Dogway's Filters Packs

[anton_foy]

https://forum.doom9.org/showthread.php?p=1953018#post1953018

It needs a low pass recover pass that I will add in the future to avoid blending/motion blur effect.

I know this is an old post but did you add this feature later on?

[Dogway]

https://forum.doom9.org/showthread.php?p=1953018#post1953018

I know this is an old post but did you add this feature later on?

I guess it's the 'LFR' option, the higher the more you recover.

I'm taking a few off days, will resume to finish the HDR functions and the scale_space logic, I've been brewing some ideas.

[VoodooFX]

What is "SMPTE_legal" and "nmod" in your Stabilization Tools Pack?
Btw, Stab there duplicates with original, maybe rename it to Stab4.

[Dogway]

Stabilization Tools Pack is considered legacy. It needs a serious rehaul but it's not easy task.
nmod (https://github.com/Dogway/Avisynth-Scripts/blob/58cd3bc0aba9d41dbdad913095783fd490d5d0f8/ResizersPack.avsi#L1200)() is to round numbers to multiples of modulo. Reduces boilerplate code like (n/mod)*mod and adds some features.
SMPTE_legal (https://github.com/Dogway/Avisynth-Scripts/blob/58cd3bc0aba9d41dbdad913095783fd490d5d0f8/TransformsPack%20-%20Transfers.avsi#L1185)() is to convert clip range into SMPTE legal range (from full saturated range) or the inverse operation. Also adds some features like SMPTE+ (aka broadcast range), and decoupled Y/UV conversion.
What is the original Stab?

[VoodooFX]

There is no mention of "ResizersPack.avsi" & "TransformsPack - Transfers.avsi" in requirements, I was wondering why it's not working.

Original Stab: http://avisynth.nl/index.php/Stab

[Dogway]

It's slightly updated internally but didn't upload cause it needs a bigger refactor. I just uploaded what I have but it's still considered legacy.

[guest]

Does Whitepoint() require MT, it's just that when I use it, or add it another script, the encoding speed drop's of a cliff :(, especially with 4K HDR.
It's not pushing the CPU anywhere near 100%.

[kedautinh12]

Add prefetch

[guest]

Add prefetch

I was just coming back to update my post...there is an option within RipBot where you can add certain filters/scripts to a multi-thread option, I guess it's a prefetech of some kind, and it's made a HUGE difference, the CPU is running at 100% most of the time :)

[Dogway]

If you want more speed you can exploit the function by replacing YUV_to_RGB() and RGB_toYUV() with avsresize.

[guest]

If you want more speed you can exploit the function by replacing YUV_to_RGB() and RGB_toYUV() with avsresize.

Thanks, but I think I'm going to need an example (or 2)

Sorry.

[Dogway]

Replace lines 1484,1485 and 1582,1583 with:
z_ConvertFormat(pixel_type="RGBP"+string(ID[7]),resample_filter="spline36")
and
z_ConvertFormat(pixel_type="YUV420P"+string(ID[7]),resample_filter="spline16")
respectively

[guest]

Replace lines 1484,1485 and 1582,1583 with:
z_ConvertFormat(pixel_type="RGBP"+string(ID[7]),resample_filter="spline36")
and
z_ConvertFormat(pixel_type="YUV420P"+string(ID[7]),resample_filter="spline16")
respectively

This doesn't make too much sense to me...

This is changing GradePack.avsi directly, correct ??

https://github.com/Dogway/Avisynth-Scripts/blob/master/GradePack.avsi

So these are lines 1484 & 1485 :-

https://github.com/Dogway/Avisynth-Scripts/blob/master/GradePack.avsi#L1484

rgb ? rgbp ? c : ConvertToPlanarRGB(c) : \
YUV_to_RGB(c, mat[0], tv_in=tv, tv_out=false, kernel="lanczos", b=3)

And these are lines 1582 & 1583 :-

https://github.com/Dogway/Avisynth-Scripts/blob/master/GradePack.avsi#L1582

rgb ? rgbp ? last : MatchClip(c,internal=true) : \
RGB_to_YUV( mat[0], tv_in=false, tv_out=tv, kernel="bicubic", b=-0.5, c=0.25, Jab=PixelType(c))

How can replacing 2 lines with one, work...I don't understand :(

Do I need to add avsresize to the script ??

I'm clearly in the wrong place, here.

