QTGMC Deinterlacing Script (v3.384) - Page 12

odyssey · 21st December 2016, 16:15

Quote:

Originally Posted by Groucho2004

Consider using Avisynth+ instead of SEt's MT. At this stage it's quite stable and tests have shown good results multi-threading QTGMC().

Just got Avisynth+ up and running. I get about 24fps with this script, but apparently with some spare CPU left (not reaching 100%):

SetFilterMTMode("DEFAULT_MT_MODE", 2)
SetFilterMTMode("DGDecode_mpeg2source", 3)
DGDecode_mpeg2source(...)
AssumeTFF()
ColorMatrix(hints=true, threads=0)
QTGMC(Preset="Slow")
Prefetch(4)

Is it possible to tweak better for performance?

Groucho2004 · 21st December 2016, 16:32

Quote:

Originally Posted by odyssey

Just got Avisynth+ up and running. I get about 24fps with this script, but apparently with some spare CPU left (not reaching 100%):

SetFilterMTMode("DEFAULT_MT_MODE", 2)
SetFilterMTMode("DGDecode_mpeg2source", 3)
DGDecode_mpeg2source(...)
AssumeTFF()
ColorMatrix(hints=true, threads=0)
QTGMC(Preset="Slow")
Prefetch(4)

Is it possible to tweak better for performance?

Since you have a i7-6700K I'd try a prefetch value of 6 or 8. However, watch the memory consumption of the script. Also, set both "EdiThreads" and "DftThreads" to "1" in QTGMC().

Most importantly, higher CPU usage doesn't necessarily mean better performance and/or efficiency.

ndjamena · 23rd December 2016, 02:46

For whoever it may concern:

BUG (or at least an inconsistency)

Code:

EdiQual    = (EdiMode == "NNEDI3" && EdiQual > 2 ) ? 2 : EdiQual  # Smaller range for EdiQual in NNEDI3

Code:

(EdiMode == "EEDI3+NNEDI3") ? Input.EEDI3( field=-2, mdis=EdiMaxD, threads=EdiThreads, U=CEed,V=CEed, \
       sclip=Input.NNEDI3( field=-2, nsize=NNSize, nns=NNeurons, qual=EdiQual, threads=EdiThreads, U=CEed,V=CEed ) ) : \

The script fails to check the maximum EdiQual value for NNEDI3 when used with EEDI3.

real.finder · 23rd December 2016, 08:16

Quote:

Originally Posted by ndjamena

For whoever it may concern:

BUG (or at least an inconsistency)

Code:

EdiQual    = (EdiMode == "NNEDI3" && EdiQual > 2 ) ? 2 : EdiQual  # Smaller range for EdiQual in NNEDI3

Code:

(EdiMode == "EEDI3+NNEDI3") ? Input.EEDI3( field=-2, mdis=EdiMaxD, threads=EdiThreads, U=CEed,V=CEed, \
	sclip=Input.NNEDI3( field=-2, nsize=NNSize, nns=NNeurons, qual=EdiQual, threads=EdiThreads, U=CEed,V=CEed ) ) : \

The script fails to check the maximum EdiQual value for NNEDI3 when used with EEDI3.

yes, it is, fix it in my edit

tsftd · 28th December 2016, 16:21

OK, so having spent several hours messing around through Google and not being able to get it to work, hopefully someone here can help me.

I'm encoding VHS/Super8 rips of old home movies, and looking to properly deinterlace them, so I'd like to get QTGMC working. I'm using megui (dev updates), and not trying to use multithreading. I've used Megui for years, but haven't really messed around with external filters/plugins before.

My AviSynth Script Creator template looks like:

Code:

SetMemoryMax(1024)
<input>
<crop>
<resize>
QTGMC( Preset="Slow" )
<denoise>

The outputted .avs file is:

Code:

SetMemoryMax(1024)
LoadPlugin("C:\Megui\tools\lsmash\LSMASHSource.dll")
LWLibavVideoSource("C:\encoding test\test2.avi")
#crop
#resize
QTGMC( Preset="Slow" )
#denoise

Please note that trying to manually load most of the DLLs results in an error -- eg,

Code:

'LoadPlugin("C:\Megui\tools\avs\plugins\SSE2Tools.dll")' gives:
AviSynth script error:
LoadPlugin: unable to load "C:\Megui\tools\avs\plugins\SSE2Tools.dll", Module not found.
Install missing library?

