Originally Posted by stevevid
Thanks! My head is starting to align with my body, but I'm still partially walking backwards.
In the script, you used pp=0 to skip looking for combing artifacts instead of the default pp=6. What are the cues to determining which approach to use?
The decimate function is set to remove 10 out of 20. Is this related to the 2:3 pulldown pattern? Although this seems to be assuming a 1:2 pattern. What determines the M & N settings?
Thanks for your knowledge and patience,
Those film captures were made from film shot at 16 or 17 fps, or something like that. You don't use 3:2 pulldown on that speed. I first ran a script with only TFM enabled, and after looking at the remaining frame flow it was evident that some duplicate frames as well as "interlaced" pulldown sequences were used. After running TFM and looking at the results, there was an irregular pattern remaining which broke down to 10 duplicates in every 20 frames. You determine the M in N pattern by looking at frames and counting in VirtualDub
. If I recall, I made a spreadsheet with about 200 frame numbers and counted those marked as dupes. It turned out to be 10 dupes every 20 frames, sometimes 2 dupes in a row, a few times 3 dupes in a row (it was not a regular 1-2 pattern. I left it up to TDecimate to find the 10 dupes in every group of 20).
I believe the final flm speed came out to 17fps. In Avisynth I used AssumeFPS() to increase the speed to 18.98 fps and applied pulldown with DGPulldown
, which would have been some wildly unusual pulldown sequence that DGPulldown "knows" how to calculate) to make the film play at 29.97fps for DVD. 19.98 fps (19000/1001) is the slowest speed that DGPulldown can work with accurately.