On Thursday, 12 June 2008 at 23:54, Angus Lees wrote:
> My intention with the fmvs originally was to provide a separate 'fmv.wz'
> which could just be dropped into the right directory - as other's have
> suggested.  Look for the movie and fall back to current behaviour if not
> found (and when the movie ends).  Distros-and-whatever would presumably
> package it separately, and (when the legal situation was murky) users could
> choose to download/distribute it themselves.
> 
> The patch I sent some time ago supported both .ogg and .rpl formats, again
> my intention being allowing those few people with the original game to use
> their .rpls (or something) and the rest of us can use .oggs.  And the code
> is structured to allow other formats if something better turns up.

Since the reencoded movies are quite a bit smaller and can now
definitely be redistributed, the RPL code isn't really needed anymore, I
think.

For playback I'd prefer using FFMPEG to the OGG libraries directly,
since this will make it easier for others to make movies for mods
without lossy reencoding into obscure formats.

> For reference, a no-fancy-options recompression of the rpl files into ogg
> ends up at about this resulting size:
> 
>  187M    sequences_ogg.zip
>  777M    sequences_rpl.zip
> 
> So .ogg theora/vorbis is a pretty big saving in size.  sequences_ogg.zip was
> generated via my rpl2avi wine program with the original eidos dlls and then
> reencoded using ffmpeg2theora - if you're interested in the resulting file
> or any of the pipeline just ask.

What codec was the intermediate AVI? If it was lossy, the double
reencoding degraded the quality/size more than necessary (i.e. with
direct encoding to the target format the files could be smaller or
of better quality (or even both)).

> As for the original patch, it was deliberately against 2.0, since I figured
> that was where the unmodified game was (this may not be the intention
> anymore however).  I tried briefly to port it to svn head, but I don't think
> I finished the job.
> 
> I can easily throw what I've got into an svn branch and everyone can hack on
> it, although there seemed little interest from other developers last time..

I guess because of the uncertainty regarding the movies it never made it
into SVN and thus fell by the wayside.

> My code always had a strange opengl bug I could never track down: after
> playing an fmv, the closest LOD textures were corrupted. As far as I could
> work out I was correctly resetting the texture page and other obvious things
> - I figure I wasn't cleaning up sufficiently after the various YUV opengl
> shenanigans and I expect someone who actually understands GL would be able
> to spot it fairly easily.  The RPL dec130 decoder also always had some
> output corruption I could never work out.  The ogg/theora decoder works fine
> though.  Oh and I hadn't reverted enough of our hacks to be able to show the
> windowed research videos again, but that should be quite possible.

Is the YUV stuff really necessary? I don't really know the OGG libs, but
FFMPEG includes yuv2rgb conversion that could be used as a baseline, and
the OpenGL YUV converters as optional optimizations. That also has the
advantage of working always (if not that fast, but for the low-res FMVs
that shouldn't matter much) - IIRC none of the mplayer OpenGL YUV
variants works correctly on my system.

-- 
* dark has changed the topic on channel #debian to: Later tonight: After
  months of careful refrigeration, Debian 2.0 is finally cool enough to
  release.

_______________________________________________
Warzone-dev mailing list
[email protected]
https://mail.gna.org/listinfo/warzone-dev

Reply via email to