"Max ray intens" vs "Clamp output"

[Art48]

Hi,

Now I basically know what the difference between those two is. At least I think so.

The max ray intensity will clamp all secondary rays to that value.
The clamp output will clamp the final rendered result to that value.

What I want to is reduce jaggy edges around very bright highlights reflected in car paint for example.

Now I think using the max ray intensity would do too much in terms of darkening the image, since we work with reflective and refractive GI caustics, for example getting a bright light into some rims to brighten them up and let the light bounce inside of them. Would it be better to take the clamp output approach in this case? We usually light our scenes in a plausible way and dont neccassarily need details in the overblown areas (floating point etc)

[joconnell]

Max ray is for reflection and refraction rays and where you get tiny speckles in your highlights, its normally down to an individual ray on a glossy surface hitting something really bright and overpowering a single pixel in the final render. Max ray will cut off those really high values instead of having to use loads more reflection samples to get a smooth result.

Clamp as you say will cut off all bright values, regardless of whether they're in the normal lighting or in the reflect / refract so I wouldn't typically use that if you're going to do a lot of comping afterwards. If you do want to clamp you could do something like set it at 3 or 4 so you still have plenty of high values in your final float files.

What might also work quite well for you, especially as you've mentioned cars is using the burn value on the reinhard colour mapping. A few people were getting some nasty render times on car renders where they had a light source inside of a headlight that needed to be bright enough to illuminate the road it was shining on, but also came from quite a small bulb size source. The last buckets near these areas were taking quite a while to get through. What Vlado mentioned is that you can start using the burn value of the reinhard colour mapping to start rolling down the very burnt out values in a, soft, organic manner (kind of similar to film response) and when the image sampler is trying to anti alias around those highlights, it's got less of a range to deal with. It'll bring down your render times pretty significantly if you try values around 0.5 or 0.25. You'll start to lose some overall brightness after a while which can be restored easily in post, but you'll get far quicker results. Just note that any kind of colour mapping other than linear or reinhard with burn set to 1 (which is actually linear at that value) will break the maths needed to 100% recombine passes in post - it won't be far off but it'll be a tiny bit off. If you're okay with that then you'll get some good benefits.

The last thing people do is add on some glow to those uber bright areas - normally around highlights you'll get the reflected light catching the atmosphere a tiny amount, so a few layers of tight, medium and broad / soft glow can take the zingy edge off of the highlights - CG is often way too sharp to start with so a bit of softening adds realism.

[Art48]

Thanks for your comment.
We don't do the real sick comping of all layers. We typically use the beauty as base, then apart from the obvious masks mostly just reflection/specular/lightselect elements to do some tweaking.
So, the clamp output doesn't interfere witrh reflective GI caustics right?

I will also try the Reinhard approach, seems like a good idea aswell.

[joconnell]

Clamp is an overall thing that kills any pixel values over the brightness set, regardless of where they come from.