Dogway's Filters Packs

[guest]

VDub2 is the recommended one. Make sure you use the correct version (32-bit or 64-bit) depending on your system and Avisynth setup. If it fails, it should give you a meaningful error message as well. We cannot see what you see unless you tell us.

You must have replied to this one, before I edited my post...

Yes, I opened a few scripts, and even the ones that work for me threw up a error message pop up.

I used VDub264, as I'm using Windows 10 x64 & Avisynth 3.7.1.

I would have to take a screen shot of the error message (or whatever it is).

One thing I did notice, it seemed to reference mvtools2 in most msg's.

[kedautinh12]

Maybe your CPU Xeon error with x265 and x264

[guest]

Maybe your CPU Xeon error with x265 and x264

I doubt it would happen on all of them.

I tried a different build of x264 & x265...same thing.

But it only seems to be BM3D...Knlmeans & the new DG filters work.

I know that BM3DCPU won't work due to no AVX2, but BM3DCUDA should !!!

[StainlessS]

One thing I did notice, it seemed to reference mvtools2 in most msg's.
Very mystical, so you want everybody to just guess what the problem might be.
Post the error message. [just write down the text, no screenshot]

[guest]

Very mystical, so you want everybody to just guess what the problem might be.
Post the error message. [just write down the text, no screenshot]

OK StainlessS, see what you make of this :-

Avisynth open failure:
Cannot load file 'C:/RipBot264/Tools/Avisynth plugins/Scripts/Custom/%AVISYNTHPLUGINS%/mvtools/mvtools2.dll'.
Platform returned code 126:
The specified module could not be found.

(C:\RipBot264\Tools\Avisynth plugins\Scripts\Custom\TemporalDegrain.avs,line 2)


Avisynth open failure:
Cannot load file 'C:/RipBot264/Tools/Avisynth plugins/Scripts/Custom/%AVISYNTHPLUGINS%/mvtools/mvtools2.dll'.
Platform returned code 126:
The specified module could not be found.

(C:\RipBot264\Tools\Avisynth plugins\Scripts\Custom\SMDegrain E-H BM3D CPU.avs,line 2)

https://imgur.com/23BJTtn

Why are some like this /, and others like this \....

Now both of these scripts work (on some of my PC's)

So I don't know why they have errors.

It does suggest that maybe there's something wrong with mvtools2.dll, but what ??

[StainlessS]

Platform returned code 126:
You are missing some requirement/runtime.

Here is the AIO (all in one) CPP runtime setup:- https://github.com/abbodi1406/vcredist/releases
Updated 23 days ago [has all cpp runtimes, smaller, less crap than M$ collection of runtimes, both x86 and x64 - recomended].
Above, probably the problem.

EDIT: "\" and "/" just depends what issues the error message, one is DOS style slash, other is unix style slash.

[guest]

You are missing some requirement/runtime.

Here is the AIO (all in one) CPP runtime setup:- https://github.com/abbodi1406/vcredist/releases
Updated 23 days ago [has all cpp runtimes, smaller, less crap than M$ collection of runtimes, both x86 and x64 - recomended].
Above, probably the problem.

EDIT: "" and "/" just depends what issues the error message, one is DOS style slash, other is unix style slash.

Hi, I only updated this a couple of days ago...I make sure I keep up to date on most things.

Got from MS, tho.

I will give this a shot, and report back, probably tomorrow, now.

Cheers & thanks.

[StainlessS]

I think that Mvtools also requires either libfftw3f-3.dll or fftw3.dll for certain modes.
EDIT: https://github.com/pinterf/mvtools
External dependencies:

libfftw3f-3.dll (or renamed to FFT3W.DLL) from http://www.fftw.org/ or look at ICL builds at http://forum.doom9.org/showthread.php?t=173229 It is used only at specific dct parameter values

May require Microsoft Visual C++ Redistributables https://www.visualstudio.com/downloads/

[guest]

I think that Mvtools also requires either libfftw3f-3.dll or fftw3.dll for certain modes.
EDIT: https://github.com/pinterf/mvtools

Yes, I'm pretty sure I have all that installed, but of course I will double check.

