Hi,
I think 8-bits per channel is the lowest mantra will quantise, even with a deep raster. “256 colours” usually involves some sort of colour lookup table. ie colour-num to (R,G,B) value. So there isn't an obvious/standard mapping from bits-per-channel (R,G,B) to 8-bit (ie. 8 doesn't divide evenly by 3). Well, unless you included “A” as well. Heh. Then you could apply 2bits to each channel (R,G,B,A) but that would look terrible..
Don't do this.
In other words, generally speaking, a good gif converter selects the best 256 colours from an analysis of the spread of colours in a 24-bit image. For example, if your image was all shades of blue, it would be a waste to allocate red colours in the table of 256-colours.
So, yes, you'd most likely want to do the conversion after a regular mantra render.
There used to be a “Posterise COP”. I can't find it in the latest Houdini. I see a Quantise COP, but note that it will just reduce the colour variation in the image, not the bit depth. The image remains 8-bit per channel after the Quantise Operation. Still, this op should provide a pretty good hint to any gif converter (and you can interactively see what you'll expect to get).
When I tried saving a file from COP's in .gif format I got an empty file. Similarly when I saved a .tif and tried converting it with Houdini's cmd-line based “icp”.
Nb: SESI: Unfortunately, the current
$HFS/houdini/FBio file which ships with Houdini is set to use “imgcopy” to make .gif format files. The
imgcopy program was an
sgi/IRIX thing, so it won't work on linux and windows. If you have another converter (say, like Image Magick
www.imagemagick.org) then it should be fairly straightforward for you to edit this file so that Houdini knows how to make .gif files. There's comments in the FBio file which explains how it works.
i hope that helps,
ben.