How to handle super bright highlight?

[joelly3d]

If a light and a surface hit each other just right, it can create a white spot that really hangs during render. In several scenes over the last few months, a super bright white highlight, typically on metals and glass, will hold up a render.

What is the "correct way to handle this from a render theory standpoint?

I don't want to turn on clamping in the Color Mapping settings and I don't want to tweak the material so that it dims the other sections too much. Dimming a light can change the scene too much.

Typically, I just wait it out, but is there a better way to handle this?

Any thoughts appreciated.
-Joel

[^Lele^]

Currently, there is no efficient way to handle these automatically (but there will be, at least to some degree, with a minimal difference in output result.).

What you can do, at the cost of some slight visual difference, is to ever so slightly reduce glossiness for the shader(s) in question.
Attached a (admittedly artificial, albeit quite pure) sample: the first sphere shader is a pure metal with a Gloss of 0.99, the Second of 0.975. The light is reflected with a peak pixel value of ~150.0f in the middle of the highlight.
They are both rendered with 1-10000 (10k) and a N.T. of 0.01
The first image took 62.3s to converge.
The second took 20.2s.
Peak energies were not that much different, and so it isn't the rolloff (call it Tail, if you wish) of the highlight (third image, exposed down).

Notice mileage may vary, as highlights may contain any amount of energy, so striving for plausible settings remains paramount.
The latitude of visual differences will also very much depend on the shape of the geo, and resulting highlight, so in some cases you may be generous with the lowering of glossiness, others you may need to be restrained.
Regardless, If you can live with these visual differences, this could prove quite the life-saver.

Edit: if you render a shot with plenty of DoF and/or Motion Blur, this trick could be used even a bit more, as the blurring generated by those effects will likely overpower the visual differences introduced by the lower gloss.
Sampling less in camera-blurred areas is something Reyes Renderman used to do to extremely beneficial effect.

[joelly3d]

Thanks for the tip and supporting visuals.

[^Lele^]

Thanks for the tip and supporting visuals.

You're most welcome!
I'd be grateful if you let me know if this helped in real-life scenarios: i haven't tried this "in battle" for a while now.

[^Lele^]

It occurs to me just now that some competitor has this trick of capping max glossiness either hardcoded into the BRDF (i think old Corona did this, but don't quote me on it.), or exposed in the render settings (Arnold.).

To plug our current hole, here's a one-liner that will turn all of the current scene's v-ray materials with a gloss above 0.975 to that value. Use with care (save before running it, or use the undo entry it will create, if you do not like the results.), and feel free to edit the *two* numerical values to suit your needs.

for m in (getclassinstances vraymtl) where (m.reflection_glossiness > 0.975) do (m.reflection_glossiness = 0.975)

[joelly3d]

Thanks I will keep that handy. I realize i'll have to alter my maps with a color correction to limit in the same manner. all good thanks!

[kosso_olli]

It occurs to me just now that some competitor has this trick of capping max glossiness either hardcoded into the BRDF (i think old Corona did this, but don't quote me on it.), or exposed in the render settings (Arnold.).

Is this something where research is being done?