[MPlayer-users] Re: 60p -> 24p conversion for ATSC

[pfile at pacbell.net]

> The few times I noticed it occuring (during frame by frame
> examination) was during the "Presented in HD" overlays used at the
> beginning of a show, the animation for the overlay is never on the
> same timing as the video (as though the video were driven to 60p then
> the overlay added afterwards with no though as to the video), which
> ends up screwing up the filter.

i have noticed this too. what eventually ends up coming over the air 
has been transcoded up the wazoo and overlayed at 60fps, etc. etc.

>
> I've also noticed that the WB does some real strange frame timings
> that require manual intervention. This doesn't occur for all WB
> stations, but for the one I get OTA it does. The frame sequence will
> change with every repeat so no amount of settings seem to catch each
> change, especially when the length of the sequence changes...
>

i've never looked at WB, but i'm not surprised based on what i know 
about D*. D* really has some aggressive mpeg. for 29.97 content they 
usually only broadcast 2 fields for every 3, and use the RFF flag. also 
they keep alternating between BFF and TFF. this seems to confuse the 
mplayer/mencoder mpeg demuxer when the content was telecined 24fps - it 
keeps switching framerates. however, i've found that just forcing 
mencoder to treat the framerate as 30000/1001 and then using pullup, 
the right thing seems to happen. having said that, the nice thing is 
that on some of the movie channels, they send straight "progressive" 
interlaced 24fps material just like a DVD.

so following up again, i happened to try the tinterlace,pullup,softskip 
chain on the finale of AD. this was one huge file with 4 eps in it; 
most of the file was fine but there were a few places where pullup had 
turfed the wrong frames. this is on a pre-edited TS file, so its 
possible that something got hosed up due to native 60fps commercials. 
also i am not necessarily running the latest version of mencoder (i 
think i'm on pre7) so i'll try with CVS and see if that's any better. i 
guess the tinterlace thing is a hack anyway so we're outside of the 
norm for what pullup is supposed to handle?

as an alternative to pullup, i tried the decimate parameters you gave 
and something really went wrong. pretty much the whole file had the 
wrong frames skipped, and based on the log, decimate was not dropping 
anything. was it supposed to be 2:1000:1600:.001? or should min and max 
be swapped? i just noticed that my script was calling a pretty downrev 
version of mencoder, so i'll try again now with something more recent.

rob