[brodie_geers]

So I've probably spent more time tweaking thinSSS settings that most since it's come out but they remain a mystery to me. I've read the manual and I have a pretty good grasp on what each of the properties do, but how that translates into the actual material often seems counterintuitive. Let me show some examples and see if any of you can help me out.

For the purposes of these tests what I'm eventually trying to get to is a ThinSSS material which will be used for the white part of a dandelion among some grass. So it'll be white but fairly translucent.

1. Why does lightening the Scattering color darken the material color?


2. Why isn't this material white?


3. Why does Roughness have such a dramatic affect on how translucent the material is?


4. Why does changing the Attenuation from 1mm to 1m have so little affect on translucency? Note: values under 1mm did make the object gradually more opaque as one would expect.


5. What's the proper Nd for a leaf? Nd has a large impact on translucency but I'm not sure what an appropriate value is.

6. Why does Ref0 make such a large difference on translucency?


-Brodie

[Half Life]

That's alot of questions and I can't answer them all right now -- try bringing them one at a time.

the most important answers are these:

a) Raising roughness will increase the influence of reflectance 0 (same as always)

b) Raising the attenuation above 20mm is pointless as 10 mm is the maximum thickness possible with ThinSSS... if you set the thickness for 1mm then the attenuation should not be higher than 2mm or you will see no difference from 2mm.

c) Avoid changing the asymmetry (from 0) until you have nearly the result you seek through other settings -- it's very sensitive (extreme settings are unrealistic).

d) Use my ThinSSS ghost technique as it will give much better and more controllable results in nearly every instance ( http://www.maxwellrender.com/forum/view ... 84#p337684 )

Best,
Jason.

[brodie_geers]

Giving that a shot. It does seem to help although many of the settings are still a mystery to me. For example...

7. Why does Ref0 in the Base layer make an enormous impact on translucency? Black = translucent, White = Opaque Note: the same holds true for the Ghost layer

[Half Life]

I'm not seeing your thinSSS settings there.

I will say I suspect (and have suspected for a while) a bug in thinSSS that causes the thickness parameter to not work properly.

It's also worth knowing that thinSSS always works better with images/maps.

Sometimes you can ignore the rules and get ok results anyway -- but true realism may not be possible since thinSSS is really a "cheat" of the system... true volume SSS is far superior (if less practical sometimes).

Best,
Jason.

[brodie_geers]

So far here's my list of things that affect the level of opacity. Left side makes it more opaque, right side makes it more translucent...

[edit: I got this part backwards the first time around]
Ref0 Lightness
Light/Dark

Note: There is not a corresponding slider for Lightness within the Maxwell Material editor. Black has a lightness of 0 (Translucent) and White has a lightness of 100 (Opaque) but colors also contain their own lightness value. Dark green will be more translucent than light green (ie. all else being equal, a light green leaf will be more opaque than a dark green leaf). Blue will be more translucent than yellow. Lightness values can be checked in Photoshop. Changing lightness in either Base or Ghost layer has an affect but the higher the roughness the more the affect it will have.




Roughness of Base Layer
Rough/Smooth

Note: This value takes precedents over Dark/Light. If you set roughness to 0 you'll have maximum transparency which can't be overridden by setting Ref0 to Black.

Asymmetry
+1/-1

-Brodie

[brodie_geers]

Here's another tidbit I'm seeing regarding the relationship between Scattering Color, Coef, and Asymmetry

Positive Asymmetry gives the object more of the color of the Scattering. If the Coef is low it will give it a little of this color, if the Coef is high it gives it more of this color.

Negative Asymmetry give the object the OPPOSITE of the Scattering color. Coef has the same relationship as above (more Coef = more color influence).

This affect seems much more noticeable in backlight situations than frontlight.



-Brodie

[brodie_geers]

I'm not seeing your thinSSS settings there.

I will say I suspect (and have suspected for a while) a bug in thinSSS that causes the thickness parameter to not work properly.

It's also worth knowing that thinSSS always works better with images/maps.