Cheers

[kedautinh12]

OK StainlessS, see what you make of this :-



https://imgur.com/23BJTtn

Why are some like this /, and others like this \....

Now both of these scripts work (on some of my PC's)

So I don't know why they have errors.

It does suggest that maybe there's something wrong with mvtools2.dll, but what ??

From your log, i seen this errors from mvtool2 not bm3dcuda

[guest]

From your log, i seen this errors from mvtool2 not bm3dcuda

Exactly, but if there was really a problem with it, I'm sure it would affect other filters & scripts.

Anyway, StainlessS has suggested a few things as well, so I will see how I go next time.

Cheers

[guest]

I think that Mvtools also requires either libfftw3f-3.dll or fftw3.dll for certain modes.
EDIT: https://github.com/pinterf/mvtools

OK, installed that CPP runtime setup package, and got a fresh mvtools2, and VD2 throws up the exact same error....

[StainlessS]

Well if you are using Avisynth x64, you need [from Dependency Walker]

x64 Visual C ++ Redistributable for Visual Studio 2015 package.
MSVCP140.dll
VCRuntime140.dll
VCRuntime140_1.dll

In your system32 [installed with x64 VS 2015 redist, or AIO package]

Also need x64 versions
libfftw3f-3.dll (or renamed to FFT3W.DLL) [some plugs require one name, some the other]
in system32

Other than that, I know of no other dependencies.

[WolframRhodium]

I have several PC's with nVidia GPU's ranging from a GTX 660, 760, 770, 780ti, and a 980ti hybrid, I used to have an MSI 1080 Gamer X, but it developed errors, and is basically un usable.

Thanks. The test6 build (https://github.com/WolframRhodium/VapourSynth-BM3DCUDA/files/8061593/BM3DCUDA_AVS-test6.zip) should work on 780ti and 980ti. Other gpus are too old to be supported.

980ti still requires driver >=452.39.

[tormento]

test6 build
Thanks! I will try it ASAP.

Did you statically linked VC libraries? The dlls seem to me a bit huge in size. Is it possible to have the dynamic version too?

[tormento]

Other gpus are too old to be supported.

I did some benchmarks with SMDegrain on test 5 and test 6 with a 1080p video and x264 --preset slow.

SetMemoryMax()
SetFilterMTMode("DEFAULT_MT_MODE", 2)
LoadPlugin("D:\Eseguibili\Media\DGDecNV\DGDecodeNV.dll")
DGSource("F:\In\2_11 8,1 Departures\departures.dgi",ct=24,cb=24,cl=0,cr=0)
ConvertBits(16)
SMDegrain (tr=3, thSAD=300, refinemotion=true, contrasharp=false, PreFilter=6, plane=4, chroma=true)
fmtc_bitdepth (bits=8,dmode=8)
Prefetch(6)

test 5 5.91 fps

test 6 6.10 fps

tr=4, thSAD=400

test 5 4,99 fps

test 6 4,71 fps

tr=5, thSAD=500

test 5 4,45 fps

test 6 3,58 fps

Any idea about the decreased performance?

[guest]

Thanks. The test6 build (https://github.com/WolframRhodium/VapourSynth-BM3DCUDA/files/8061593/BM3DCUDA_AVS-test6.zip) should work on 780ti and 980ti. Other gpus are too old to be supported.

980ti still requires driver >=452.39.

I'm puzzled that the other 700 Series GPU's are too old, when they are the same gen & series as the 780 Ti, and use the same driver ?!?!?

Also, there is a new nVidia driver(s) out for these cards.

600 & 700 Series cards for W10 & W11, is 472.98 (Std & DCH)

900 Series W10 & W11, is 472.12 Std or 511.70 DCH

Also, I noticed that you've "remove avx requirement", now as BM3D for Avisynth didn't support AVX before, how can it be removed, UNLESS you mean you've added AVX support.

Anyway, I will be testing new build(s) soon.

[Dogway]

@tormento: Try with ex_BM3D() only without other filters, and do like 4 or 5 runs for each to calculate the average to get reliable numbers.

[guest]

Has SMDegrain v3.4.7d been changed back to BM3DCUDA by default ??

