movie-making for the compleat novice

[David R. Blythe]

In article <26471.27318f0c at kuhub.cc.ukans.edu> arritt at kuhub.cc.ukans.edu writes:
>Here's our situation: We have a SGI 4D/25, are buying a 4D/35, and want to
>be able to make videotapes of results from our numerical models.  We
>ordered something called a "genlock" board with the 4D/35 because we were
>told we would need it to dump the screen to a VCR.
>Once we get our hardware set up, what kind of software do we need?
>Do we need a special animation package to make it all work?
>
>
>What we'd like to do is to keep using the same graphics programs we're
>using now to generate "stills", and dump the data a frame-at-a-time
>to a VCR.  (Fortunately we need to make only fairly short
>movies, maybe a minute or so at a time.)  If we have to use
>a special graphics program, we'd need to either (1) recode all of our
>stuff ourselves, and like I said, we're theorists rather than programmers
>(i.e. it'd take us forever, IF we could get it to work at all); or (2) hire a 
>programmer (actually we'd like to hire a programmer anyway, but can't get 
>the $$$ given the current environment for research funding).  
>                               

First, the genlock card serves two functions:  it produces composite
video suitable for recording on a north american vcr (NTSC).  Second,
it provides the ability to synchronize the iris video signal with
a sync signal provided by a external source.  This is used to produce
a very stable video signal while recording.  To make use of the board
look at the example code in 4Dgifts/examples/video for setting up
the board.  Note that the image size will be reduced to the 646x485 pixel
region in the lower left hand portion of the screen rather than the normal
1280x1024.  (i.e. the constants XMAX170 and YMAX170 in <gl/gl.h> can
be used to set up the window.)

You can record directly from the iris by putting it into NTSC mode
and connecting the composite ouput cable from the cg2 (genlock) board
to your vcr.  However, this set-up will only allow you to record
in real-time not frame at a time.  So if you can't display your images
fast enough you lose.  You will invariably have pauses and other
interference from other programs running on the system, etc so this
technique is only useful for making quick and dirty videos or demonstration
videos where you are going to interact with the program.  It also helps
if you have lots of memory to hold the images in memory for rapid 
sequencing through them.  For an example of that, look at the program
4Dgifts/iristools/imgtools/movie.

If you record frame at a time, then you need fancier equipment.  First of all
you need a vcr which is capable of doing frame acurate recording.  This is
usual done on the vcr as an insert edit.  Second the vcr needs to be remotely
controlled and you need something to do the actual control (an animation
controller).  The vcr also should be kind of heavy-duty or you will wear it out
pretty quick.  Such a vcr usually has glass heads, among other things.  We (and
many other places) use a Sony BVU-950 3/4" u-matic vcr.  It costs about $17K
canadian and is ``industrial quality'' i.e. better than home use (+ or - some
caveats) but not broadcast quality.  Other (cheaper) possibilities include
various S-VHS vcrs from JVC, etc.  I am not too up on the part of the equipment
range.  Best of all would be the Sony LVR-5000 video disk recorder for about
$25K US, which can record very very fast relative to tape and very high quality
(i.e. broadcast quality). Tape machines have to do a 5-second preroll to make
sure everything is up to speed and in sync before recording so it takes
~10 seconds to record each image :-(.

The animation controller is the intermediary between the workstation software
and the vcr and is used to position the tape, tell the vcr when to record, etc.
Diaquest, Lyon Lamb, and Videomedia make pretty reasonable stuff  (we have a
Lyon-Lamb Mini Vas which is pretty commonly used).

Finally you need the software to display images and talk to the animation
controller.  This is the cheapest part.  You can get a copy of gplot
from the pittsburgh super computer centre (ftp from calpe.psc.edu) which
can display NCAR cgm files on a number of machines and has a driver
for a diaquest animation controller and sony video disk recorder.  (Last
I looked though the sgi gl display driver still needed some work).  I believe
minivas drivers are also available although I forget from where (we
have one which we have never distributed to anyone before).  I think some
of the NCSA tools have hooks in them for video recording (at least they
can display in NTSC sized windows in near real time for some data sets).
It probably wouldn't take much to add drivers for frame at a time.
The tough part is converting your data into their annoying HDF format.

Unfortunately, I think you are still going to have to spend a few more
$$ for additional equipment and find someone with a modest understanding of
video to help you get it set up.

Hope this helps you out.

	david blythe
	ontario centre for large scale computation
	drb at clsc.utoronto.edu