Selur's Little Message Board

Hello Selur,

When I'm using Avisynth filter and I select the following filters

Deen1b1_Deen
FFT3DGPU
Deblock
LSFmod
aWarpSharp
FineDehalo
FastLineDarkenMOD4
Spline36Resize

and then I select Filtering->Misc->Custom Filter Order

I get the following error: Unknown filter 'aSharp' in custom filter order!

I attached the script generated and the Debug file.

Thanks,
Dan

Ah, that's a typo, which I think I already fixed in my dev version.
Will build a new dev package and send you a link in ~15min.

Cu Selur

Send you a link.

Cu Selur

Hello Selur,

I installed the dev version, and now the problem has been fixed.

I'm using for testing the following file

https://file.io/7daaniEpSD7g

and I attached the script used.

On this file the quality of up-scaling is comparable with the one obtained using A.I upscaler like RealESRGAN-anime, but is much faster.

I have 2 questions:

1) Are available GPU versions of AVS filter used in the attached script (at least for some of them) ?
2) It is possible to write an equivalent Vapoursynth script ?

Thanks,
Dan

(27.12.2022, 12:07)Selur Wrote: [ -> ]Ah, that's a typo, which I think I already fixed in my dev version.
Will build a new dev package and send you a link in ~15min.

Cu Selur

Hi, Selur.
Could you please send it to me too? I'm facing the same problem described by the OP. Sad

Quote:1) Are available GPU versions of AVS filter used in the attached script (at least for some of them) ?

Looking at your script:

Code:
LWLibavVideoSource("E:\VideoTest\Anime_Upscale\ToonResize\scooby_doo.e.i.mostri.del.cinema.2013.italian.dvdrip.mkv",cache=false,dr=true,format="YUV420P8", prefer_hw=0,repeat=true)

# current resolution: 720x544

# filtering

Deen1b1_Deen(mode="a3d")

FFT3DGPU(ow=16,oh=16,mode=1,precision=0)

Deblock(planes="yuv")

LSFmod(Smode=5,Smethod=3,Lmode=4,soft=-2,edgemaskHQ=true)

aWarpSharp()

FineDehalo()

SuperToon()

# scaling to 1280x968

Spline36Resize(1280,968)

# Converting from 8 to 10bit for encoder

ConvertBits(10)

# setting output fps to 25.000fps

AssumeFPS(25,1)

There are no GPU filters.
Deen was never ported to Vapoursynth, since KNLMeansCL and other filters do a better job.
FFT3DGPU is Avisynth only
There is no gpu version of Deblock, or aWarpSharp.
FineDehalo is a script which uses Masktools and Removegrain, both are cpu only.

Quote:2) It is possible to write an equivalent Vapoursynth script ?

Probably not, but one can probably do better. You just would have to look into it.
I would probably start with:

MCDegrainSharp to remove the grain and sharpen it a bit,
KNLMeansCL on Y only and distance 0 with a relatively high value to remove the main luma noise from flat areas.
GLSL AdaptiveSharpen to sharpen the lines

so something like:

Code:
# removing grain using MCDegrain

clip = mcdegrainsharp.mcdegrainsharp(clip=clip, csharp=0.50)

# denoising using KNLMeansCL

clip = core.knlm.KNLMeansCL(clip=clip, d=0, h=5.00, channels="Y")

# adjusting color space from YUV420P8 to YUV444P16 for vsGLSLAdaptiveSharpen

clip = core.resize.Bicubic(clip=clip, format=vs.YUV444P16, range_s="limited")

with open("i:/Hybrid/64bit/vsfilters/GLSL/parameterized/adaptive-sharpen.glsl") as glslf:

  glsl = glslf.read()

glsl = glsl.replace('#define curve_height    1.0', '#define curve_height    1.5000')

glsl = glsl.replace('#define anime_mode      false', '#define anime_mode    true')

glsl = glsl.replace('#define overshoot_ctrl  false', '#define overshoot_ctrl  false')

glsl = glsl.replace('#define video_level_out false', '#define video_level_out true')

clip = core.placebo.Shader(clip=clip, shader_s=glsl, width=clip.width, height=clip.height)

If you like the line thinning of aWarpSharp, I would alternatively try 'Thin (GLSL-A4K)' and instead of SuperToon for the line darkening I would try 'Darken (GLSL-A4K), since GLSL filters are usually really fast.

Cu Selur

Dan64

Selur

Selur

Dan64

Nero

Selur