[WolframRhodium]

I'm puzzled that the other 700 Series GPU's are too old, when they are the same gen & series as the 780 Ti, and use the same driver ?!?!?

Their compute capabilities (https://developer.nvidia.com/cuda-gpus) (a term used by NV to denote the availability of hardware features) are different. CUDA 11 requires cc >= 3.5 (and deprecates 3.5 <= cc <= 5.0, which may be dropped in CUDA 12).


Also, I noticed that you've "remove avx requirement", now as BM3D for Avisynth didn't support AVX before, how can it be removed, UNLESS you mean you've added AVX support.

Anyway, I will be testing new build(s) soon.

AVX (not AVX2) is required in the previous avs+ builds. Now the requirement is restored in test7 (https://github.com/WolframRhodium/VapourSynth-BM3DCUDA/files/8065185/BM3DCUDA_AVS-test7.zip). test6 is accidentally build with debug config so it should not be used for production, sorry for the mistake.

[guest]

Their compute capability (https://developer.nvidia.com/cuda-gpus) (a term used by NV to denote the availability of hardware features) is different. CUDA 11 requires cc >= 3.5 (and deprecates 3.5 <= cc <= 5.0).



AVX (not AVX2) is required in the previous avs+ builds. Now the requirement is restored in test7 (https://github.com/WolframRhodium/VapourSynth-BM3DCUDA/files/8065185/BM3DCUDA_AVS-test7.zip). test6 is accidentally build with debug config so it should not be used for production, sorry for the mistake.

Hello again, well, I have some news for you then.

I got the latest nVidia drivers for the 600 & 700 series, and your "test#6", and also your #35 & #36 releases on Git.

Now Dogway has changed things around again, I think, yet to deal with his latest versions, anyway I tested #6, still the same, then I tried #35, and it worked on the GTX680 SOC (I made a typo yesterday, I haven't got a 660)...then I tried #36, and it worked too.

I will try "test#7", and let you know.

Cheers

"Test #7" doesn't work for me, but it doesn't matter if the others do...

I just noticed that you posted some more builds on Git, #38 & #42, so I got them, and neither work for me, in fact they didn't even work on the 980Ti PC, where ALL the others do.

[Boulder]

Do you have any recommended deblocking function from your package regarding crappy VC1 sources? I'm currently working on the Star Trek TOS blu-rays, and the quality is far from good. I was thinking that some kind of deblocking (+ possibly adding noise to deceive the brain) might be useful here. Avoiding any serious detail loss would be a big plus.

[kedautinh12]

I think sample is needed. I prefer SmoothD2c, oyster and autodeblock

[Dogway]

I rarely denoise spatially, so usually I apply deblocking as prefilter (a strong deblock is fine) and let SMDegrain clean up the rest. Where block matching fails you are gonna get blocking so use also a deblocker for 'mfilter'.

[Boulder]

Here are two sample clips, the first one shows the blocking in the flat background areas. To my eyes, it looks rather sharp, maybe it's something typical of VC1. The second clip has some closeups and darker areas. It doesn't look as bad because the blocking is not as apparent in motion.

https://drive.google.com/file/d/15PFYjZB2JgFMtGydRr8eA2vg1TxoqyxB/view?usp=sharing
https://drive.google.com/file/d/1ws2B4t9ZiaO1e-LUwBH8Vi2-ntWRQwQf/view?usp=sharing

[tormento]

Try with ex_BM3D() only without other filters, and do like 4 or 5 runs for each to calculate the average to get reliable numbers.
Yes, I always average on 5 runs.

I did a x264 comparison test because, at the end of the day, it's the global performance what matters mostly. I know there could be some minor fluctuations due to video compressibility but on the very same file, with the very same script, unless a big plugin engine modification, they are negligible.

