A question for the material wizards:
Which faux glass node setup is more effective for convincing but fast-rendering Cycles glass?
so basically, the only difference is reflection ray vs shadow ray?
And the Math node operator: Add vs Maximum.
1 Like
You need to expand the question.
What application is it for. The solution for a glass vase with transparent shadows is different to that for architectural glass panels for the side of a building for example.
No single solution is the best - it depends on the application.
For architectural glass panels for example (particularly for windows) I probably wouldn’t use the glass shader at all, preferring instead to have a Fresnel mix of glossy and transparent shaders.
If I wanted coloured glass, but with white reflections (as it should be for a dielectric material), i’d use a Fresnel mix of refraction and glossy (the glass shader alone in this application would colour the reflections red as well - which is not physically accurate).
2 Likes
Interesting, thanks for your insights!
I’m looking for the best balance between convincing, realistic glass properties and rendering speed (less noise).
1 Like
for best results, learn glassblowing
3 Likes
Maybe I should mark that as the solution. 
I use single pane (one sided geometry) for all my window type glass. Always starts out like this:
You get the “fresnel” darkening of the shadow for free, but having that would be an extremely rare requirement. If there are lots of internal patterned light points being reflected in them in daylight (I do office spaces), or line continuations that needs breaking, I’ll add some random normal modification to the panes; from inaccurate mounting and wobbles in the glass itself. Something like this I guess:
Downsides:
- No double reflections. Nobody cares about them anyway unless you render out detail views for a glass frame producer or something - that’s not what makes or breaks the render.
- The “fresnel” calculation is not “perfect”. But again, not a soul would ever notice. Try real fresnel IOR 1.55 (about typical for glass, and here only affects how reflections are mixed in) side by side with that layer weight calculation. Not even you would notice the difference in a blind test, let alone any client. But using layer weight you don’t have to deal with inverted IOR for backfaces and so on.
- No roughness for glass refraction since it doesn’t use refraction. If that is a requirement, swap out Transparency for Refraction and use geometry/incoming for its normal (for thin geometry). However now you have to deal with shadowing separately again.
- Single sided thin pane, so you won’t ever see any internal reflections on the edges or darkening on the exterior edges. For glass panes as separators I’d use a different technique using either real glass with volumetric absorption and high bounce count (64+), or some fakery to simulate the edge.
1 Like
I think that moony’s first response was on point, at least with regards to architecture. Basically, "what is the minimum that’s needed to persuade you that “a piece of glass is there?” You see – some kind of reflection, and some amount of reduction in the hue-and-saturation in the objects that are immediately “behind” the glass. As soon as you get anything that’s "convincing enough," stop. You don’t need to simulate actual glass – at least, not in this case.
2 Likes