x x y - 9.9 * +",U=2,V=2).repair(shrp,12,0).mergeluma(shrp,1.0-xstr)




return(shrp)
}


Which settings are now closest to my script?
Is 750 Ti capable of using Madshi`s algorithm for sharpening?

Thanks in Advance!

[macycat]

It looks like I found a problem created between madVR v0.89.12 and v0.89.13. It took some time to track this down since I hadn't upgraded in quite a long time.

The problem occurs with 1080i59 content (recorded with HDPVR), but does not seem to be a problem with 1080p23 content. While playing a 1080i59 video file, if I double click on another 1080i59 video file to play it in the same video player window, I get a blank screen with sound from the new video. I also previously was able to drag the video player window from one screen to another screen with no problems (just a minor pause in playback), but that now usually gives a blank screen with audio playing with an occasional DirectX fails to render crash.

MPC-HC 1.7.10 (64-bit)
LAV Filters 0.67.0
madVR 0.89.12 - ok
madVR 0.89.13 - fail

Windows 7 64-bit
NVIDIA GTX 670 video card

madVR settings:
D3D9 overlay
NV12, 8 bit, 4:2:0
movie 29.970 fps (says source filter)
movie 1920x1080, 16:9
touch window from outside
chroma > Jinc AR
image < Lanczos3 AR
vsync 16.68ms, frame 16.68ms
matrix BT.709 (says upstream)
primaries BT.709 (says upstream)
limited range (says upstream)
deinterlacing on (settings)
...

I haven't updated for quite a while, but with all the great work being done, I thought I would give the 3D playback a try. That is when I noticed the problem described above.

Thanks for all the great features and hard work.

[iSeries]

Hi,

Couple quick questions for the experts, if your gpu could either run nnedi3 32 neurons + sharpen edges, nnedi3 32 neurons + superres strength 1, or super-xbr + superres strength 3, which would you choose?

Also, when using doubling, say 720p to 1440p down to 1080p, how is chroma scaled? Is it scaled just to the final output size of 1080p or is it also upscaled to 1440p with luma before downscaling to 1080p?

[Warner306]

Hi,

Couple quick questions for the experts, if your gpu could either run nnedi3 32 neurons + sharpen edges, nnedi3 32 neurons + superres strength 1, or super-xbr + superres strength 3, which would you choose?

Also, when using doubling, say 720p to 1440p down to 1080p, how is chroma scaled? Is it scaled just to the final output size of 1080p or is it also upscaled to 1440p with luma before downscaling to 1080p?

If you are upscaling 720p -> 1080p, I would use NNEDI3 32 neurons + SuperRes(1). For larger scaling factors, I would use super-xbr100 + SuperRes(3). But that is just an opinion.

The chroma is upscaled to the native luma resolution (1080p) before image doubling. If using NNEDI3, the luma alone is upscaled to 1440p before downscaling. If using super-xbr, the luma and chroma are doubled together to 1440p.

[adhara]

Hi,

What do you think about these settings for 1080p --> 1080p content ?
I would love to get the best settings for my 65" TV (use GTX970 Nvidia card with 4Gb memory).

Chroma: NNEDI3 128 neurons
Image: Jinc3 + AR
Image doubling: Off
Upscaling refinement: Off
Artifact removal - Debanding: Medium/High
Image enhancements: crispen edges (0.6)
Dithering: Error Diffusion 2

[Warner306]

Hi,

What do you think about these settings for 1080p --> 1080p content ?
I would love to get the best settings for my 65" TV (use GTX970 Nvidia card with 4Gb memory).

Chroma: NNEDI3 128 neurons
Image: Jinc3 + AR
Image doubling: Off
Upscaling refinement: Off
Artifact removal - Debanding: Medium/High
Image enhancements: crispen edges (0.6)
Dithering: Error Diffusion 2

Those are my settings. It is hard to judge whether those are the "right" settings. I think it mostly comes down to your preference for the use of added sharpening. Some prefer higher sharpening values, combining an edge sharpener with enhance detail, or using sharpen edges + thin edges + enhance detail.

It comes down to your eyes. But I like the set-up you have.

[DD51]

remove ffdshow and make sure a nightly build from lavfilter is used not the internal lavfilter version from mpc-hc.

Thanks huhn...
The lav "nightly" build did the trick...can't believe it was that simple :mad:

Going to do some tests to see which version I prefer between PDVD and via MADVR and MPC-HC.

My test is watching on a 9ft wide screen and a JVC projector

[har3inger]

Hi,

Couple quick questions for the experts, if your gpu could either run nnedi3 32 neurons + sharpen edges, nnedi3 32 neurons + superres strength 1, or super-xbr + superres strength 3, which would you choose?

Also, when using doubling, say 720p to 1440p down to 1080p, how is chroma scaled? Is it scaled just to the final output size of 1080p or is it also upscaled to 1440p with luma before downscaling to 1080p?

I recommend sxbr 100 and sr3 for 720p going up. For 480p to 1080p I recommend nnedi32 with no superres. Superres is more important than nnedi32 vs sxbr on high quality clean sources. However, on dirty and low resolution sources, sr just increases source artifacts and introduces aliasing, and I generally dislike it. It seems to like a low image detail:pixel resolution density to avoid aliasing, and small resolution video have too high a density.

Madshi, is it possible to implement profile selection that detects power settings, i.e., plugged or unplugged laptop?

