Telecide(guide=?)

[tripex]

Hi

I normally come from the PAL side and don't have to deal with it but assuming time does NOT matter what is the safest way to deal with FILM material which is not close enough to 100% FILM? I don't want to process it with DGIndex b/c I still spotted comb effects in the final avi although it was 99.??% FILM. So in my understanding I have to handle it with Telecide() even it is slower. But why? Further I have the option of switching the guide on or off. So first what is the difference between DGIndex's FILM process and Telecide(guide=1) and second what is the difference between guide=1 and guide=0? Obviously it's the guide but what does Telecide do w/o the guide exactly? Is there ANY disadvantage regarding safety? As I said, I don't really care about speed, I rather just don't trust guidance b/c of my negative experience with DGIndex.

CU
Andy

[buzzqw]

another way is to use the TFM(d2v="your_film.d2v").Tdecimate()
it is great of film source

for telecide ... i suggest to re read the manual, Neuron2 has done a great work with this filter and documentation!

BHH

[neuron2]

I normally come from the PAL side and don't have to deal with it but assuming time does NOT matter what is the safest way to deal with FILM material which is not close enough to 100% FILM? It depends too much on your source to give a definitive answer. If there are not too many combs left after doing Force Film in DGIndex, you could just do it anyway and follow it with FieldDeinterlace(full=fall), which will clean up just the combs that sneak through. If it's really bad, then you'll want to use a good field matcher such as Telecide() or TFM(), followed by decimation, i.e., external IVTC.
I don't want to process it with DGIndex b/c I still spotted comb effects in the final avi although it was 99.??% FILM. So in my understanding I have to handle it with Telecide() even it is slower. See above.
But why? Suppose you took an interlaced clip and applied 3:2 soft pulldown. Then you undo the pulldown (perfectly) with DGIndex Force Film. The result would, not surprisingly, be combed. So, all it means is that the source was not fully progressive. That can come from video portions mixed with film, but also from orphaned fields at bad edit cuts.
Further I have the option of switching the guide on or off. So first what is the difference between DGIndex's FILM process and Telecide(guide=1) DGIndex's film process works (essentially) by just ignoring the soft pulldown flags. Telecide() actively tries to match up the fields. The field matching process is heuristic and can be helped if you know what the pattern should be, i.e., 3:2 pulldown.
and second what is the difference between guide=1 and guide=0? Obviously it's the guide but what does Telecide do w/o the guide exactly? See above and read this (the 6-25-2003 entry):

http://neuron2.net/journal/journal2003.html

Is there ANY disadvantage regarding safety? As I said, I don't really care about speed, I rather just don't trust guidance b/c of my negative experience with DGIndex. Telecide() guidance and DGIndex's Force Film are totally different things and shouldn't be compared in that way. Generally guidance helps if you know it's regular 3:2 pulldown.

[tripex]

Hi neuron2

Thanks for the infos! I see more clear now. Since I don't use DGIndex's simple FILM processing whenever the material is below 99.9% (I could sustain a few seconds but not 10) I use Telecide anyway in many cases.
As I understand guide=1 has the focus to "track the positions" of the known 3:2 pattern whereas guide=0 does not rely on any patterns and tries as smart as possible to match the fields regardless a pattern at all.

I used to change Telecide(order=1,guide=1).Decimate() to Telecide(order=1,guide=0).Decimate()

It takes much longer but I always thought it's the perfect way and I still understand it as the perfect way even for slow motion movies, because even it might match wrong you can't see it because it can happen only at static scenes so it doesn't matter, at most theoretically, right?

In other words or as a question: Assuming we have the worst clip (but mainly regular FILM); can the result with guide=0 be practically (means visible for good human eyes, no still) worse compared with the result produced by guide=1?
If so how I can easily prevent it and pick the proper choice every time?

CU
Andy

[neuron2]

I have test clips that match better with pattern guidance so I know it can be useful. But there isn't going to be a universal prescription. You're going to have to judge matters yourself for each specific stream.

[tripex]

You're going to have to judge matters yourself for each specific stream.

I hope you understand, that I don't want to spend that much time. So in your opinion as the Telecide programmer, is it better (safer) for the most FILM material to leave the guide on or switch it off? I have never seen a visible difference in the results anyway but I'd "feel better" to know doing it right for the majority of cases.

CU
Andy

[neuron2]

For pure soft 3:2 telecine of progressive material I would use DGIndex Force Film together with FieldDeinterlace(full=false) to catch bad edits.

For pure hard 3:2 telecine of progressive material I would use pattern guidance.

I can't comment on what works best in the "majority of cases" because I seldom do encodes of anything more than 10-second sports clips for didactic purposes.

[TheRyuu]

another way is to use the TFM(d2v="your_film.d2v").Tdecimate()
it is great of film source

for telecide ... i suggest to re read the manual, Neuron2 has done a great work with this filter and documentation!

BHH

I'd recommend this route too. I've had nothing but good results using it that way (unless it's soft telecined, then you can just use dgindex force film).
tfm and tdecimate are part of TIVTC (found here. (http://bengal.missouri.edu/~kes25c/), documents on it found here. (http://avisynth.org.ru/docs/english/externalfilters/tivtc.htm))

Although the ultimate decision on what filter you use will be up to you.