[kedautinh12]

You can change
YUV_to_RGB(c, mat[0], tv_in=tv, tv_out=false, kernel="lanczos", b=3)to
z_ConvertFormat(pixel_type="RGBP"+string(ID[7]),resample_filter="spline36")

And
RGB_to_YUV( mat[0], tv_in=false, tv_out=tv, kernel="bicubic", b=-0.5, c=0.25, Jab=PixelType(c))to
z_ConvertFormat(pixel_type="YUV420P"+string(ID[7]),resample_filter="spline16")

[guest]

You can change

Thankyou my friend, I will try this tomorrow :)

[StainlessS]

You can change
...
to
z_ConvertFormat(pixel_type="RGBP"+string(ID[7]),resample_filter="spline36")
And
...
to
z_ConvertFormat(pixel_type="YUV420P"+string(ID[7]),resample_filter="spline16")


Looks to me like too many closing parentheses ")"

[kedautinh12]

Looks to me like too many closing parentheses ")"

I just copy from Dogway :D

[VoodooFX]

Looks to me like too many closing parentheses ")"
Looks OK for me.

[StainlessS]

Yeah, sorry, I missed this one,
z_ConvertFormat(pixel_type="RGBP"+string(ID[7]),resample_filter="spline36")
Old Age :(

[VoodooFX]

Old Age :(
Happens, we will train ChatGPT on your posts, and after you are gone we'll replace you with StainlessS-GPT. It will continue to help us with the scripts and maybe even it will write the new nice plugins for us. :sly:

[StainlessS]

Somebody just released a Chat Bot thingy to tell if essay/article was written by human or bot.
Was in the press either yesterday or day before.

EDIT: Maybe it was an Android/iPad app, accessing a chatbot whotsit.
Think I maybe saw it linked via CodeProject stuff I get via email.

[guest]

Thankyou my friend, I will try this tomorrow :)

Originally Posted by kedautinh12
You can change

YUV_to_RGB(c, mat[0], tv_in=tv, tv_out=false, kernel="lanczos", b=3) original line 1485
to
z_ConvertFormat(pixel_type="RGBP"+string(ID[7]),resample_filter="spline36")
And
RGB_to_YUV( mat[0], tv_in=false, tv_out=tv, kernel="bicubic", b=-0.5, c=0.25, Jab=PixelType(c)) original line 1583
to
z_ConvertFormat(pixel_type="YUV420P"+string(ID[7]),resample_filter="spline16")

So despite all the comments about ), I am getting an error with Line 1485.

And before anyone ask's, I cannot provide an error log :(

https://github.com/Dogway/Avisynth-Scripts/blob/master/GradePack.avsi#L1485

Which strangely is the first one to be changed :(

So there's a possibility that line 1583 will produce a similar error.

[guest]

Yeah, sorry, I missed this one,
z_ConvertFormat(pixel_type="RGBP"+string(ID[7]),resample_filter="spline36")
Old Age :(

Speaking of which :-

It would be nice if this was updated :-

https://forum.doom9.org/showthread.php?t=176546

[kedautinh12]

Don't ask age of ssS, it's no meaning. He/she is a android :D

[StainlessS]

All I/we need is a little squirt of oil:- https://www.youtube.com/watch?v=louBM-Mix7s

EDIT: To below: thanks, nice of you to say so :)

EDIT: To Below below: So, was not a total waste of time, I thought it way too subtle for the average D9'er. :)

[guest]

All I need is a little squirt of oil:- https://www.youtube.com/watch?v=louBM-Mix7s

HeartlessS

EDIT: To above: You're welcome, I got the connection from the ending of that YT clip... :D

[anton_foy]

In MCDeblockFFT3D I noticed that you use "pre" (which is converted to 8bits) before MCompensate followed by convertbits(bi,etc...). Is there a particular reason that you need to use mcompensate in 8bits here? Also when I tried this script but replaced with only deblock() after compensation I get weird blocking artefacts on higher motion.

[Dogway]

In MCDeblockFFT3D I noticed that you use "pre" (which is converted to 8bits) before MCompensate followed by convertbits(bi,etc...). Is there a particular reason that you need to use mcompensate in 8bits here? Also when I tried this script but replaced with only deblock() after compensation I get weird blocking artefacts on higher motion.

If I don't do that I have to create another super clip for the original bitdepth clip. All in all turns to slow down the script quite a bit, FFT3D is already hella slow anyway...

