Originally posted by nicola_barbano
View Post
This is a minor problem for the progressive sampler (it'll take longer to converge.), but it could be a deal breaker for the traditional photon mapping.
Corona suffers from the same kind of issue, however: put a caustic-enabled object far away from your subject, even not seen by the camera, and Corona will have to take it into account, meaning your caustics will immediately look different. (see attached).
The two images show the same number of passes.
The first scene contains only what you see (some cosmos asset with refractive properties), the second also has a glass, caustic-enabled sphere somewhat far from the subject.
The number of photons reaching the subjects is dramatically lower, and so you can tell most of the left-side caustics are absent, and the right-side diffraction isn't nearly as resolved.
It's likely it will converge eventually, but there's no magic at play with either V-Ray or Corona: it's all down to proper set-up (and, in some case, the amount of memory, or time, available.).
Comment