Advice need. Append video in AviSynth.

[Vitaliy Gorbatenko]

code:
#
SetMemoryMax(1024)
#
# PLUGINS
SetWorkingDir("C:\system\Knot\AviSynthPlugins\")

import("!Allinclude.avsi")
import("srestore2.avsi")

# SOURCE
mpeg2source("H:\Encoding\test.d2v", cpu2="oooooo", info=3)
a=colormatrix(interlaced=true, hints=true)
pa=a.trim(0, 28053).animeivtc(mode=1,aa=0)
pb=a.trim(28054, 28110).TempGaussMC_beta1u().srestore2(frate=23.976)

return (pa+pb)#.crop(8,0,-6,0).Spline64Resize(880,496)

pb is added correctly to pa


#
SetMemoryMax(1024)
#
# PLUGINS
SetWorkingDir("C:\system\Knot\AviSynthPlugins\")

import("!Allinclude.avsi")
import("srestore2.avsi")
# SOURCE
mpeg2source("H:\Encoding\test.d2v", cpu2="oooooo", info=3)
a=colormatrix(interlaced=true, hints=true)
pa=a.trim(0, 28053).animeivtc(mode=1,aa=0)
pb=a.trim(28054, 28110).TempGaussMC_beta1u().srestore2(frate=23.976)
pc=a.trim(28111, 0).animeivtc(mode=1,aa=0)
return (pa+pb+pc)#.crop(8,0,-6,0).Spline64Resize(880,496)

pb is absent in video :confused:
Why it is impossible to add a piece (pb)?

[Vitaliy Gorbatenko]

SRestore2 is same as SRestore, but without memory leak.

[MadRat]

Normally when you append videos you do it like this:


video1 = mpeg2source("filenameA.ext")
video2 = mpeg2source("filenameB.ext")
video3 = mpeg2source("filenameC.ext")

finalvideo = video1 ++ video2 ++ video3


I'm doing this off the top of my head so I hope I don't have it wrong.

[Vitaliy Gorbatenko]

That's just the point! I cannot add clips in to one. The second is lost. return(a+b) work! return(a+b+c) don't work.

If not to use difficult functions - that so works:

pa=a.trim(0, 28053)
pb=a.trim(28054, 28110).Invert()
pc=a.trim(28111, 0)

return (pa+pb+pc)

syntax error - Does not appear and on an exit I have the necessary number of frames..

[Gavino]

My memory leak fix has now been incorporated in the official SRestore (http://avisynth.org/mediawiki/upload/7/7d/Srestore.avsi) version. I see that my srestore2 has been renamed srestore_inside to avoid confusion. (Incidentally, it doesn't avoid the use of ScriptClip, it just makes the ScriptClip string very much smaller.)

Vitaliy, can you confirm you have a self-consistent srestore function (perhaps renamed by you to srestore2?).
When you say pb is missing, what do you mean exactly? Does the total framecount match only pa+pc?

[Vitaliy Gorbatenko]

Sorry for late answer... Yes, with official version the same situation.
pb missing - i'm see in stream pa and pc part.

And it's not FPS problemm:
f_int=colormatrix(interlaced=true, hints=true)
a=f_int.trim(0, 1177).animeivtc(mode=1,aa=0).AssumeFPS(23.976)
b=f_int.trim(1178, 1243).TempGaussMC_beta1u().srestore(frate=23.976).AssumeFPS(23.976)
c=f_int.trim(1244, 0).animeivtc(mode=1,aa=0).AssumeFPS(23.976)
a+b+c

don't work too.

[Gavino]

Perhaps the problem is with animeivtc - it may be that it doesn't work if called more than once in a script.

[thetoof]

1 - why don't you use something like this to avoid unnecessary multiple calls of the same function (for a starter, it wastes less memory)?:
ivtced = f_int.animeivtc(mode=1,aa=0)
bobbed = f_int.TempGaussMC_beta1u().srestore(frate=23.976)

ivtced.trim(blah)+bobbed.trim(blah)+ivtced.trim(blah)

2 - why do you need to do bob+use blend-aware decimation on a section and not the others?

[Vitaliy Gorbatenko]

1. I'm not very smart enough... In this chain of filters is difficult to determine from what and at what frame is cut, so wanted to cut the source, and then perform deinterlacing.

2. video source has mastering errors (or codec bug...) most 3:2 and some parts pure interlaced..

[thetoof]

Can you send a small sample (5-10mb) of the unprocessed source? I'd like to check those interlaced sections.

[Vitaliy Gorbatenko]

Short sample here : http://www.mediafire.com/?micxtojzoz0
Frames 454-498 are not normal.

[thetoof]

heh, I see what you mean. Looks like they added the "shaking" after applying 3:2 pulldown, creating 30i motion.

-I wouldn't recommend going to 24fps, as you have 60fps (after bobbing) worth of motion
-Creating a VFR clip would be the way to go imo.
I'm having a problem detecting the 30p sections with tdecimate because of the peculiar way it was done, so I'll have to ask in tivtc's thread.

edit - I created a new cfr mode for you. http://www.sendspace.com/file/nt1n7z Since it won't really matter and won't be noticed that frames are dropped since the motion is only "frame shaking", try this :
animeivtc(mode=1,e1=454,i1=503,e2=628,omode=1,credconv="drop")

[Vitaliy Gorbatenko]

At a lot better! Could you make these changes on an ongoing basis. Because these errors occur often enough though, but trouble with them very much.

[thetoof]

I don't get what you want to say

Which changes, which errors?

[Vitaliy Gorbatenko]

Sorry for my bad english... You created a new cfr mode for me.. I want this mode remains in FUTURE versions .. Since these moments interlacing from time to time appear in the source material ...

[thetoof]

Yes, sure. As you might have noticed, I sent you the "dev copy", meaning it is the script I'm currently working on. There are a few things to test and correct for vfr and once it's done, the changes and corrections (for both cfr and vfr) will be uploaded as 2.01.