Preview AVS Script in the AviSynth Script Creator yields the error:

Code:

Avisynth script error:
Script error: there is no function named "Interleaved2Planar"
(QTGMC.avsi, line 350)

Trying to run the .avs script in megui yields the error:

Code:

The file C:\encoding test\test2.avi\test2.avs cannot be opened.
Please make sure it's a valid AviSynth script and that AviSynth is properly installed
Error message for your reference: Can't read the video stream

Instructions followed:

From the OP in this thread

From http://avisynth.nl/index.php/QTGMC

Troubleshooting from http://forum.doom9.org/showthread.ph...48#post1485948

Troubleshooting from http://forum.videohelp.com/threads/3...eird-artifacts

Troubleshooting from https://www.vegascreativesoftware.in...-84017/?page=3

etc...

Files in my plugin folder:

Code:

AddGrainC.dll
colors_rgb.avsi
dfttest.dll
DirectShowSource.dll
EEDI2.dll
eedi3.dll
FFT3DFilter.dll
fftw3.dll
libfftw3f-3.dll
masktools2.dll
mvtools2.dll
nnedi.dll
nnedi2.dll
nnedi3.dll
QTGMC.avsi
RemoveGrainSSE2.dll
RepairSSE2.dll
SSE2Tools.dll
SSE3Tools.dll
SSETools.dll
TCPDeliver.dll
TDeint.dll
VerticalCleanerSSE2.dll
yadif.dll

And yes, fftw3.dll and libfftw3f-3.dll are both in the SysWOW64 folder as well.

I am using the built-in avisynth (2.6); I have an external AviSynth with the same plugins in the folder, but the current MeGui doesn't appear to have an option to force usage of the external AviSynth.

And yes, MeGUI works fine if I remove the 'QTGMC( Preset="Slow" )'

Any help would be much appreciated, as I've pretty much exhausted my troubleshooting abilities and will have to settle for standard YADIF if someone here can't help me.

kuchikirukia · 28th December 2016, 16:51

Quote:

Originally Posted by tsftd

Please note that trying to manually load most of the DLLs results in an error -- eg,

Code:

'LoadPlugin("C:\Megui\tools\avs\plugins\SSE2Tools.dll")' gives:
AviSynth script error:
LoadPlugin: unable to load "C:\Megui\tools\avs\plugins\SSE2Tools.dll", Module not found.
Install missing library?

Are you sure they're not in avisynth_plugin?

Also you can replace removegrain and repair with rgtools.

StainlessS · 28th December 2016, 17:10

SSE2tools requires MSVCR71.dll from VS 2003 Dot Net I think. (EDIT not sure about the other versions of SSETools)
see here:- http://forum.doom9.org/showthread.ph...ll#post1781169

EDIT: And you only need one of the SSETools (I'de choose SSE2) in your plugs folder.

Groucho2004 · 28th December 2016, 17:35

Quote:

Originally Posted by StainlessS

SSE2tools requires MSVCR71.dll from VS 2003 Dot Net I think.

By the way, ricktendo's AIO also installs these DLLs (msv*70.dll, msv*71.dll).

StainlessS · 28th December 2016, 18:42

Quote:

Originally Posted by Groucho2004

By the way, ricktendo's AIO also installs these DLLs (msv*70.dll, msv*71.dll).

oooooOOO, thats new I think.

Groucho2004 · 28th December 2016, 18:44

Quote:

Originally Posted by StainlessS

oooooOOO, thats new I think.

Don't think so. These were included the first time I used it (last year?).

StainlessS · 28th December 2016, 19:09

Thank you G2K4, guess I can now get rid of my vs 2002 and vs 2003 suppositories

LigH · 28th December 2016, 20:29

But, but, but ... they are mostly obsolete since RGTools substitutes the whole "Repair" pack all together.

Groucho2004 · 28th December 2016, 22:44

Quote:

Originally Posted by LigH

But, but, but ... they are mostly obsolete since RGTools substitutes the whole "Repair" pack all together.

