Silent source decimation

[nixo]

Hi,
I have little experience with this type of source and I'd be grateful if someone could take a quick look.
As far as I know the original source *should* be 18fps but 20fps is certainly a possibility. Typically, I would use TFM/TDecimate but I can't figure out a consistent pattern.

http://www53.zippyshare.com/v/7356479/file.html

--
Nikolaj

[poisondeathray]

For a silent film, why not use a duplicate frame remover? You don't have to worry about sync. e.g. multidecimate , dedup (in non vfr mode, without timecodes), or similar filters, then AssumeFPS() for any framerate

But another problem looks like there are dropped frames. You could use mvtools2 to interpolate and insert , but it would need manual identification

[nixo]

For a silent film, why not use a duplicate frame remover? You don't have to worry about sync. e.g. multidecimate , dedup (in non vfr mode, without timecodes), or similar filters, then AssumeFPS() for any framerate


That's an interesting idea. I was being a little unclear, though; the source does have a music soundtrack and I think I'll try and keep it.
I think I might try running it through direct264 and just live with the results.


But another problem looks like there are dropped frames. You could use mvtools2 to interpolate and insert , but it would need manual identification

Yeah, there are plenty of missing frames. It's a Criterion BD, so I'd assume a better source simply does not exit. Going through the whole thing manually is a little too much work just to make a tablet-friendly backup, I think.
Thanks for taking a look.

[poisondeathray]

Theoretically, you should be able to keep it in sync with the soundtrack if the duplicates were inserted from the process of making it 29.97. You can calculate the framerate from the ratio of (remaining frames / original frames) * original frame rate) . In this small clip , it works out to be 18.03 =~ 142/236 * 30000/1001, but i bet on a longer sequence it will average out to 18

The accuracy should be high for duplicate removers - because these are almost identical duplicates (the difference in duplicates is in the compression noise). If you look at the log file, the difference between duplicate, and non duplicate is large (lots of headroom), so the threshold settings are easy to tweak

[nixo]

Oh, that's pretty cool! I'll take a look at the various filters.

--
Nikolaj

[johnmeyer]

It sure is a complicated pulldown pattern. It (more or less) repeats every 60 fields, with 24 fields needing to be decimated. This standard IVTC script seems to do the right decimation, although I haven't tested it thoroughly:

loadPlugin("c:\Program Files\AviSynth 2.5\plugins\TIVTC.dll")

AVISource("e:\fs.avi").killaudio()

AssumeTFF()
tfm(display=false)
tdecimate(display=false,mode=0,cycleR=24,cycle=60)
AssumeFPS(12)
#separatefields()


There are definitely some jumps, such as at the beginning of the second scene where she walks towards the cage. In one frame there is a big gap between her body and the plate on the wall, and in the next frame she has moved forward by such a large amount that her body completely obscures the plate. This jump is in the original video, and is not a function of poor decimation by the script above.

So, if I were doing this, I'd do the decimation with the above script, and then use one of several scripts that have been posted in this forum which attempt to look for missing frames, and then interpolate between those frames. As already pointed out, since this is silent film, it won't matter if you interpolate a few too many frames because you can simply adjust speed later on. I chose, in the script above, to set playback at 12 fps, since that was common in a lot of early silent, hand-cranked silent films.

BTW, the jumps may be due to damaged frames which may have simply been decimated by the people who transferred the film.

[nixo]

Thanks, that looks good and seemingly takes care of dupes all the way through.

--
Nikolaj