[Bf-committers] Re: render passes & workflow

[Ed Halley]

On Dec 6, 2006, at 9:41 PM, GSR wrote:
 > PS: The more I think about it, the more I feel it is highly  
related to
 > the DupliMats idea (the ID could be the seed for the randomization,
 > for example) and should be tackled at the same time, or at least lay
 > the base for future DupliMats work.

Per DupliMats:

    http://mediawiki.blender.org/index.php/BlenderDev/DupliMats

As the guy who tried to implement DupliMats (before nodes were merged
and available), I completely agree, but there are issues.  Object IDs
can and should be assigned when objects are created, and they should
also be overridable, as discussed here.

The crucial bit that was a big problem in DupliMats was that *derived*
objects, such as those created by particle systems or dupliverts or
dupliframes (or now dupligroups, I expect), must ALSO be (1) divergent
if desired, (2) stable from frame to frame for animation, (3) stable
when the artist edits the base lists by adding and subtracting and
fiddling with unrelated objects.

An object-id pass would indeed generate 32bit "pixels" which identify
the most-covering or most-cameraward object ID in numeric form.
However, I really never considered "viewing" these buffers as RGBA
colors.  They're usually for post-processing purposes, such as
replacing UV textures without full rerendering work.  Imagine
rendering a flag once (with RGBA, U, V, id layers), and then replacing
180 nationalities as a batch pixel-processing job without Blender
involvement at all:  For all pixels with id=fabric, resample texture
at U,V.

--
[ e d @ h a l l e y . c c ]