Of course they are obsolete. People simply don't know that there are newer (and better) versions. Just look at the first post of this thread, it still promotes the use of these outdated plugins and the linked packages also contain them (I'm not blaming Vit, he has not been around for a while).

A sticky in the Avisynth usage forum with the current recommended plugins may be a good idea.

tsftd · 29th December 2016, 05:20

Solved! Thank everyone for their help, I never would have been able to do it without you!

All msvcr requirements are installed in system.

RemoveGrain & Repair replaced with RGTools

SSE & SSE3 removed (had added as several troubleshooting threads as linked said that they resolved issues for the posters)

Yes, all plugins are in "C:\Megui\tools\avs\plugins"

It has no trouble loading RGTools.dll or QTGMC.avsi, in the same folder.

The original issue is resolved, however now:

Code:

AviSynth script error:
mt_makediff: unsupported colorspace. masktools only support planar YUV colorspaces (YV12, YV16, YV24)
(QTGMC.avsi, line 776)
(QTGMC.avsi, line 386)

Any other ideas?

New config:

Code:

LoadPlugin("C:\Megui\tools\avs\plugins\RgTools.dll")
SetMemoryMax(1024)
<input>
<crop>
<resize>
QTGMC( Preset="Slow" )
<denoise>

Note:
Using Masktools2.dll from http://avisynth.nl/index.php/MaskTools2 (x86 version)
Also tried mt_masktools-25.dll and mt_masktools-26.dll (each time, removing other masktools versions and trying it with only the target in the plugin directory)

Edit:
<SOLVED> via inserting "ConvertToYV12(interlaced=true)" before calling QTGMC:

Code:

LoadPlugin("C:\Temp Programs\Megui\tools\avs\plugins\RgTools.dll")
SetMemoryMax(1024)
<input>
<crop>
<resize>
ConvertToYV12(interlaced=true)
QTGMC( Preset="Slow" )
<denoise>

LigH · 29th December 2016, 10:23

Oh, you wrote "VHS"; so we should have assumed that your source filter returned YUY2 instead of YV12? Well, yes, QTGMC may support YUY2, but with some quirks. And the latest versions just try to substitute the slower "packed pixel" format YUY2 with the faster "planar" format YV16, available in AviSynth 2.6 or AviSynth+.

ErazorTT · 3rd January 2017, 03:47

I have made some tests and it appears to me that the internal dither of dfttest is significantly faster (~10%) than the construct currently used in version 3.34 with lsb and ditherpost.
Are there any side effects or why isn't it beeing used?

real.finder · 3rd January 2017, 08:26

Quote:

Originally Posted by ErazorTT

I have made some tests and it appears to me that the internal dither of dfttest is significantly faster (~10%) than the construct currently used in version 3.34 with lsb and ditherpost.
Are there any side effects or why isn't it beeing used?

don't know which QTGMC you use but in my edit if you didn't set lsbd=true then it will work with internal dither

ndjamena · 3rd January 2017, 08:36

Ideas:

1: Black and White mode.
There are a bunch of "ignore chroma" settings that all need to be set independently. It would be easier if you could just say it once and have the script itself apply all the necessary settings. Plus, there doesn't seem to be any way of disabling Chroma EDI completely. The lowest setting is "Bob" and for black and white, even that is overkill.

2: Search Presets. It should default to the same as the main preset, but search results can be shared across instances. If you don't write in the exact same settings for each instance of QTGMC the search settings are rejected and the script aborts. If you run something like srestore with an incredibly slow preset for final output and an incredibly fast preset for detection, the motion detection results can't be shared across the two instances without writing in a bunch of manual overrides.

ErazorTT · 4th January 2017, 17:08

Quote:

Originally Posted by real.finder

don't know which QTGMC you use but in my edit if you didn't set lsbd=true then it will work with internal dither

No, setting lsbd=true uses ditherpost, not dfttests internal dither. As can be seen looking at the source of version 3.34 (Line 501):

Code:

dnwindow =    (Denoiser == "dfttest") && (NoiseProcess != 0) && lsbd ? dnWindow.ditherpost(mode=6, U=ChromaNoise?3:2, V=ChromaNoise?3:2, slice=false) : dnWindow

One can see that in case lsbd is true, and thus dfttest outputs a 16bit clip, the clip is dithered using ditherpost and not by using dfttest's internal dither.

