PAL to NTSC with avisynth - blended frames

[nesburf]

Hi all,

Trying to convert a PAL dvd to NTSC using this script.

SetMemoryMax(512)
LoadPlugin("C:\Program Files (x86)\AviSynth\plugins\DGDecode.dll")

setmtmode(6,8)
mpeg2source("VTS_02_1.d2v")
setmtmode(2)
assumetff()
QTGMC(preset = "slow", EdiThreads=4)
super = MSuper(pel=2)
backward_vec = MAnalyse(super, isb = true,overlap=4, blksize=16,search=3,searchparam=4)
forward_vec = MAnalyse(super, isb = false,overlap=4, blksize=16,search=3,searchparam=4)
backward_vec = super.MRecalculate(backward_vec, overlap=2, blksize=8,search=3,searchparam=4)
forward_vec = super.MRecalculate(forward_vec, overlap=2, blksize=8,search=3,searchparam=4)
MFlowFps(super, backward_vec, forward_vec, num=60000, den=1001, ml=100)#59.94p
lanczosresize(720,480)
separatefields().selectevery(4,0,3).weave()
converttoyuy2(interlaced=true)

The problem I'm having is that at every shot change there seems to be 1 or 2 blended frames of the 2 separate shots. Is there anyway to tweak this script that may fix the issue?

Thanks

[manono]

The best thing you can do is to provide a 10-second sample so we can have a look, one with steady movement. You can cut it using DGIndex.

[nesburf]

Hmmm after looking at it again it may just be a crappy source. I did a double rate deinterlace (QTGMC) on the original source and there seems to be some frame blending on shot changes. So that just translates over when I do my conversion.

[jagabo]

Your source may be an NTSC to PAL conversion with blended fields. The blending can be mostly eliminated with SRestore().

[nesburf]

Hmm after looking over this some more, it almost seems like a straight telecide() seems to restore the frames. This gives me progressive 25fps video, but then I still need to get it to NTSC somehow. I'm guessing a slow down to 23.976 is in order?

I've attached a sample.

[jagabo]

Yes, it's just an out-of-phase PAL cap. You just need to get it back into phase with Telecide or TFM. Assuming you're making a DVD, you could soft telecine from 25 fps to 29.97 fps. Ie, encode at 25 fps with a progressive 720x480 frame, then use DgPulldown to add 3:2:3:2:2 pulldown flags. That will keep the original running time so you don't have to change the audio.

[manono]

Or, if it's a movie, you could easily slow it to 23.976fps with a simple AssumeFPS(23.976) at the end of the script and then slow the audio to match. That's what I always do for for film sources. If it was created at 25fps then, yes, I'd do as jagabo suggests.

Now do you understand why, when asking questions about how to treat a video source, you should always provide a sample without even being asked?

[nesburf]

Thanks Jagabo, that seems to produce the best output.

Manono - that's my bad, I should know better and should have provided a sample off the bat, my apologies.