[MPlayer-users] Problems encoding interlaced DVD

[D Richard Felker III]

On Sat, Sep 06, 2003 at 12:26:49PM -0700, Corey Hickey wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> Chris Phillips wrote:
> >[Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> 
> >Ripping the DVD works fine (using vobcopy).
> 
> Out of curiosity, is there an advantage to using vobcopy rather
> than mplayer dvd://1 -dumpstream   ?

AFAIK, not really. Maybe if you plan to rip the vobsubs, though.

> >Simple
> >playing with mplayer shows bad interlacing, but the 
> >-vf pullup option fixed the problem OK. 
> >
> 
> If pullup fixed the problem, that probably means the interlacing
> was telecine. However, you might want to verify that. Play the video
> with -fps 2 (or -fps 1) and see if progressive and interlaced frames
> follow each other in this pattern:
> 
> p,p,p,i,i,p,p,p,i,i
> 
> That's the usual pattern for 3:2 pulldown, which can be removed easily.
> However, I've seen some weird (usually old) movies that follow a
> different pattern, and sometimes there are movies where every frames is
> interlaced. In these cases, you may be stuck with using the
> deinterlacing filter of your choice, such as -vf pp=lb, pp=md, pp=fd,
> etc.

No, as long as it's *any* kind of pulldown (<= 30fps converted to 60
fields/sec or <=25fps converted to 50 fields/sec), pullup is the
correct filter to use and should give you (essentially) lossless
non-interlaced frames. Deinterlace filters should *only* be used when
the original content is filmed with a TV camera (50/60 field/sec
interlaced video), since they throw away lots of information and
create artifacts that will waste bits when encoding...and in that case
you're probably better to just keep the interlacing and use -vf pp=lb
at playback time.

> You said, however, that playing the movie with pullup worked ok, so
> perhaps encoding with pullup would work. I haven't tried pullup since
> the first iteration, and Rich made some changes since then. By now it
> might work quite well, and I'm sure it will before too long.

The basic algorithm is pretty much done, although fine-tuning options
will be added later. The only problem is that the mplayer-g1 wrapper
for pullup is a bit incomplete, and g1's design limitations prevent it
from working as well as it would work in g2.

> I never use 3-pass encoding, so I can't say for sure, but this
> might be an effect of the first (audio) pass expecting more frames
> per second of video. When you filter the video, this reduces the number
> of frames, and makes the audio out sync. For now, try encoding with only
> one pass (to keep it simple), and once you get things worked out, then
> try going back to 3-pass. 3-pass might work properly if you use the ivtc
> filter for the first pass, too.

You cannot use 3pass with any sort of inverse-telecine filter (or any
filter that adds/drops frames). Perhaps this should be made clear in
the fine manual...

Rich