+ Reply to Thread
Results 31 to 38 of 38
but computational complexity is lowest or higher using yadifmod/nnedi3 instead of yadif (X86)?
NNEDI3 is the "bottleneck" by far , whether x64 or x86 (unless there is memory bottleneck for x86)
There are some GPU variants that can speed it up , NNEDI3 OCL (open CL), and optimized CPU znnedi3 (for vapoursynth), and NNEDI3CL (vapoursynth CL version) but overall it's still very slow
(If you want fast, QTGMC, CUDA variant, KTGMC can run 5-10x faster than CPU QTGMC if you have a fast Nvidia GPU ; this can be must faster than yadifmod2+nnedi3 ; but it requires a separate cuda avisynth setup)
Last edited by poisondeathray; 15th Jan 2020 at 17:02.
I have do some proof: NNEDI3 is too heavy to keep the playback in sync my NLE (it's a NLE so high of level that I don't say the name so not to advertise
yadif seems the best quality but I have this strange problem in some clip with details (is a mystery of how strangely possible is this)
an alternalitve isCode:
I wonder yadifmod does do a deinterlace without without recourse to an external plugin like NNEDI? or is there another deinterlacer that is not so heavy like NNEDI that do a deinterlace of yadif quality?
LWLibavVideoSource("C:\Users\Administrator\Desktop\Nuova cartella (5)\canon422 XF300 1920x1080 50i 422.MXF") yadifmod(order=1,mode=1,edeint=nnedi3(field=3))
LWLibavVideoSource("C:\Users\Administrator\Desktop\Nuova cartella (5)\canon422 XF300 1920x1080 50i 422.MXF").yadifmod(order=1,mode=1, LWLibavVideoSource("C:\Users\Administrator\Desktop\Nuova cartella (5)\canon422 XF300 1920x1080 50i 422.MXF").edeint=nnedi3(field=3))
yadif is known for these - deinterlacing artifacts, or "marching ants" . The quality really isn't that good. It's fast, that's it. All fast deinterlacers will have those types of artifacts. You have to decide where you want to make tradeoffs . In general, you cannot have high quality and speed ( unless it's GPU KTGMC, with a fast Nvidia GPU)
If you omit edeint; ie. Yadifmod(order=1, mode=1) it is the same as yadif . Same quality. Same speed. Same artifacts. The external NNEDI3 is for fixing those artifact areas. It's used for interpolation instead of yadif's own algorithm . You can use other filters for edeint , but some are even slower like EEDI3 .
Or use a physical file instead of "live" avfs . There is additional overhead with avfs too
I try use
Import("v:\automazioneclip\avisynth\plugins\IResize.avsi") LoadPlugin("v:\automazioneclip\avisynth\plugins\LSMASHSource.dll") LoadCPlugin("V:\automazioneclip\AviSynth\plugins+\LeakKernelDeint.dll") LoadCPlugin("V:\automazioneclip\AviSynth\plugins+\yadifmod\yadifmod.dll") LoadCPlugin("V:\automazioneclip\AviSynth\plugins+\NNEDI32019\nnedi3.dll") LWLibavVideoSource("C:\Users\Administrator\Desktop\c116\C0116.MXF") Yadifmod(order=1, mode=1)
yadifmod: args 0 and 4 must be clips!
what can it be?
something like the catfunction can be help?
Sorry, that's for yadifmod2 , not yadifmod . original yadifmod requires edeint specified
yadifmod2(order=1, mode=1) is same as yadif(order=1, mode=1)