[Warner306]

I recommend sxbr 100 and sr3 for 720p going up. For 480p to 1080p I recommend nnedi32 with no superres. Superres is more important than nnedi32 vs sxbr on high quality clean sources. However, on dirty and low resolution sources, sr just increases source artifacts and introduces aliasing, and I generally dislike it. It seems to like a low image detail:pixel resolution density to avoid aliasing, and small resolution video have too high a density.

Madshi, is it possible to implement profile selection that detects power settings, i.e., plugged or unplugged laptop?

I agree SuperRes seems to matter more than the choice between super-xbr and NNEDI3. But I disagree that 480p content is too low quality to benefit from SuperRes.

I get the most impressive results of any content using the chain of super-xbr100 + AR + SuperRes(3) + crispen edges(1.0) when applied to fairly low bitrate streaming video. This is a resize of 540p -> 1080p -- the result of which looks like HD video, to me. As long as the source is clean, its resolution shouldn't matter.

I think sharpening, in general, looks most natural when combined with large amounts of upscaling. This allows for more aggressive values of each sharpener and more control over the end result.

[har3inger]

It does depend. I have a lot of scenes where I can identify where superres makes thin smooth lines (after nnedi3) into thin jagged lines. These happen a lot more often when the lines are 1-3 px, as they often are in 480p or lower. Otherwise, yes, superres helps a lot to sharpen the image at 480p.

[hu1kamania]

I've been having trouble getting a clip to play smoothly. It's a sample of the sort of content I see frequently while watching TV.

http://www.filedropper.com/sample2

The first half of the clip works perfectly with Film Mode forced on, but I get judder when the golf program resumes. If I force Video or allow MadVR to auto select the method, the reverse is true. Allowing LAV to deinterlace, and disabling MadVR's deinterlacing, I get smooth playback throughout. Though I lose the benefit of IVTC, something the first part of my clip needs.

I get the same behavior in EVR. Is this a limitation with DXVA2 hardware deinterlacing?

I had asked the same question on the LAV Filters thread, but with some guidance I decided to bring my question here as I believe there might be a problem with DXVA2 hardware deinterlacing.

[Light001]

Firstly id like to say "Yay" i finally made it haha :), Excellent reading so far, and the added security of random questions and 5 days before posting is nice. Now what i am here for is because my brain has melted and id love to have some peace of mind where it once was. I have been using MadVr and MPC-HC, I Attempted Ezoden's guide for Video Euphoria aswell as mike jung. Honestly Im not sure what to be looking for, Overall the quality seems to be nice either way With Whichever rendering setting i have used, i always aim to see more green then red. my issue however is with the jittering and stuttering if you could call it that, i could not figure out reclock. but basically i have a BENQ xl2420t monitor that i used tft centrals color calibration for in windows, i enjoy that, but am still driven to madness over how bright windows 10 is and its not in the cards for them to just give us a darn black theme. Anyway thats another story.
So with my monitor, it is 120hz, and these movies are 23.9 something , So im assuming that thats whats causing the ugly sluggish skittering/stuttering/jittering nonsense when the scene's are in motion. The quality is nice but this drives me bonkers...and i also find that 120fps with SVP is just way to fast and looks unnatural for film and ya...so can i find maybe a 60fps or anything slower that will not cause this affect? I have a good chunk of other questions and such, just looking for some help. I have $14 CAD saved in my pay pal to whomever can help me solve my biggest issue or give me the best "For DUmmies" version of why or what can or cannot be done. (possibly pictures,Click here,Do this,Try this,what looks like this with updated versions) type deal. Also i cannot seem to get reclock to uninstall :(. And my Specs are this. i5 3550 running at 4.1ghz, Gtx 980, 16 gig ram and windows 10 64bit. 2tb+ssd128. :stupid: Ty!

[adhara]

Those are my settings. It is hard to judge whether those are the "right" settings. I think it mostly comes down to your preference for the use of added sharpening. Some prefer higher sharpening values, combining an edge sharpener with enhance detail, or using sharpen edges + thin edges + enhance detail.

It comes down to your eyes. But I like the set-up you have.

Would it be possible to update your wiki page with the recommanded settings for high profile (970/980 Nvidia video card) when outputing in 1080p ? You did the same for 2160p output.
I do not see a big difference between NNEDI 128 neurons and Super-XBR in term of quality. Super-XBR looks better in term of sharpen and less GPU asking.

Regards.

http://forum.kodi.tv/showthread.php?tid=222576

[bcec]

It does depend. I have a lot of scenes where I can identify where superres makes thin smooth lines (after nnedi3) into thin jagged lines. These happen a lot more often when the lines are 1-3 px, as they often are in 480p or lower. Otherwise, yes, superres helps a lot to sharpen the image at 480p.

On a slighly related note, would sr3+nnedi3@64 (or sr3 + sxbr) be recommended over plain nnedi3@128 on a 1080p -> 2160p upscale?

[aufkrawall]

On a slighly related note, would sr3+nnedi3@64 (or sr3 + sxbr) be recommended over plain nnedi3@128 on a 1080p -> 2160p upscale?
Yes, anytime.

[baii]

On a slighly related note, would sr3+nnedi3@64 (or sr3 + sxbr) be recommended over plain nnedi3@128 on a 1080p -> 2160p upscale?
Even with the highest end cards, I think they may struggle at those setting. Super res is super expensive at 4k. Definitely impossible for 60fps content.

Sent from my 306SH