Disable reflective caustics

[maru]

In some cases like lots of reflective surfaces + intensive lights, fireflies appear and don't go away after many passes. A quick way to get rid of them is to use the rayswitch material, put the reflective material into direct visibility slot and put material's copy without reflectivity into GI slot. How about automatizing this process? Like a "no caustics" checkbox next to material's reflectivity. Or something like additional MSI for reflective materials? Quick example in attachment. Better example here (picture by Tanakov).

[romullus]

Could you attach this scene, maru? I have one thing in my mind that i'd like to test.

[maru]

Here it is, slightly modified. I noticed strange things are happening to these objects when they are reflected or refracted ad messed up a little with their settings but generally it's the same thing.

[maru]

Now this is getting a bit hardcore when you want to do this with a glass object with absorption.
+10 to hardcore if it's reflected/refracted in another object. :/

[Stan_But]

yep
you cut all secondary reflections without reflection in GI slot

[maru]

Ok, actually it looks quite fine with:
-the material without reflectivity in GI slot
-the normal material in all the other slots
Sometimes interactive rendering acts strange when switching materials in rayswitchmtl and this was the cause of my problems. It looks ok in final rendering.

[Stan_But]

You can try also decrease MSI to 10 for example. It helps too in same cases

[maru]

As far as I remember Keymaster once wrote that MSI does not affect caustics generated by materials with 1.0 glossiness. But I may be wrong.

[romullus]

I think this is bug, at least i reported it some time ago: https://forum.corona-renderer.com/index.php/topic,5169.msg35149.html#msg35149

I attached two pictures. both are rendered with exactly the same settings. The only difference is that clean version is rendered with normal exposure from start to finish and second one is started with -20 EV and after few passes exposure was set back to default.

[Stan_But]

As far as I remember Keymaster once wrote that MSI does not affect caustics generated by materials with 1.0 glossiness. But I may be wrong.

Of course. MSI 1 is bad. But MSI 10 may be a good solution in this case. In the past MSI 10 even was default

I think this is bug, at least i reported it some time ago: https://forum.corona-renderer.com/index.php/topic,5169.msg35149.html#msg35149

I attached two pictures. both are rendered with exactly the same settings. The only difference is that clean version is rendered with normal exposure from start to finish and second one is started with -20 EV and after few passes exposure was set back to default.

You mean if EV was changed during rendering then we get extra noise?

[romullus]

Kind of. Also, i've found that extreme overexposure is harmful as well as underexposure:

feature=youtu.be

[maru]

I think this is bug, at least i reported it some time ago: https://forum.corona-renderer.com/index.php/topic,5169.msg35149.html#msg35149

I attached two pictures. both are rendered with exactly the same settings. The only difference is that clean version is rendered with normal exposure from start to finish and second one is started with -20 EV and after few passes exposure was set back to default.

I don't think this is a bug. Pardon my lack of technical knowledge but as far as I remember and understand, sampling is "optimized" based on initial exposure settings so changing it during rendering will always result in some quality loss. You should find the exposure you want and then hit render again.

[borisquezadaa]

Good to know.
I always thought there was something fishy with changing the exposure.

[Juraj]

I don't think this is a bug. Pardon my lack of technical knowledge but as far as I remember and understand, sampling is "optimized" based on initial exposure settings so changing it during rendering will always result in some quality loss. You should find the exposure you want and then hit render again.

Where did you read this ? I thought progressive mode in Corona simply samples identically each pixel, since there is no adaptivity. The "initial" (what is that ? 0 ?) exposure thus wouldn't matter at all. It works the same way for Progressive mode in Vray, which would otherwise sample adaptively based on pixel's brightness in set exposure, but in Progressive, it just oversamples everything equally.

[Ondra]

the MSI value is relative to the initial exposure... so changing exposure by 20 stops after starting render will produce weird behavior