OTOY Forums

Let us know when you get home, if you're talking gibberish

treddie wrote: Someone remarked that they had this hard edge problem too and it only happened when they had a small light on inside a car. For me, I happen to have the turn signal and rear taillights backed up by a very weak emitter so that they would not be so dark. Not sure if that is causing it, but it was interesting that someone else had a problem with emitters and hard edges in the shadow catcher.

That was me, and it was definitely what was causing the weirdness. However, as it said it was fixed in March (in Lightwave anyway). Not sure which version you are using. It may be the problem has crept back

Yes, both are correct. That is Color Theory. In terms of exactly what Octane does when calculating shadow color may not be as obvious as that, but if MLT is TRULY physically correct, then it all reduces down to the basics of the Theory of Light and Color when all is said and done.

xxdanbrowne wrote:

gordonrobb wrote:I thought that the colour of the shadow was affected by the diffuse colour of the shadow catcher object. No?

I could be talking crap here but it seems to me that both could be true.
i.e. the color of the shadow is affected by the color of the light source
AND/OR
the color of the shadow is affected by the diffuse color of the shadow catcher.

But again, I could be talking gibberish. No way to test it till I get home.

Nope, the shadow is always black only. The color of the matte material is currently totally ignored and doesn't have an effect on the shadow color. The only thing that has an effect is the opacity.

@abstrax,

Now you got me curious. I'm pretty sure I saw colored shadows. Now I have a burning NEED to go test it...

xxdanbrowne wrote:@abstrax,

Now you got me curious. I'm pretty sure I saw colored shadows. Now I have a burning NEED to go test it...

Maybe in the past. The behaviour of matte material has changed multiple times of the last 20 or so releases.

abstrax wrote:

xxdanbrowne wrote:@abstrax,

Now you got me curious. I'm pretty sure I saw colored shadows. Now I have a burning NEED to go test it...

Maybe in the past. The behaviour of matte material has changed multiple times of the last 20 or so releases.

No it's cool. Obviously you're the expert. I just have to see it with my own eyes so I don't think I'm crazy...

In essence, I think we are talking about the same thing. By "shadow color" I mean the base color with whatever is mixed in with it to darken that color. I believe (correct me if I am wrong) that you are referring to "shadow color" as the color (black) that is placed OVER the base color.

Technically, the shadow color (your definition of it) would become a problem if it were correctly a combination of both the compliments of the light source color and the base color, since that shadow overlay needs to be able to be dropped over ANY color background in post. If the base color of the target background is different than what was in the original render, the color combination would look out of place and just wrong. So I can understand how a simple % of black is necessary in 3D work because that shadow overlay has to be able to be applied to any prospective background image during compositing.

You can get away with that much easier in RGB than in CMY since in CMY work, adding black to darken something deadens the color and everything has a gray cast to it. You can always tell if a painter knows his color or not from that alone. In painting (and even when working in CMYK in PShop) you should NEVER add black to darken a color until you have added the maximum amount of compliment and STILL need to take the darkness down a notch. The only exception is in CMYK work, where black is only added due to the white spaces between printer dots washing out the colors in the image. If the colors were laid down continuously (no individual dots but a continuous glaze), black would actually darken the image too much and again, would add a gray cast to everything. That is one reason why print media has a narrower color gamut then RGB...The black contaminates the color intensity, but it is necessary to help counteract the white spaces between ink dots. A case of "you can't have your cake and eat it too".

I assumed when first working with the shadow catcher that the gray tone to the catcher surface was normal. So my entire workaround was based on that, which ended up being a wild-goose chase. Turns out this is a bug as discussed in my other thread, "I think this is a Real Shadow Catcher problem this time":
http://render.otoy.com/forum/viewtopic.php?f=9&t=40027

Can there be a fix so that these emitters do not cause this? Otherwise, you have to do two to three passes to get one final image:
1. A shadow catcher render for just the shadows, where any emitters in the subject need to be turned off.
2. A beauty pass for just the subject with its emitters on, w//o shadow catcher on.

If you want the best result, a third pass:
3. A mask image for the subject needed to erase "swelling" of the outline of the subject when the beauty pass is layered over the shadow pass in PShop.

gordonrobb > Yes, that WAS you!

As for color in shadows, here is a perfect example. Notice that the shadow color on the left in the background image (environment) has cool colors in it. Again, this is light source and base color compliments mixed in with the base color itself. That is nature at work captured by the camera. Then look over to the right where you have the shadow catch shadow...It is just black.

Although the shadow catcher color is wrong, there is no easy way around this in a renderer. After all, if you plan on taking the subject with its shadows and placing them over another image entirely, not related to the environment image (in other words, not a backplane image that came with the environment image) then that coloring will look wrong to one degree or another.

The black shadow here, under the car certainly looks wrong, but it is probably easier in PShop to get chroma into that shadow by using Channel Mixer or Selective Color than trying the same thing when the shadow started out with color. The reason is that shifting color in a colored area means parts of that area may start to shift completely into wrong colors when trying to adjust overall. That is a common color correction issue. In the case of shadows, probably best to add color to a black shadow than an already colored one.

Unfortunately, if you go to the extreme of creating authentic colored shadows in PShop from shadows created with Shadow Catcher, then you will need to run a separate render of a mask of your subject, so that you can color correct only the shadows without affecting the subject. The faster alternative is to create your mask by hand in PShop and feather it properly. Some people are better at that than others.

Then again, for the BEST results in PShop, the Shadow Catcher shadow layer should be set to Multiply mode so that details in the background are not weakened. You can't have the subject set to Multiply, so that means either splitting your shadows from your subject into two separate layers via the mask (which means you will get a halo around the subject, unless you tinker with the PShop selection with Expand/Contract), or running a THIRD render of just the subject w/o shadow catcher on, for the beauty pass. If the latter, you will still need the mask then to erase out "swelling" of the subject due to Multiply mode expanding the apparent contour of the subject in the shadow catch layer by a few annoying pixels, beyond what you see on the beauty pass layer above it.

No matter what, for anal quality there will be a fair amount of work to do in PShop. Here is a PShop color-corrected version: