I would tend to agree with François, ambient occlusion is function of 3-space where ao(x,y,z) is the proportion of the solid angle around that point that contains geometry that blocks ambient light. All geometry, even transparent geometry, blocks ambient light. So, if you are seeing darkening on a refracting or reflecting surface it is because you are not combining ambient occlusion correctly, not because the AO has been calculated incorrectly. A refracting or reflecting object really does block ambient light 100% and only lights other surfaces through caustics (or itself through transmission).
If a surface is partially refractive or reflective, but also diffuse, then only the diffuse part should be affected by AO. If it is because the object has alpha transparency then that is not a physically based kind of lighting and then there really isn't a "correct" answer. People might give different answers if you asked them should the object give AO to itself vs contributing to other objects in the scene. From Brecht's explanation it sounds like Cycles just ignores this kind of transparency for global illumination purposes. On Sun, Apr 22, 2012 at 5:21 PM, Brecht Van Lommel <brechtvanlom...@pandora.be> wrote: > Augustin is right, it works, but only for the Transparent BSDF. In > Cycles, Transparent is defined to work as, "render as if there was no > geometry there", which also affects AO and shadow rays. > > The case could be made that the Glass BSDF should also work this way > for AO rays, but it gets fuzzy then, where depending on the particular > settings of a BSDF it may seem a good/bad idea to make it transparent. > Right now there's a clear and consistent rule at least, and you can > control it with the light path node. > > Brecht. > > On Sun, Apr 22, 2012 at 11:32 PM, François T. <francoistarl...@gmail.com> > wrote: >> As AO is geometry based, as far as I know, it is not supposed to make >> difference on its transparency, as at least the basic algorithm only check >> for mesh data, but not the material. >> >> >> 2012/4/22 Agustin Benavidez <agustinbenavi...@gmail.com> >> >>> I am pretty sure it does work with transparent BSDF, and not with caustics >>> i.e. using glass bsdf , so if you are using glass bsdf, use light path node >>> to switch to transparent bsdf at least for shadow/diffuse rays or whatever >>> ray you need. >>> Cheers >>> >>> >>> 2012/4/22 Tobias Oelgarte <tobias.oelga...@googlemail.com> >>> >>> > I made some experiments with Cycles today and I noticed that AO sees all >>> > faces as opaque, like the internal Renderer did. That means that a face >>> > with glass material (entirely or nearly entirely transparent) will be >>> > dark (itself) at parts close to other geometry. Should AO not have much >>> > less impact on such surfaces? Otherwise it can be hardly combined with >>> > transparent materials. >>> > >>> > Do you see room for improvement or should it stay this way? >>> > >>> > Greetings from >>> > Tobias Oelgarte >>> > _______________________________________________ >>> > Bf-committers mailing list >>> > Bf-committers@blender.org >>> > http://lists.blender.org/mailman/listinfo/bf-committers >>> > >>> _______________________________________________ >>> Bf-committers mailing list >>> Bf-committers@blender.org >>> http://lists.blender.org/mailman/listinfo/bf-committers >>> >> >> >> >> -- >> ____________________ >> François Tarlier >> www.francois-tarlier.com >> www.linkedin.com/in/francoistarlier >> _______________________________________________ >> Bf-committers mailing list >> Bf-committers@blender.org >> http://lists.blender.org/mailman/listinfo/bf-committers > _______________________________________________ > Bf-committers mailing list > Bf-committers@blender.org > http://lists.blender.org/mailman/listinfo/bf-committers _______________________________________________ Bf-committers mailing list Bf-committers@blender.org http://lists.blender.org/mailman/listinfo/bf-committers