Sometimes you can ignore the rules and get ok results anyway -- but true realism may not be possible since thinSSS is really a "cheat" of the system... true volume SSS is far superior (if less practical sometimes).

Best,
Jason.

The "rules" are exactly what's puzzling me. I can set a camera to ISO 200 and know that my image will be brighter than 100. I can set a material to Roughness 80 and know that will make it a bit shinier than 97. But in ThinSSS land dark green is more transparent than light green, green scattering might look red, some things that seem like they should have great influence don't and vice versa.

I'm going through so many tests I don't have the exact settings but I think they were something to this effect. I bumped the Asym to -1.0 to amplify the affect.



-Brodie

[Half Life]

What exactly are you trying to get at Brodie? I'm not sure if you want to make flowers or understand thinSSS settings.

If you are trying to understand thinSSS settings -- I typed up a really long response but tossed it because I personally think either thinSSS is broken or it needs to be moved off into it's own UI like coatings and emitters because it does not strictly follow the rules of SSS.

ThinSSS works best with images, you will have a hard time creating basic materials with thinSSS -- the scattering slot was added to thinSSS (and is not accessible any other way)for a reason.

Best,
Jason.

[Half Life]

leaf.zip

This is a good example of why I came up with the "Ghost" approach -- wrap the original leaf on the simball to compare and look at it lit from the front and you will see it looks very unnatural... with the ghost technique you can get a good result for lighting on both sides.

I tried for months to produce a usable result (in all lighting) with thinSSS and performed thousands of tests -- the Ghost technique is the easiest and best way to get anything useful out of thinSSS (as it is now).

It's simple and it works.

Best,
Jason.

[brodie_geers]

What exactly are you trying to get at Brodie? I'm not sure if you want to make flowers or understand thinSSS settings.

Well, both. What spawned this was wanting to make some white translucent dandelion type flowers to scatter around my grass. In honesty, at this point my best bet, is almost definitely to simply use a translucent material rather than ThinSSS. However, this is a dilemma I've long wanted to understand because more importantly than my white flower is grass and leaves. I might have lucked out and got my grass where I want it but leaves provide a more difficult problem because each tree has it's own leaf map with different light/dark values and colors. Maintaining settings and simply replacing the maps seems to have a pretty poor outcome, presumably because of all of these issues I'm finding out as I go. I feel like if I could get my head around what settings affect what then when a new leaf comes along I might be able to get a decent result without all the hassle.

It's quite frustrating that so many factors seem to affect the result in such a huge way. You'd think all you have to do is get the material looking right and then add a few ThinSSS properties (my leaf is about yey thick and yey translucent), no big whoop. But that doesn't seem to be the case and each of the ways I've found to create ThinSSS seems to have it's own issues. A one layer ThinSSS has one set, the default ThinSSS leaf has another set, your GhostSSS has another set, and a leaf that JD graciously created for me some time ago (I go through this frustrating experience a few times per year) has another set.

If you are trying to understand thinSSS settings -- I typed up a really long response but tossed it because I personally think either thinSSS is broken or it needs to be moved off into it's own UI like coatings and emitters because it does not strictly follow the rules of SSS.

If ThinSSS was acknowledged as broken or at least in drastic need of development then it would ease my burden considerably.

ThinSSS works best with images, you will have a hard time creating basic materials with thinSSS -- the scattering slot was added to thinSSS (and is not accessible any other way)for a reason.

Granted, but I started out with what was essentially an almost totally white .tga file for my white material. It seemed a little sillly to stick with that for testing purposes so I moved to colors. And with your Ghost SSS the main issue I see is that as I mentioned in a post above (which I've edited because I got it backwards the first time) the Lightness of the color has a strong affect on the translucency of the material. For example, I've taken your material tweaking only a couple properties to be able to get some translucency and then replaced your map with a simple checker map. The two colors are the same value of green although with the lightness adjusted.



If this were, say a leaf with a thick dark green base and thin light green tips, you'd end up with the thick base being more translucent than the thin tips simple because of the color.