[anton_foy]

If I don't do that I have to create another super clip for the original bitdepth clip. All in all turns to slow down the script quite a bit, FFT3D is already hella slow anyway...

Ok thanks I see now. But still I don't understand why I get these strange artifacts.

[Dogway]

Just checked with default MCDeblockFFT3D() and it's not blocking but bad block matching of MCompensate. This was a very early approach to motion compensated deblocking back in the days. Currently I would recommend you to feed a deblocked prefilter to SMDegrain, so MDegrain can do the temporal deblock, and probably feed the same clip to Mfilter to deblock spatially on non-matching blocks.

[madey83]

Just checked with default MCDeblockFFT3D() and it's not blocking but bad block matching of MCompensate. This was a very early approach to motion compensated deblocking back in the days. Currently I would recommend you to feed a deblocked prefilter to SMDegrain, so MDegrain can do the temporal deblock, and probably feed the same clip to Mfilter to deblock spatially on non-matching blocks.

@Dogway

Could you please provide some example how to add deblocking into SMDegrain and Mfilter?

[Dogway]

It's straight forward:
pre=Deblock_QED(quant1=30, quant2=40) # You can (it's safe to) go overboard or use more destructive deblockers here
SMDegrain(2, 200, mode="MDegrain", prefilter=pre, mfilter=pre)

You might be tempted to simply use deblock before smdegrain (as input), but that's blurrier.

[madey83]

It's straight forward:
pre=Deblock_QED(quant1=30, quant2=40) # You can (it's safe to) go overboard or use more destructive deblockers here
SMDegrain(2, 200, mode="MDegrain", prefilter=pre, mfilter=pre)

You might be tempted to simply use deblock before smdegrain (as input), but that's blurrier.

Thank you 😊

[Emulgator]

Just throwing in a suggestion to extend QTGMCp requirements list
# AviSynth 3.7.3 (because of propcopy extended parameters)
(I was searching today for dependencies, it took some good time to get these all together and finally serving the first frame.)

And while we are at it:
Who was unlucky so far to get Groucho's sysinfo64.dll running:
sysinfo64.dll wants msvcr100.dll 10.0.30319.1 (18.03.2010 809KB) in system32
https://forum.doom9.org/showthread.php?p=1981167#post1981167

[kedautinh12]

I remember still use normally QTGMCp with lastest ver of sysinfo. I think you can try install latest vc++ pack
https://github.com/abbodi1406/vcredist/releases

[Dogway]

Visual C++ 10.0.30319.1 ? I don't have that but VCPP 2010 10.0.40219.325 and it works.

Maybe it works with any of the 2010 family.

I will try to check when propcopy extended params were implemented, because if that's so I'll need to promote all scripts to 3.7.3 requirement.

[kedautinh12]

Visual C++ 10.0.30319.1 ? I don't have that but VCPP 2010 10.0.40219 and it works.

Maybe it works with any of the 2010 family.

I will try to check when propcopy extended params were implemented, because if that's so I'll need to promote all scripts to 3.7.3 requirement.

Cause you have later ver, you meet error only when have former ver

[Dogway]

@Emulgator, the propcopy extended features belong to 3.7.2 (https://github.com/AviSynth/AviSynthPlus/releases/tag/v3.7.2), unless it's something else that requires v3.7.3.

EDIT: It must be the new NNEDI3CL requirements (https://github.com/Asd-g/AviSynthPlus-NNEDI3CL#requirements), r3682 (r3689 recommended) and VCPP. The exact VCPP I don't know, maybe someone can have a look at dependency walker. And r3682 is basically v3.7.3 right?

[LeXXuz]

New year and I think it's time to do some house keeping. :D

What is the deal with this AVS+ 3.7.3?

I updated QTGMC+ and all other things that have been changed/updated recently (gee, I really could use an auto-updater for all those files :D) because I could not get QTGMC+ to work, even with default settings. MeGUI simply crashed upon script load and sadly didn't tell me where it hurts.

Now, after updating the files QTGMC+ requires 3.7.3 but there is still 3.7.2 the latest version on avs-plus.net :confused:

And if I'm just at it: which version of NNEDI3CL/EEDI3CL should I use and (where to get)?

Also which version of mvtools? Still the kinda old 2.7.45? What about DTL's latest versions? I tried those, but couldn't get them to work with SMDegrain.

I kinda lost overview with all those updated files, I know. :o

[kedautinh12]

I use latest ver of all with Megui normally. About MVtool still use Pinterf stable ver than DTL beta build

[kedautinh12]

If Megui crash don't alert error. You can try AvsPMod, errors will appear

[guest]

New year and I think it's time to do some house keeping. :D

What is the deal with this AVS+ 3.7.3?

I updated QTGMC+ and all other things that have been changed/updated recently (gee, I really could use an auto-updater for all those files :D) because I could not get QTGMC+ to work, even with default settings. MeGUI simply crashed upon script load and sadly didn't tell me where it hurts.

Now, after updating the files QTGMC+ requires 3.7.3 but there is still 3.7.2 the latest version on avs-plus.net :confused:

And if I'm just at it: which version of NNEDI3CL/EEDI3CL should I use and (where to get)?

Also which version of mvtools? Still the kinda old 2.7.45? What about DTL's latest versions? I tried those, but couldn't get them to work with SMDegrain.

I kinda lost overview with all those updated files, I know. :o

Yes, there were certainly a LOT of updates...

I also haven't had any success with DTL's mvtools builds :(

https://gitlab.com/uvz/AviSynthPlus-Builds

https://github.com/Asd-g/AviSynthPlus-NNEDI3CL/releases/

[kedautinh12]

With NNEDI3CL 1.0.6 need bin from here to work
https://github.com/Asd-g/AviSynthPlus-NNEDI3CL/blob/main/src/nnedi3_weights.bin

[LeXXuz]

Yes, there were certainly a LOT of updates...

I also haven't had any success with DTL's mvtools builds :(

https://gitlab.com/uvz/AviSynthPlus-Builds

https://github.com/Asd-g/AviSynthPlus-NNEDI3CL/releases/

Thanks TDS. Up and running. :)

However, I still can't get subpixel=4, pel=4 to work with gpuid=0 in SMDegrain. Still that "nnedi3wrap: Field must be 0 or 1 for scaling" error in ResizersPack, line 678. Again, subpixel 1,2,3 work fine.

Updated SMDegrain and ResizersPack to latest version (and about a dozen other script files from Dogways repo :D)

[kedautinh12]

Thanks TDS. Up and running. :)

However, I still can't get subpixel=4, pel=4 to work with gpuid=0 in SMDegrain. Still that "nnedi3wrap: Field must be 0 or 1 for scaling" error in ResizersPack, line 678. Again, subpixel 1,2,3 work fine.

Updated SMDegrain and ResizersPack to latest version (and about a dozen other script files from Dogways repo :D)

Can you share full scripts??

[kedautinh12]

Thanks TDS. Up and running. :)

However, I still can't get subpixel=4, pel=4 to work with gpuid=0 in SMDegrain. Still that "nnedi3wrap: Field must be 0 or 1 for scaling" error in ResizersPack, line 678. Again, subpixel 1,2,3 work fine.

Updated SMDegrain and ResizersPack to latest version (and about a dozen other script files from Dogways repo :D)

just add propSet("_FieldBased",0) before SMDegrain, and it's work well. It's same error when use Deep_Resize after Deinterlaced with TIVTC, you need add propSet("_FieldBased",0) between Deep_Resize and Deinterlaced filter to work well. I know you don't rep me cause you think me noob:p

[Dogway]

I'm adding a fix for nnedi3wrap() and SMDegrain(), uploading in a few minutes, but for Deep_Resize() TIVTC should be updated with frameprops, not my fault that progressive is still tagged as interlaced.

[LeXXuz]

I know you don't rep me cause you think me noob:p

I never said nor think that! :confused: I appreciate your readiness to help others. ;)

I'm adding a fix for nnedi3wrap() and SMDegrain(), uploading in a few minutes, but for Deep_Resize() TIVTC should be updated with frameprops, not my fault that progressive is still tagged as interlaced.

Thanks Dogway for the quick response. And please never ever think someone here is blaming you for anything. Quite the contrary, your tireless efforts with all your scripts are very much appreciated and a tremendous help! :thanks:

[Dogway]

Just updated the remaining ResizersPack. Hope all issues are ironed out now.

[LeXXuz]

Just updated the remaining ResizersPack. Hope all issues are ironed out now.

https://abload.de/img/screenshot2023-01-181tzf0a.png (https://abload.de/image.php?img=screenshot2023-01-181tzf0a.png)

Please don't shoot me. :D