First of all, I'm not part of the corona vs fstorm movement, I just want to make that clear.
But I've been doing regular fstorm tests ever since it got released. The fstorm tone mapper was dumped as ACESlog lut when fstorm was released, the problem is that corona does not have a standard loglut "decoder". Logluts are super sensitive, just look at this example
as you can see in this example, you can't just load a random log lut and get 1:1 output. We want ACESlog, because it behave as EV. Don't confuse this with the hyped ACES tone mapper.
ACES log is just like PBR, we can try to ignore it, but it is the standard in cg.
I've been doing regular tests the past few weeks, because I want to nail the difference between these two renderers. The render settings have been the same.
"PT + PT
32 Light sample
20 gi clamp
25 ray depth
Tone mapper off"
The main difference is that fstorm has "diffuse roughness", and it is at 0.8 default. After some tests, it is clear that fstorms max roughness setting at 0.99 yields the same result as corona. Fstorm has softer shadows than corona because of this.
If Corona want to match this, we need a diffuse roughness parameter.
But there is another factor, and I don't know what it is yet.
fstorm seem to do something that looks like a secondary gi bounce. open the example pictures in photoshop /affinity and toggle between them. fstorm is 5 RGB brighter in the shadow areas, and this is always the case. I've been doing ALOT of tests, and fstorm always bounce GI more. reflection is disabled, and fstorms reflection render element is black. I don't know if this is a fact, but I feel this help the noise % go down quicker.