To complete the table:

SetMemoryMax()
SetFilterMTMode("DEFAULT_MT_MODE", 2)
LoadPlugin("D:\Eseguibili\Media\DGDecNV\DGDecodeNV.dll")
DGSource("F:\In\2_11 8,1 Departures\departures.dgi",ct=24,cb=24,cl=0,cr=0)
ConvertBits(16)
SMDegrain (tr=3, thSAD=300, refinemotion=true, contrasharp=false, PreFilter=6, plane=4, chroma=true)
fmtc_bitdepth (bits=8,dmode=8)
Prefetch(6)

test 5 5.91 fps
test 6 6.10 fps
test 7 6,58 fps
test 8 6,03 fps

tr=4, thSAD=400

test 5 4,99 fps
test 6 4,71 fps
test 7 5,01 fps
test 8 5,56 fps

tr=5, thSAD=500

test 5 4,45 fps
test 6 3,58 fps
test 7 4,29 fps
test 8 3,64 fps

I did, anyway, some tests with AVSMeter on plain ex_BM3D() too.

SetMemoryMax()
SetFilterMTMode("DEFAULT_MT_MODE", 2)
LoadPlugin("D:\Eseguibili\Media\DGDecNV\DGDecodeNV.dll")
DGSource("F:\In\2_11 8,1 Departures\departures.dgi",ct=24,cb=24,cl=0,cr=0)
ex_BM3D()
Prefetch(6)

test 5 18,58 fps
test 6 16,55 fps
test 7 19,47 fps
test 8 19,56 fps

But now comes the funny part. If I modify the script as:

SetMemoryMax()
SetFilterMTMode("DEFAULT_MT_MODE", 2)
LoadPlugin("D:\Eseguibili\Media\DGDecNV\DGDecodeNV.dll")
DGSource("F:\In\2_11 8,1 Departures\departures.dgi",ct=24,cb=24,cl=0,cr=0)
ConvertBits(16)
ex_BM3D()
fmtc_bitdepth (bits=8,dmode=8)
Prefetch(6)

test 8 20,86 fps

and with

fmtc_bitdepth (bits=10,dmode=8)

test 8 20,45 fps

P.S: I'd like really to try a version of your scripts where ConvertBits is replaced by fmtc_bitdepth, as I find it faster.

[Boulder]

I rarely denoise spatially, so usually I apply deblocking as prefilter (a strong deblock is fine) and let SMDegrain clean up the rest. Where block matching fails you are gonna get blocking so use also a deblocker for 'mfilter'.

I started experimenting with SmoothD2c (since I've already used SmoothD2 back in the day), line 988 in deblockpack.avsi throws an error that 'fs' is unknown.

zw1 = (Cpr == 0) ? sd1.ex_Ylevels( 0, 1.0, 255, 0, 0, fs) : sd1.buildZmask(cpr=cpr)

After setting false there, the next error is from line 999: SmoothD2: input to filter must be in YV12. This is after loading the source in DGSource, so it's regular 8-bit YV12 data.

[gispos]

Hello Dogway,
I don't know if you will read my post in the confusion of postings here, but I'll throw it in with the other postings.:)

As a note on the AvsPmod slider function definetions:
All parameters listed in the function header must also be present in the slider templates.
Otherwise if the user sets a parameter that is not in the definition, the slider window remains empty (the function is not recognized).

For parameters that should not be displayed or for which no default value should be given, the parameter must be enclosed in a square bracket.
As an example ex_contrast and "UV".

(
clip,
float "cont"=0.00 (-1.00 to 1.00),
float "pivot"=127.5 (12.0 to 192.0 by 0.5),
float "sat"=1.00 (0.00 to 3.00),
[int "UV"],
bool "tv_range"=True,
bool "show"=False
)

[Dogway]

@Boulder, AFK today, I will check SmoothD2c tomorrow, I'm not sure if it still was in WIP or not.

@gispos, very nice I didn't know the trick to omit defaults, as for brackets I use them occasionally but I don't always include all args into sliders as to not make them "busy".

Man, little things here and there and still haven't resumed the SceneStats() filter. I plan to do a few "minor" updates tomorrow and probably resume it.

[Boulder]

Also an issue with GrainFactory3mod.avsi: if you set the tv_range parameter, an error "I don't know what 'tv' means" is shown.