-Brodie

[Half Life]

If this were, say a leaf with a thick dark green base and thin light green tips, you'd end up with the thick base being more translucent than the thin tips simple because of the color.

This is where the thickness map becomes very important -- I just hacked the original image in the example above, but in reality you would need to do some (possibly extensive) editing in Photoshop to get a really good result.

But yes, using a simple photo can never get it right -- this is only a very poor approximation... the real stuff is incredibly complex and to think that a few parameters and a hacked up photo can do the real thing justice is not realistic.

The problem is the thickness map, scattering map, and transmittance map would need to be very precise and you would likely need multiple layers with opacity masks to really try to puzzle together something truly good... but I question the point -- this might be an excellent example of "losing the forest for the trees".

Best,
Jason.

[brodie_geers]

I see what you mean about the thickness map but it's overly complicated even there.

You'd expect black = minimum thickness, white = maximum thickness, and 50% gray would be half way inbetween.

Not the case.

The thickness is originally based on whatever the lightness value is of the Ref0 in the Base layer. So 50% gray will give you the same result as having no thickness map (light green = thick/opaque, dark green = thin/translucent). To get an even thickness you have to take that map, stick it into the thickness slot and invert it. So for my checker map it went from light green/dark green to light pink/dark pink which evened it out. But you can't very will manipulate thickness from a pink map. I assumed desaturating it to -100 would give the same result however it didn't. The dark green became slightly thinner again and the image was more noisey.

So if your Min/Max is 1/2 you can't say black=1, gray=1.5, white=2.0. In reality it will be more like black=1-1.5ish (depending on the color in Ref0 in that location), gray=1.3ish-1.7ish (depending on...), white = 1.5ish-2.0 (depending on...).

However none of that is quite right either because if you set min/max to 1/1 you still won't get an even transparency because what the Ref0 is affecting isn't really thickness but some actual transparency value. And don't forget, you can't just "choose" a value for Ref0 because you want your leaf to be a particular shade of brown/green/yellow.

It's maddening. It's like trying to make a car paint material but the more red you make it the more it translucent it becomes. What?

-Brodie

[Half Life]

See why I say it needs it's own separate UI like coatings and emitters... but here's the thing -- trying to too hard to "know" exactly what to do numerically is probably a waste of time. Just use your eyes and tweak until it looks good... after all these are just leaves there's bound to be thousands of them so the likelihood of anybody looking at one too closely is pretty minimal. Same with grass -- I mean after all most people just post it in anyway instead of sweating these types of details.

Oh, and don't get me started on Car Paint...

Best,
Jason.

[brodie_geers]

I can see why a separate UI could be nice but in honesty it seems like the current UI would be fine if it worked like you'd think it would without all the quirks (as you alluded to one of the huge issues with ThinSSS is that you can get it looking good when backlit but then it'll turn out way to dark and dull when frontlit - such is the case with the default ThinSSS leaf). So long as they solve the issues, I don't much care how it looks in the end.

They are small but in the end the totality of leaves and grass takes up a huge chunk of the scene and can drastically alter its look. One need only look at the best exterior archviz stuff currently available. One of the major differences between it and what people where doing not so long ago isn't the concrete material or detailed patio furniture but rather the believably and detail of the vegetation.

I've been post processing my grass in for awhile but not because it necessarily looks the best nor because I'm always in a time crunch and couldn't afford the time to put in 3d grass and let it render with thinsss. It's because I've been unable to get a very believable look in Maxwell so far. I'm honing in on it but it's elusive. Likewise, I've found a fairly believable setup for my leaves but it's not ideal and it took days (not hours) of testing.

-Brodie

[Half Life]

Yes, ideally thinSSS would work exactly as real SSS with the addition of the virtual thickness map -- generally this is what the UI leads you to expect because all the other elements work together in a way that makes sense except thinSSS.

My reasoning to move it out is primarily to cement in the users mind that we are in a completely special mode of Maxwell materials once we enter thinSSS -- which is not really obvious with the current setup and it muddies the water.

Best,
Jason.