Hello.
Some glasses from Ikea, the Pokal model.
Regards.
This a practice? Just roughness map?
See:
Hello.
A quick update and my conclusions: Corona handles dielectrics very well. It’s very difficult to achieve seamless integration of caustics with the HDRI environment without losing realism. You either have clearly defined caustics, which sacrifices realism, or you have realistic lighting, which sacrifices caustic definition. However, this is common to the generic path tracing algorithm.
Best regards.
Can you explain what you mean by that? Corona should simulate caustics based on the HDRI you are using. In your examples, the HDRI seems to have a large light source, like a window, which casts soft shadows, which means that caustics should also be soft.
Are you using the fast caustics solver?
Hi.
To obtain well-defined caustics, we are forced to reduce the size of the light sources to physically incorrect values. This generates a photon density of frequency that is too high for unidirectional Path Tracing engines to reconstruct smoothly. The result is an artificial contrast: either you have blurry caustics (physically plausible but imperceptible) or you have sharp caustics (visible but clipped and aliased). In my opinion, this is an intrinsic limitation in the dynamic range management of current photon solvers integrated into unidirectional engines.
Bidirectional engines do not suffer as much from this problem because they trace the light from both ends simultaneously, but in unidirectional engines, the Fast Caustics Solver has to resort to interpolations that fail when the light source is a mathematical point (Sun Size < 1.0).
Regards.
Errm… do you have any evidence to support those claims? If you’d like, we can check your scene and find out whether you are getting the expected caustics results or not.
Also asking once again: are you using Corona’s caustics solver or just path tracing?
Hi.
Path tracing + Path tracing
Caustics from environment
Caustics Solver
Usefulness ratio 50000
It’s just for fun and learning. I’m testing several render engines, both bidirectional and unidirectional. They all use the same algorithms and have similar problems. Your photon solver is incredible, and its ease of use is unsurpassed. This is just an observation about the architecture of Photon Mapping and Dynamic Range. Here’s a summary of my experience:
Render Engine Architecture Handling Sharp Caustics The Real Problem
Corona / V-Ray Unidirectional + Solver Fast and clean on large surfaces. Small suns (Sun Size < 1.0) create energy spikes (clipping) that the reconstruction filter struggles to average.
LuxCoreRender Bidirectional (VCM) The gold standard for complex caustics (pools/diamonds). Extremely slow for complete architectural scenes. It struggles a lot with complex HDRI skies.
Maxwell: Pure bidirectional renderer. Physically perfect; doesn’t need solvers. Infinite rendering. Sharp caustics generate a “salt and pepper” noise that takes hours to disappear.
Arnold (GPU/CPU): Unidirectional. Very poor with caustics; requires visibility cheats. It’s simply not a tool designed for high-frequency architectural caustics.
My favorite is Corona.
Fstorm and Octane, especially Octane, have hybrid algorithms I’m unfamiliar with. From what I can see, Fstorm doesn’t calculate caustics, and Octane seems to have a photon solver similar to yours, but being GPU-based, I suspect it’s much less efficient with complex bounces.
The scene is very simple: three glasses, a plane, an HDRI, and a corona light.
Best regards.
Thank you for the details. Those are some interesting observations. I think our solver does a pretty good job producing realistic caustics in terms of both their brightness and sharpness, but if you have some results where it fails, or isn’t that great, please do share them!
Hello.
It’s much more complex than simple reflections and refractions (Ray Tracing). Understanding Path Tracing isn’t about understanding how to draw an image; it’s about understanding how radiance is transported through a scene while respecting the energy conservation defined by the BRDF of each material.
Beautiful reflections.
Best regards.
why then, do your images look ‘uncanny’? (RhetoricalQ)
(finger prints look as if etched on or forensically powdered)
tip, if you really wanna get into ‘ze nitty gritty’ of the light transport simulations in the spectral realm, and not just another shitty RGB cloister fukk w/o abilities to trace all the paths (SDS+), see yourself to the SPPM w/in old LuxRenderer
&BTW, LC’s bidir VCM was never finished
A quick update.
I’ve been testing several engines for a while now, and I’m sticking with Corona. All the menus are so well explained. You don’t need a thousand-page manual, everything is super intuitive and exactly where you expect it to be. Clear and intuitive menus, no manual needed, interactive rendering that flies, and network rendering so robust it could practically make a Spectrum render. Pure power without the hassle. We’ll keep learning.
Cheers.
The girl’s model:
https://free3d.com/es/modelo-3d/woman-statue-496.html?dd_referrer=https%3A%2F%2Fwww.google.com%2F
Cheers.