Sub Division Surfaces

[WHiTeRaBBiT]

greetings,
I was just wondering if a MLT renderer could be capable of doing
sub division surfaces with dispacement mapping (micro pixels) on the fly
(like the way renderman prman renderers do)? I really like how these
techniques work in renderman to create very complicated shapes with
very small project files. Especially if you combine this with object
instancing. Just curious if this is possible or if the render
algorithms require standard geometry to do what it does. Just thought
it would be interesting if the strengths of renderman and indigo could
be combined.

Anyway just daydeaming and curious if anybody else knew anything
else about this.

[zsouthboy]

How does that work, exactly?

Renderman tesselates say, a plane for you, and then applies a displacement map to the triangles?

[WHiTeRaBBiT]

Yes.
http://en.wikipedia.org/wiki/Micropolygon

The scene geometry is tesselated based on its orientation
to the pixels of final render and then run through shaders
such as displacement maps and so on ...
Whats so neat about this is displacements are done at the
pixel level no matter the original mesh density, producing
very detailed models. Combine this with rendertime SDS and
model and material instancing you can get very detailed
scenes with very small project files. One of the problems with
using a renderfarm is that if the project files get really big
the startup time for test rendering can get pretty long (takes a
while to serve big files to many computers).
I've never used Maxwell so I dont know if it can do any of this,
I was just thinking combining these renderman features
with the realism of MLT rendering made sense for large
complex scenes with detailed geometry (especially on a
renderfarm).

Actually I was just daydreaming and wondered if anyone else
has thought/heard of anything similar!

[OnoSendai]

Hmm.. subdivision surfaces could be very cool.
I'm not sure if the on-demand tessellation is a viable option though, the number of ray/object intersections in a path tracer/MLT renderer will be much greater than in a renderer like renderman. But scene initialisation time subdivision would be cool, and could still be view-driven.

It would help to alleviate some of the issues with shading normals as well, which are a monster hack.

[WHiTeRaBBiT]

Thats what I wasn't sure of.
It seems that most of the geometry I'm seeing is fairly simple
compared too something you would see in renderman. The
only problem I see is SDS without displacement mapping
doesn't offer any real benefits. Rendertime SDS and DM are
usually used together to create a greater amount detail in the
model than can be viewed in the modeller but shows up during
the render. Oh well, maybe someday!!

One simple thing that may help with project size in complex
scenes would be the ability to create xml geometry files
and call them in the main project xml with transformations.
Then a complex object could be broken down (for example
leaves repeated over a bush).

[OnoSendai]

One simple thing that may help with project size in complex
scenes would be the ability to create xml geometry files
and call them in the main project xml with transformations.
Then a complex object could be broken down (for example
leaves repeated over a bush).

You can already do this

[WHiTeRaBBiT]

Very Cool

Does anyone know if theres a way to do this in the blender
exporter? Would I need to export to OBJ file and then hand
edit main xml? Not using test 6 yet, can it be done with that
in blender? Or does the exporter know if an object instance
is being used in blender and automatically do it. Searched
forums and docs but didn't see that .

Sorry still a indigo noob