Tests and questions on SSS

For tests its easier to scale the model instead.

I wish,

Blenders manual would be more explicit in how to use the SSS.

I have a rough grasp about it - but not really a full understanding.

is three layer SSS worth the trouble in Blender? The results hardly seem better than single layer SSS. I know mental ray uses a shader that makes use of the three layer specifically. Is this material layering in Blender a valid equivalent

After multiple tests yesterday I agree in the confusion and I am also confused. The theory is:

  • Scale: Indicates the size of the mesh
  • Radius: Indicates the radius from the point of incidence of light that each component is able to penetrate in the mesh.

In my tests done in 2.49 months ago it was working that way.

But in 2.5 I was needing to use radius of 2kilometers for a scale of 0.001 so I think there is a bug in the way radius shows us the units, instead of kilometers probably they must be called milimeters (inverted probably?) But I think it is better don’t even talk about this bug to developpers and wait for the rewritten of the render that is planned when they have time.

Yesterday I found that instead of keep a scale and then variate the radius it is easier in 2.5 keep the radius and then variate the scale. I keep radius in 4, 2 and 1 cm and then I chosen a scale of 0.02 for dermis and 0.4 for backscattered. More testing to do and I will post my result in Jose Conseco realistic head test thread.

Its even more confusing, Bao2
I used 2.49 , 2.5x
If you set up SSS in 2.49, works the same in 2.5 LOL

But all these numbers, well never mind, they are just numbers. Km etc is nonsense. I just put them anywhere I like. Its not rational. IOR close to 0.4 is not rational at all but it worked for me, especially when single layer-pass is used.
BTW
For tests, I strongly recommend all of you to use multiple single render outputs and combine them to Ps or gimp. This is easier to understand the principles. Then construct a similar process to nodes editor.
Its easier than having to wait all passes every time. Levels, curves, transparency, screen, color etc are our friends.

Results of my tests haven’t convinced me so far, keep trying. I didn’t like sintel’s SSS actually.

@michalis do you mind if other people post tests in this thread ( I might want to over the weekend or next week)

about the scale

the formula for scale is : scale = size of object in blender units/actual size of object in mm.

so if you model a scene with a man, a bunch of grapes, a glass of milk, a marble statue all in proportion to each other and to the same consistent scale than their sss shader will have the same scale factor.

to illustrate if you model to a scale of 1 BU (blender units) = 1m

a man of a height of 1760mm (5 feet 9inches) will be 1.76BU
if he is standing next to the Michelangelo’s David which is 5170mm tall (about 17feet) it will be 5.170BU

both will have the same SSS scale as scale = 5.170/5170=1.76/1760=0.001.

This next part I am not 100% sure off but I think that scale value is used to scale the values of the scattering radius which should be given in mm and will depend on the material you are simulating. I reasoned it this way if 1bu = 1mm you have a scale of 1. any value multiplied by 1 doesn’t change so in radius value will not be changed if but if scale is 0.001 (1 bu = 1m) than the are made smaller by a factor of 0.001. in essence they are scaled from mm to meters

I forgot to say that I have Color Management turned on. Without it on it renders exactly like 2.49 but with it on SSS comes crazy. I was testing the renderer in 2.5 and liked much more with Color Management on. But clearly that breaks tricks that worked in 2.49, like SSS. No problem, when developpers have time Ton already said that a rewritten of the renderer is in the TO DO list.

When light penetrates matter like skin, illuminates inside in a spheric way from the point of incidence. IOR I am sure determines if it is a sphere or a flattened sphere (less radius in depth than to the sides).

That is a good idea. What I was doing was rendering a little area of the image (using shift+B in camera view to select the area) so every test is 10 seconds.

It would be weird if it had ultrarealistic skin and the rest no. They said they were looking for the look the movie has, not photorrealism. I find the skin of Sintel fitted her in almost all the shots. My scene favourite is the argentinian team one with all that rain and Sintel struggling, and this scene has in my opinion a strange skin but I imagine other skin in that situation would not work.

@michalis do you mind if other people post tests in this thread ( I might want to over the weekend or next week)

Of course not, its the tests topic actually. My pleasure.
@Bao2
I meant all sintel movie characters, all this SSS, it looks like a rubber or other toy plastic material. If this is a kind of stylistic approach, I doubt it. They tried to keep it low via PP. See our new banner…

Im curious what exactly is sss, I have a vauge Idea of what it is, I think it stands for sub surface scattering but Im not sure.
If you guys already talked about this disregard this comment.

@billymuncher sss is subsurface scattering wikipedia has a nice article that explains what it is all about

Imagine you have in real life a cube of wood and one of wax or cheese, side by side. Imagine both cubes are lighted from above. In the wood one you can’t see light penetrating and lighting the sides, in the wax/cheese one you see the sides lighted because light penetrates the material and scatters inside it in a spheric way from the point where light enters. That is the subsurface (below the surface) scattering. This “buble” of light inside the mesh goes from maximum light in the point the right of light enters and goes decreasing in strength. Radius is the distance from the point of incidence of the ray to comlete fading. I was telling above that IOR probably indicated the shape of this sphere (really a hemisphere) but probably is more the speed that the ray strength fades, like a bias.

I find Sintel skin in many shoots a little waxier than human skin is. In the argentinian shot (the one with rain) I find the skin lacking a little wax. But I think the main problem is the dirt because is unnatural (perhaps the SSS destroyed it, eating the detail, and when you lack detail you obtain a CG look). But as I said if Sintel were with awesome skin it would be weird because the rest of materials in the movie would look not so awesome. I am sure they lowered the photorrealism at purpose (and also more easy to complete the movie).

Here are two captures with good SSS. I really like to know how they did it. Its a combination of lighting and AO that helps. I’m also testing the SSS using AO only. The area around eyelids, usually a cooler- grayish tone effect, here is OK but still visible. This area should be almost red.

Attachments


I find them “soft” like sintel skin is some high tech new rubber material. A real face in the position of Sintel in that image doesn’t look so soft and absorbing light. But it fits in my opinion much better than if they make a real face in that light condition. With that “soft” look it creates emotion in you about the place as a refuge from the wilderness she came. I don’t see problems.

About lacking colors it would be fixed coloring the texture used correcting the wrong tones, it is not about a SSS problem and easily fixable. But I don’t think anybody complain about coloration in Sintel face because it is obvious they didn’t go for photorrealism. I am sure they said to Ben Dansie many times to correct shaders and go for a more soft looking.

It will be interesting when the files be released in this scene to try modifications and see if they work or look worst than imagined they would look.

Bao2

that soft rubber look is quite often in some recent animation film
in particular when they shoot for a straight cartoon CG look.

works well to make everything I don’t know more child like.
reminds me about the soft focus / filter on photography

What do you mean by that? I think blender’s retopo tools are excellent. Are you using them correctly?

I think he might talk about the sketch tool we had once.

Otherwise besides automatic options I also feel that the
tool is pretty decent since it is like well a sketch on a surface tool.

Maybe he mean auto retopo.

Yes, I mean we almost have auto retopo, a good one. Shrinkwrap is so close…
BTW I don’t find blender retopo tools excellent. 3DC and topogun are excellent tools.

Shrink wrap has nothing to do with autoretopo(which means remeshing),it’s a way to project/snap points from a
surface to another,it works really well for reconstructing multires levels of subdivision but you have to do the mesh(actually it’s a pity that this modifier can handle only points,because it could be used to transfer uv coordinates,bone weights or vertex colors between meshes with different topology but same shape,the algorithm is more or less the same)

So topgun is basically like a auto face generator when you draw,
and a shrink wrap when u use the subdivision tool?