What I mean is replacing the following code (starting at line 496):

Code:

dnWindow = (NoiseProcess == 0)      ? NOP() : \
           (Denoiser == "dfttest")  ? noiseWindow.dfttest( Y=true, U=ChromaNoise, V=ChromaNoise, sigma=Sigma*4, tbsize=noiseTD, threads=DftThreads, lsb=lsbd ) : \
        (Denoiser == "KNLMeansCL")  ? KNLMeansCL( lsbd ? noiseWindow.Dither_convert_8_to_16() : noiseWindow, d=NoiseTR, h=Sigma, lsb_inout=lsbd, device_type="GPU") : \
                                      IsYV12(noiseWindow) ? noiseWindow.FFT3DFilter( plane=(ChromaNoise ? 4 : 0), sigma=Sigma, bt=noiseTD ) : \
                                      noiseWindow.ConvertToYUY2().FFT3DFilter( plane=(ChromaNoise ? 4 : 0), sigma=Sigma, bt=noiseTD ).ConvertToYV16()
dnwindow =    (Denoiser == "dfttest") && (NoiseProcess != 0) && lsbd ? dnWindow.ditherpost(mode=6, U=ChromaNoise?3:2, V=ChromaNoise?3:2, slice=false) : dnWindow
dnwindow = (Denoiser == "KNLMeansCL") && (NoiseProcess != 0) && lsbd ? dnWindow.ditherpost(mode=6, U=ChromaNoise?3:2, V=ChromaNoise?3:2, slice=false) : dnWindow

With this:

Code:

dnWindow = (NoiseProcess == 0)      ? NOP() : \
           (Denoiser == "dfttest")  ? noiseWindow.dfttest( Y=true, U=ChromaNoise, V=ChromaNoise, sigma=Sigma*4, tbsize=noiseTD, threads=DftThreads, dither=lsbd ) : \
        (Denoiser == "KNLMeansCL")  ? KNLMeansCL( lsbd ? noiseWindow.Dither_convert_8_to_16() : noiseWindow, d=NoiseTR, h=Sigma, lsb_inout=lsbd, device_type="GPU") : \
                                      IsYV12(noiseWindow) ? noiseWindow.FFT3DFilter( plane=(ChromaNoise ? 4 : 0), sigma=Sigma, bt=noiseTD ) : \
                                      noiseWindow.ConvertToYUY2().FFT3DFilter( plane=(ChromaNoise ? 4 : 0), sigma=Sigma, bt=noiseTD ).ConvertToYV16()
dnwindow =    (Denoiser == "dfttest") && (NoiseProcess != 0) && lsbd ? dnWindow : dnWindow #noop since dithering already done by dfttest internally
dnwindow = (Denoiser == "KNLMeansCL") && (NoiseProcess != 0) && lsbd ? dnWindow.ditherpost(mode=6, U=ChromaNoise?3:2, V=ChromaNoise?3:2, slice=false) : dnWindow

The dithering by dfttest is significantly faster for my system than by ditherpost.

LigH · 4th January 2017, 17:28

Quote:

Originally Posted by ErazorTT

No, setting lsbd=true uses ditherpost, not dfttests internal dither.

That was the point of "if you didn't set lsbd=true then it will work with internal dither". Negation.

23rd December 2016, 02:46	#223 \| Link
ndjamena Registered User Join Date: Sep 2012 Posts: 366	For whoever it may concern: BUG (or at least an inconsistency) Code: EdiQual = (EdiMode == "NNEDI3" && EdiQual > 2 ) ? 2 : EdiQual # Smaller range for EdiQual in NNEDI3 Code: (EdiMode == "EEDI3+NNEDI3") ? Input.EEDI3( field=-2, mdis=EdiMaxD, threads=EdiThreads, U=CEed,V=CEed, \ sclip=Input.NNEDI3( field=-2, nsize=NNSize, nns=NNeurons, qual=EdiQual, threads=EdiThreads, U=CEed,V=CEed ) ) : \ The script fails to check the maximum EdiQual value for NNEDI3 when used with EEDI3. Last edited by tebasuna51; 21st April 2017 at 12:45.

