[MPlayer-users] Greetings to all and a mencode question...

[J. J. Franzen]

D Richard Felker III wrote:

>-vfwopts codec=filename.dll
>
>Sorry it doesn't seem to be in the docs. BTW why are you using vfw for
>encoding?
>
Well, it's a long sorted tale...  I hope you've had some coffee recently...

     I'm one of the Sys Admins for the animated TV show, South Park. 
 Until recently, we were using Avid's Media Composer Version 7.x to edit 
the show.  We recently updated to Avid's Media Composer 10.x.  In doing 
so, we unknowingly broke a part of our production pipeline.  Currently, 
the way we get our animation onto the Avids is to send the rendered 
frames to a DDR, and then have the editors do a batch digitize to get 
the shots off of the DDR.  I know, it's screwy, but it's the fastest 
method we had at the time to get shots in.  And when you're turning 
around over 100 shots in a day, speed is essential.

     So what's the problem now?  Avid broke support for DDRs, saying 
that instead of having support for generic tape decks, which is what our 
DDR emulated, they now have support for each individual type of deck. 
 Of course, this means our DDRs no longer work as reliably as they used 
to, and the editors are getting constant errors when they digitize now. 
 Add in the fact that the company that made our DDRs has long since 
vanished off the planet, and it becomes obvious we need to find another 
way to get animation into our Avids.

     Which brings us to why I'm using mencode and vfw.  I need some way 
to automate encoding our rendered frames into a QT file using Avid's 
Meridian uncompressed codec, since during all the tests I've done, it 
has the fastest import time, hands down.  Example:

224 frame shot @ 720X486
QT Raw - 40 secs to import
QT Animation Codec - 34 secs to import
QT PhotoJpeg Codec - 37 secs to import
QT Avid Meridian Uncompressed Codec - 8 secs to import

     Importing the individual frames was faster than the QTs (except the 
Avid codec'd one) at about 25 secs, but the import time of that Avid 
codec'd file is just awesome, and is actually faster than the method we 
currently have in place.  Right now, I'm using QT Pro under Windows 2000 
to make the QTs, but QT Pro has no command line functionality (and is 
Win2K based).  So, what we need to do is find a way to automate the 
creation of these files under Linux.  After badgering various mailing 
lists, your project was mentioned, and it is by far and wide the most 
promising.

     After all that, here's where I'm at now.  I tried running:

mencoder -v 671_00_01.tga.\* -mf on:w=720:h=486:fps=24:type=tga -ovc vfw 
-vfwopts codec=AvidAVICodec.dll -nosound -o output.avi

and I get a seg violation...  I'm recompiling now with debug to see if I 
can get any extra info out of it...  Interesting...  It's dying at line 
173 of aclib_template.c.  Actual message is:

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 1024 (LWP 30031)]
fast_memcpy_MMX2 (to=0x380488b0, from=0x4054688a, len=2144) at 
aclib_template.c:173
173                 small_memcpy(to, from, delta);

Hmm...  Perhaps if I compile with MMX off?  Interesting.  Well, I'm 
gonna keep beating my head against this, but in the meantime, I have 
another question for you guys.

     What would you guys charge to include the functionality I'm looking 
for into your package, and how long do you think it would take?  I 
mentioned to my boss just how promising your tool looked and he told me 
that they'd be willing to pay to have someone "fix it" for us...  I 
might as well spec out exactly what we'd want a custom tool to do as 
well, just in case anyone does want to write some code for us...

Here's what we would need (based on what we do now).  A Linux command 
line utility that takes a series of frames in whatever format you guys 
are comfy with (targa is fine), takes the first frame and writes the 
text from a textfile on top of the image (The text would be something 
like "Shot Number: 155_00\nTake: 03\nAnimator: Bob\nFrames:224"), then 
copy that frame six times, then begin the actual sequence after them (we 
do this to help the editors in referencing the shots), then duplicate 
the last 6 frames as well (again for the editor's sake), then do a 3/2 
pulldown on the entire sequence (the 6 header frames + the original 
frames + the 6 tail frames), and write the whole bloody thing out to a 
QT file using the Avid Meridian Uncompressed codec.  Simple, no?  ;-p

     I know this is silly, but we're in a bind, and since I know next to 
nothing about reading/writing image or video formats, and don't have the 
time to learn, perhaps you lads could help us out.  And as always around 
here, we're in a rush, so we'd need it done soon.  Just a thought I had. 
 I don't know how much they'd be willing to pay for something like this. 
 What do you guys think would be fair?  Anyways, back to banging my head 
on that seg violation(inbetween render wrangling and helping artists). 
 Thanks for the help.  Later,

J. J. Franzen