[Boulder]

One more possible issue that I found: at least QTGMC(preset="slower", noisepreset="fast", tr2=0, inputtype=2, noiseprocess=2, grainrestore=0.5, noiserestore=0.3) changes the _ColorRange property from 1 to 0.

While looking at the props, I was thinking that should QTGMC set the _FieldBased property to 0?

[kedautinh12]

I think you need creat issues here
https://github.com/Dogway/Avisynth-Scripts/issues

[Dogway]

Thanks Boulder, just fixed GrainFactory3mod and QTGMC.
EDIT: Also fixed SmoothD2c()

[guest]

Hey Dogway,

I just noticed the update you've done with DeblockPack.avsi (that I don't use)...

Could you please provide an appropriate script or call to use this ??

Sounds interesting with the BM3D "connection".

Thanks

[Dogway]

You mean Oyster? it's feisty2's deblocker/deringing/AA, it's very slow and currently in WIP. I need to dedicate it more time. Read more here (https://forum.doom9.org/showthread.php?t=173470).

@tormento: I benchmarked fmtc_bitdepth() and resulted slower than internal, even using dmode=2.

206.9fps
fmtc_bitdepth(bits=16)
removegrain(12)
fmtc_bitdepth(bits=8,dmode=2)
fmtc_bitdepth(bits=16)
removegrain(12)
fmtc_bitdepth(bits=8,dmode=2)

208.4fps
ConvertBits(16)
removegrain(12)
ConvertBits(8,dither=-1)
ConvertBits(16)
removegrain(12)
ConvertBits(8,dither=-1)

[Boulder]

Got one more changed prop: GrainFactory3mod (at least the MIX version) switches from limited to full range.

[tormento]

@tormento: I benchmarked fmtc_bitdepth() and resulted slower than internal, even using dmode=2.
Geez... It's less than 1%... Won't we have benefits using better dithering modes that fmtc_bitdepth provides?

[Dogway]

Thanks Boulder, fixed already.

Geez... It's less than 1%... Won't we have benefits using better dithering modes that fmtc_bitdepth provides?

Maybe, but all internal down conversions to 8-bit in my scripts use dither=-1 so that's the most used. Other than that is only applied once at the script end typically. I don't think it doesn't make much difference speed wise, in quality though I do think there are benefits in using other than floyd-steinberg but haven't reached to conclusions. One simple way to check dither types is with ex_posterize().

[Zetti]

how to fix this error?

https://i.ibb.co/Q9ktf07/Screenshot-2022-02-17-160108.png

[Boulder]

how to fix this error?

https://i.ibb.co/Q9ktf07/Screenshot-2022-02-17-160108.png

Open that file in Notepad and see how it looks. I haven't had any issues with the latest ExTools.avsi so chances are there's something wrong with the format.

[Zetti]

It looks like a normal script. I can't see anything wrong (it's the first time i have looked in ExTools.avsi)

[Shinkiro]

I can't figure out what I'm not doing wrong, seem all the clips YUV420P16
mthr = 34

ConvertBits(16)
Source=last
mask1b=last.ex_edge("kirsch",6,10)
mask2b=last.ex_edge("kirsch",4,7)
mask=last.ConditionalFilter(mask1b, mask2b, "AverageLuma()",">","50").mt_lut("x "+string(mthr)+" <= x 1 >> x 1 << ?", U=1, V=1).RemoveGrain((980>960) ? 20 : 11, -1)

deb = Source.neo_f3kdb(sample_mode=2, Y=68, Cb=68, Cr=68, grainy=54, grainC=40, range=15, dynamic_grain=true)
ex_merge(deb ,Source ,mask, luma=true, Y=3, UV=3)
returns an error
CombinePlanes: source and target plane dimensions are different
(C:/AviSynth+/plugins64+/ExTools.avsi, line 7072)
(C:/AviSynth+/plugins64+/ExTools.avsi, line 527)

why does he say that it is not so, or does ex_merge not know how to handle 10-16bit?