28th December 2016, 16:21	#225 \| Link
tsftd Registered User Join Date: Dec 2011 Posts: 6	OK, so having spent several hours messing around through Google and not being able to get it to work, hopefully someone here can help me. I'm encoding VHS/Super8 rips of old home movies, and looking to properly deinterlace them, so I'd like to get QTGMC working. I'm using megui (dev updates), and not trying to use multithreading. I've used Megui for years, but haven't really messed around with external filters/plugins before. My AviSynth Script Creator template looks like: Code: SetMemoryMax(1024) <input> <crop> <resize> QTGMC( Preset="Slow" ) <denoise> The outputted .avs file is: Code: SetMemoryMax(1024) LoadPlugin("C:\Megui\tools\lsmash\LSMASHSource.dll") LWLibavVideoSource("C:\encoding test\test2.avi") #crop #resize QTGMC( Preset="Slow" ) #denoise Please note that trying to manually load most of the DLLs results in an error -- eg, Code: 'LoadPlugin("C:\Megui\tools\avs\plugins\SSE2Tools.dll")' gives: AviSynth script error: LoadPlugin: unable to load "C:\Megui\tools\avs\plugins\SSE2Tools.dll", Module not found. Install missing library? Preview AVS Script in the AviSynth Script Creator yields the error: Code: Avisynth script error: Script error: there is no function named "Interleaved2Planar" (QTGMC.avsi, line 350) Trying to run the .avs script in megui yields the error: Code: The file C:\encoding test\test2.avi\test2.avs cannot be opened. Please make sure it's a valid AviSynth script and that AviSynth is properly installed Error message for your reference: Can't read the video stream Instructions followed: From the OP in this thread From http://avisynth.nl/index.php/QTGMC Troubleshooting from http://forum.doom9.org/showthread.ph...48#post1485948 Troubleshooting from http://forum.videohelp.com/threads/3...eird-artifacts Troubleshooting from https://www.vegascreativesoftware.in...-84017/?page=3 etc... Files in my plugin folder: Code: AddGrainC.dll colors_rgb.avsi dfttest.dll DirectShowSource.dll EEDI2.dll eedi3.dll FFT3DFilter.dll fftw3.dll libfftw3f-3.dll masktools2.dll mvtools2.dll nnedi.dll nnedi2.dll nnedi3.dll QTGMC.avsi RemoveGrainSSE2.dll RepairSSE2.dll SSE2Tools.dll SSE3Tools.dll SSETools.dll TCPDeliver.dll TDeint.dll VerticalCleanerSSE2.dll yadif.dll And yes, fftw3.dll and libfftw3f-3.dll are both in the SysWOW64 folder as well. I am using the built-in avisynth (2.6); I have an external AviSynth with the same plugins in the folder, but the current MeGui doesn't appear to have an option to force usage of the external AviSynth. And yes, MeGUI works fine if I remove the 'QTGMC( Preset="Slow" )' Any help would be much appreciated, as I've pretty much exhausted my troubleshooting abilities and will have to settle for standard YADIF if someone here can't help me.

28th December 2016, 17:10	#227 \| Link
StainlessS HeartlessS Usurer Join Date: Dec 2009 Location: Over the rainbow Posts: 11,410	SSE2tools requires MSVCR71.dll from VS 2003 Dot Net I think. (EDIT not sure about the other versions of SSETools) see here:- http://forum.doom9.org/showthread.ph...ll#post1781169 EDIT: And you only need one of the SSETools (I'de choose SSE2) in your plugs folder. __________________ I sometimes post sober. StainlessS@MediaFire ::: AND/OR ::: StainlessS@SendSpace "Some infinities are bigger than other infinities", but how many of them are infinitely bigger ??? Last edited by StainlessS; 28th December 2016 at 17:14.

