ChromaReconstructor

[feisty2]

well, I just like to design some filter with avisynth and then port it to vaporsynth rather than writing it directly with vaporsynth cuz, uhh, avisynth is still more user friendly (like avspmod, extremely simple home grown scripting language...) despite all the bad stuff it couldn't get rid of.

back to the point, this thing could be used to reconstruct 420/422 to 444, more details see http://forum.doom9.org/showthread.php?t=173005

YUV444PS (full range) input is a must!!!

the quality is slightly better than avisynth version for the enhanced precision (single precision floating point vs hybrid 8bits/16bits integer)

https://github.com/IFeelBloated/VaporPlacebo/blob/master/Placebo.py (named "fulltonative")

[bebs]

Hi,

I'd like to use fulltonative in a script where i also use QTGMC.
Problem is that RGVS.dll (for core.rgsf.Repair) conflict with the default RemoveGrainVS.dll in Vapoursynth r29 as both register with com.vapoursynth.removegrainvs
Is this possible that RVGS.dll use another name as com.vapoursynth.removegrainvs ?

[feisty2]

sure, I'll change it soon as I got time :)

[dontrythisathome]

I tried your script but it is extremely heavy, I've got an i7-4770 but no dedicated GPU ( only the internal crappy Intel HD Graphics ) and it always takes up to 99% of CPU work...

I wish to see what this filter does but It's too heavy for me...

This is my script



import vapoursynth as vs
import Placebo as po

core = vs.get_core()

clip = core.ffms2.Source( '/home/aculnaig/Video/litapavc.m4v' )

clip = core.fmtc.resample( clip, css = "444" )
clip = core.fmtc.bitdepth( clip, bits = 16 )

clip = po.fulltonative( clip )

clip.set_output()



The video file is an h264 10-bit yuv420p 1280 x 720 100 frames.

I've already tried to downscale the video size but it takes always the same amount of CPU work

What can I do to lighen this filter or do I have to buy a decent GPU nnecessarily ?

[feisty2]

the script should be

import vapoursynth as vs
import Placebo as po

core = vs.get_core()

clip = core.ffms2.Source( '/home/aculnaig/Video/litapavc.m4v' )

clip = core.fmtc.bitdepth( clip, bits = 32 )
clip = core.fmtc.resample( clip, css = "444" )

clip = po.fulltonative( clip )

clip.set_output()


try reducing "a" to 16 and do it on a standard definition clip...
if that still wouldn't work, I guess there's nothing more you could do, I initially wrote this thing for some kind of experiment, not for regular video processing out there in the real world
sorry about that

[dontrythisathome]

Still no luck...even with a simple JPEG file...

I must give up...

Thanks anyway...