[MPlayer-dev-eng] Slideshow

[Daniel A. Nagy]

On Wed, Dec 11, 2002 at 02:37:28AM +0100, Herman Tamas wrote:

> some time it should b a legitim part of mplayer. but it requires
> more factorization/modularization of the source so 1 can implement
> such thing a magnitude faster. i'll b happy if can write code such as
> 
> begin eof? while
>   nextsub subpos@ >frameno seekmovie get-frame
>   decode-frame sub>osd frame2jpeg write-jpg
> repeat
> 
> & mplayer would understand it... (ofcoz not that strange naming
> convention is what im starving 4 ;)

Actually, the solution (the dirty hack, that is) I came up with first was
not THAT different from the above outlined concept:

First, I exported the subtitles in vobsub format (1), captured the frames at
timepoints extracted from the .idx file using -vo, -ss and -frames with
respective parameters (2), then hacked up TOOLS/subrip.c to rid it of the
character recognition code (3), and to save the appropriate subtitle as an
image+mask (4). Then I wrote a simple code that sticked the subtitles onto the
frames (5).

Now this is ugly for the following reasons:

1: What I really needed here were the timestamps. It'd have been nice to be
able to export only those.

2: When exporting a single frame, it seemed impossible to add the subtitle.
Would it have been possible to end the process at this stage?

3-4-5: This just stinks. All that information is available to mplayer when
playing the movie, thus I was effectively replicating already existing
mplayer functionality. Doing these steps, I actually understood, how mplayer
did it, and lifted quite a bit of code.

After doing this, I decided to implement the functionality myself (without
asking how to do it), and hacked up the jpeg driver to export only given
frames (while playing the whole movie). This way, the subtitles were in
place. But it's still very ugly. Hence my question to the list.

-- 
Daniel