[Bf-cycles] Volume shaders, mirrors that project light sources

[Brecht Van Lommel]

Hi,

On Thu, Nov 17, 2011 at 5:39 PM, Lucas <wsacul at gmail.com> wrote:
> I'm hoping http://cgi.stackexchange.com becomes public soon and has lots of
> blender support for some of these possibly remedial questions:
> How is volume shading done with cycles in the current svn trunk?
> In the following picture the transparent shader behaves intuitively but a
> glass surface shader still casts a shadow.  I think this means I need to do
> volume shading in addition or instead of surface shading, and in the node
> editor view I see a block called 'material output' with a surface and volume
> input and the surface input already hooked up, but what do I hook up to the
> volume input?
> https://picasaweb.google.com/103190342755104432973/Blender3D#5675708974274649330
> Turning off shadows works for the glass objects but is only physically
> correct if they are 100% white.

Volume rendering is not supported yet. The link doesn't seem to work,
but I think this is a case where you've got caustics that path tracing
can't render well. There's no simple solution to this, it requires a
more advanced rendering algorithm. A workaround is to make a shader
that switches to a transparent BSDF for shadow rays, using the light
path node.

> Is there any way to pass hints to send more lightpaths bouncing off one
> object to a secondary object?
> In this example there are light paths that hit the diffuse bsdf of the
> shadowed area of the plane and only a few of them hit the mirror and trace
> back to the light source I have pointed at the mirror, resulting in only a
> smattering of light reflected by the mirror.  Many many more render cycles
> would fill in the gaps but at huge cost.
> https://picasaweb.google.com/103190342755104432973/Blender3D#5675634439005970706
>
> One approach that comes to mind is to have a special shader that takes an
> object as a target, so that the peak response is always pointed towards it
> (but also has a softness factor that maybe would have to be increased
> manually if the target object is large and close, or there could be an
> automatic way of doing that).  Is that possible already?
> My workaround is to use a mirror modifier, and I think some additional
> compositing work could make it so light from that mirrored light source that
> doesn't strike the profile of the mirror is masked out.

Again the link doesn't work, but this seems another case that path
tracing can't handle well. Caustics need a more advanced algorithm
like bidirectional path tracing or photon mapping.

Brecht.