[Boulder]

It looks like a normal script. I can't see anything wrong (it's the first time i have looked in ExTools.avsi)

In my version (the latest one), the line looks like this and in column 47, there's no [ but A.

z4 = mcfa || mcf == "rMEA" ? Expr(yA[0],yA[1],yA[2],bA[0],bA[1],bA[2],"x a + 0.5 * R@ 0.00390625 * 2 + x a - dup * * y b - dup * 4 * + 255 R - 0.00390625 * 2 + z c - dup * * + sqrt") : nop()

[Zetti]

It looks also like this in the version I have downloaded today.

[Dogway]

returns an error

why does he say that it is not so, or does ex_merge not know how to handle 10-16bit?

Shinkiro, what's the size of your source clip? I think I know what the issue might be, a size where chroma is odd size, but just to be sure.

how to fix this error?

https://i.ibb.co/Q9ktf07/Screenshot-2022-02-17-160108.png

What AVS+ version are you, and AvsPmod version? Also show your script, I haven't run into such errors.

[Shinkiro]

Source BDMV

ID/String : 0 (0x0)
Format/String : BDAV
Format/Info : Blu-ray Video
FileSize/String : 8.76 GiB
Duration/String : 29 min 27 s
OverallBitRate_Mode/String : Variable
OverallBitRate/String : 42.6 Mb/s
OverallBitRate_Maximum/String : 48.0 Mb/s

Video

ID/String : 4113 (0x1011)
MenuID/String : 1 (0x1)
Format/String : AVC
Format/Info : Advanced Video Codec
Format_Profile : High@L4.1
Format_Settings : CABAC / 4 Ref Frames
Format_Settings_CABAC/String : Yes
Format_Settings_RefFrames/String: 4 frames
CodecID : 27
Duration/String : 29 min 27 s
BitRate_Mode/String : Variable
BitRate/String : 38.0 Mb/s
Width/String : 1 920 pixels
Height/String : 1 080 pixels
DisplayAspectRatio/String : 16:9
FrameRate/String : 23.976 (24000/1001) FPS
ColorSpace : YUV
ChromaSubsampling/String : 4:2:0
BitDepth/String : 8 bits
ScanType/String : Progressive
Bits-(Pixel*Frame) : 0.764
StreamSize/String : 7.78 GiB (89%)

[Dogway]

Standard 1080p. It works on my end though I can't see the issue.
What AVS+ version are you on?

Please try with the next and tell me if you still get the error:

ConvertBits(16)
Source=last
mask1b=last.ex_edge("kirsch",6,10)
mask2b=last.ex_edge("kirsch",4,7)
mask=last.ConditionalFilter(mask1b, mask2b, "AverageLuma()",">","50").mt_lut("x "+string(mthr)+" <= x 1 >> x 1 << ?", U=1, V=1).RemoveGrain((980>960) ? 20 : 11, -1)

mskY_to_YYY(last,mask,true,false,3,16)

I only got that error if I tried to use the above over an RGB clip.

[Shinkiro]

Please try with the next and tell me if you still get the error:


with this script I get the same error

CombinePlanes: source and target plane dimensions are different
(C:/AviSynth+/plugins64+/ExTools.avsi, line 7072)

I tried it on another video, the result is the same
Avisynth_3.7.2_test3, ExTools v7.5

[Zetti]

What AVS+ version are you, and AvsPmod version? Also show your script, I haven't run into such errors.

1. Avisynth+ 3.7.2 test 3

2. I do not use AvsPmod

3. I have no script yet because i getting the error when MeGUI has created index file, and trying to load the index file in "MeGUI's Avisynth script creator"

4. FFMS indexer giving me that error, but DGIndexNV indexer giving me this error

https://i.ibb.co/HK506Vq/Screenshot-2022-02-18-084553.png

If i moving your scripts out of the plugin folder, there is no problem when creating index file and basic script i MeGUI

Normally when i download updated scripts and plugins, i just replace the files in the plugin folder and im not getting any errors that i cannot fixing myself

But this time i can't find out what causing these errors

I just want to use SMDegrain

[Dogway]

Maybe MeGUI is parsing the scripts and not understanding the array syntax.
What I can ask you to do is to test your script in AVSmeter so I can know it's not a conflict with my scripts but with MeGUI.

Simply run:
"D:\Rip\workshop\AVSMeter64.exe" "Path/of/your/script.avs"