vfr & tdecimate

[juhu]

I'd like to generate a .mkv vfr file (I never did) for a 85years old silent flick which appears to have :
-some sections @ 18 fps (+7 dupe frames added for PAL authoring)
-and others @ 20 fps (+5 dupe frames added to match Pal as well)

with tdecimate defaut mode, it's rather easy to target the dupes if I tell him what to do (cycler=5,cycle=25 for 20 fps / cycler=7,cycle=25 for 18 fps)

Now I admit I'm a bit lost with all those alternative modes generating stat files, especially this 2 pass mode to end up with the famous & mandatory timecodes.txt I'll need to make my mkv

(after 1st pass using the method described here :
http://avisynth.org/VariableFrameRateVideo
I end up with EMPTY tfm.txt / stat.txt files...)

If there are tdecimate pros here, I would appreciate to be taken by hand to know how to properly use it for this case...
thanks

[foxyshadis]

http://avisynth.org/mediawiki/VFR is updated; that page is old.

Did you play the video after opening it? (Or use video analysis pass, which plays it as fast as possible.) I don't think tdec can combine hybrid and anything other than cycler=1, though. If you're manually splitting the sections, though, you could just as easily manually create a timecodes v1 file with the same frame ranges.

(script)
t18 = tdecimate(cycler=7,cycle=25)
t20 = tdecimate(cycler=5,cycle=25)
t18.trim(0,4673) + t20.trim(4674,8924) + ....

(timecodes)
Assume 18
4674,8924,20

etc.

[juhu]

thanks for the tips. Actually I didn't even know the structure of a timecode file. But as it seems it's nothing more than a text file with first-frame/last-frame/fps, yeah, I started to go for the manual method (21 parts, tedious but doable...)

one last question : how many digits are supported for fps in timecode files ? I was a bit optimistic yesterday, it wasn't 7 dupes in 25 frames, but 7 dupes in 27, ie 18.518518518 etc fps