28th December 2016, 19:09	#231 \| Link
StainlessS HeartlessS Usurer Join Date: Dec 2009 Location: Over the rainbow Posts: 11,410	Thank you G2K4, guess I can now get rid of my vs 2002 and vs 2003 suppositories __________________ I sometimes post sober. StainlessS@MediaFire ::: AND/OR ::: StainlessS@SendSpace "Some infinities are bigger than other infinities", but how many of them are infinitely bigger ???

28th December 2016, 20:29	#232 \| Link
LigH German doom9/Gleitz SuMo Join Date: Oct 2001 Location: Germany, rural Altmark Posts: 7,439	But, but, but ... they are mostly obsolete since RGTools substitutes the whole "Repair" pack all together. __________________ New German Gleitz board MediaFire: x264 \| x265 \| VPx \| AOM \| Xvid

29th December 2016, 05:20	#234 \| Link
tsftd Registered User Join Date: Dec 2011 Posts: 6	Solved! Thank everyone for their help, I never would have been able to do it without you! All msvcr requirements are installed in system. RemoveGrain & Repair replaced with RGTools SSE & SSE3 removed (had added as several troubleshooting threads as linked said that they resolved issues for the posters) Yes, all plugins are in "C:\Megui\tools\avs\plugins" It has no trouble loading RGTools.dll or QTGMC.avsi, in the same folder. The original issue is resolved, however now: Code: AviSynth script error: mt_makediff: unsupported colorspace. masktools only support planar YUV colorspaces (YV12, YV16, YV24) (QTGMC.avsi, line 776) (QTGMC.avsi, line 386) Any other ideas? New config: Code: LoadPlugin("C:\Megui\tools\avs\plugins\RgTools.dll") SetMemoryMax(1024) <input> <crop> <resize> QTGMC( Preset="Slow" ) <denoise> Note: Using Masktools2.dll from http://avisynth.nl/index.php/MaskTools2 (x86 version) Also tried mt_masktools-25.dll and mt_masktools-26.dll (each time, removing other masktools versions and trying it with only the target in the plugin directory) Edit: <SOLVED> via inserting "ConvertToYV12(interlaced=true)" before calling QTGMC: Code: LoadPlugin("C:\Temp Programs\Megui\tools\avs\plugins\RgTools.dll") SetMemoryMax(1024) <input> <crop> <resize> ConvertToYV12(interlaced=true) QTGMC( Preset="Slow" ) <denoise> Last edited by tsftd; 29th December 2016 at 05:44.

29th December 2016, 10:23	#235 \| Link
LigH German doom9/Gleitz SuMo Join Date: Oct 2001 Location: Germany, rural Altmark Posts: 7,439	Oh, you wrote "VHS"; so we should have assumed that your source filter returned YUY2 instead of YV12? Well, yes, QTGMC may support YUY2, but with some quirks. And the latest versions just try to substitute the slower "packed pixel" format YUY2 with the faster "planar" format YV16, available in AviSynth 2.6 or AviSynth+. __________________ New German Gleitz board MediaFire: x264 \| x265 \| VPx \| AOM \| Xvid

3rd January 2017, 03:47	#236 \| Link
ErazorTT Registered User Join Date: Mar 2003 Location: Germany Posts: 215	dfttest internal dither instead of ditherpost I have made some tests and it appears to me that the internal dither of dfttest is significantly faster (~10%) than the construct currently used in version 3.34 with lsb and ditherpost. Are there any side effects or why isn't it beeing used?

3rd January 2017, 08:36	#238 \| Link
ndjamena Registered User Join Date: Sep 2012 Posts: 366	Ideas: 1: Black and White mode. There are a bunch of "ignore chroma" settings that all need to be set independently. It would be easier if you could just say it once and have the script itself apply all the necessary settings. Plus, there doesn't seem to be any way of disabling Chroma EDI completely. The lowest setting is "Bob" and for black and white, even that is overkill. 2: Search Presets. It should default to the same as the main preset, but search results can be shared across instances. If you don't write in the exact same settings for each instance of QTGMC the search settings are rejected and the script aborts. If you run something like srestore with an incredibly slow preset for final output and an incredibly fast preset for detection, the motion detection results can't be shared across the two instances without writing in a bunch of manual overrides.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode