SSS scale for scattering medium

[Michael314]

Hello,
I played a bit with the SSS. For some textures and some lighting, I got too much red. I get acceptable results when raising the scale parameter to about 50 or higher.
As the recommended value is about 5 to 10 (as mentioned in Paul's blog about SSS), I wonder if the scale maybe is Poser display unit dependent. (My units are set to meter).

Best regards,
Michael

[mlru]

Poser's units don't matter. You have to find the correct physical dimesion for the effect or at least an approach.
I work with these settings (a liitle tweaked) that give a very good result under all light conditions:
http://render.otoy.com/forum/viewtopic. ... 8&start=36

The color is driven by a SSS-Map multiplied with a redish color in the transmission channel, and the depth of the effect comes from the scattering value multiplied with the scale, resulting in a 'real' depth of 9mm. (there's a thread around here, that explains that a bit further)

Uwe

[Michael314]

Hi Uwe,
Ok, so the depth of the scattering is then determined by the division of the scattering setting and the scale? I used 0.5 where you have 0.09.

I found the value of 50 through trial and error (best item to look at were the hands, I tuned the parameters until I got a slightly reddish effect on the fingers but not on the palm).
I read through your post and I agree that the color of "transmission" is crucial - with everything else same, this can give you anything from a yellow-ish ill-looking face through normal and towards a blushed appearance.
In the picture attached (DAZ Reby Sky model), I used a map for the transmission color (EC). I first tried to use the transmission strength maps (ES) as well, but then through that Octane would give me that though volumetric effects (material thickness), which I think is not available in the programs they made the maps for.

Right now I think the skin looks still a bit "dry" and uniform. A specularity map should help here, controlling specularity and roughness of the glossy skin component. While the Spec maps of most textures seem to be just a darkened version of the desaturized diffuse maps, RDNA's Vanilla Sky has some maps which could by suitable.

I have not yet tried to adjust the direction parameter, will do that later.

Best regards,
Michael

[face_off]

Ni Michael

Yes, scale * scattering is the absolute number - however I found the absorption is also important.

For your render - the scale looks too small - since there is no SSS around her nostrils or fingertips.

For specular - use filmwidth 1 and tweak to specular (to get the amount of reflection) and roughness (to get the texture). A turbulence node might give you best bump than the bump map.

Also, gamma 0.8, exposure 1.2.

There is a feature of the next release that will make this easier to tweak - you can display the previous render - so it makes it easy to compare results.

Paul

[mlru]

Hi Uwe,
Ok, so the depth of the scattering is then determined by the division of the scattering setting and the scale? I used 0.5 where you have 0.09.

I found the value of 50 through trial and error (best item to look at were the hands, I tuned the parameters until I got a slightly reddish effect on the fingers but not on the palm).
I read through your post and I agree that the color of "transmission" is crucial - with everything else same, this can give you anything from a yellow-ish ill-looking face through normal and towards a blushed appearance.
In the picture attached (DAZ Reby Sky model), I used a map for the transmission color (EC). I first tried to use the transmission strength maps (ES) as well, but then through that Octane would give me that though volumetric effects (material thickness), which I think is not available in the programs they made the maps for.

Michael, your values are way off.
The exact calculations and settings are explained here: http://render.otoy.com/forum/viewtopic. ... 80&p=67650

Without a decent setting of the transmission channel you won't get a nice SSS effect:
You'll need a grayscale SSS-Map (which can be very rough and lo-res, like 512X512) which defines what regions of the body are affected by the scattering (like Paul said: nostrils, ears, fingers). You can go without it, but then you'll get an overall waxy impression (some still might take for a wonderful skin, anyway ).
Multiply this with the color you need to have for the effect (I used a pale red, could be made something more saturated, probabliy)

The medium node tells how "deep" the light rays travel into the object, and how they get reflected. 9mm seem a reasonable value for me, though the effect gets stronger the darker the skin color is, so a fine tuning between 12mm (very pale skin) to 4mm (very dark skin) might be required. Skin sub surface doesn't reflect too much light in the direction it comes from, the main part goes through it (again, see the ears!), so the direction value has to go more to +1.

I have to disagree with Paul about the absoption. With the right transmission, scattering and directional values you don't need to touch the absobtion and can leave it to 0.0.

And, of course, you have to render in pathtracing or pmc

I've attached 3 examples with different diffuse and bump maps, all other settings like transmission, SSS or specularity are unchanged (note the impact of the effect):

[Michael314]

Hello,
happy New Year to everyone!

Thanks for the input. I've done some more tests, also with a new lighting setup.

@Uwe: Comparing your settings (0.09 * 1000 = 90) with mine (0.5 * 50 = 25), my scale was by a factor 3.5 lower than yours.

Here is a render with all light from the side, where I doubled my SSS scale (no EC maps used, just uniform red for "Transmission").

The effect can be noticed best around the nose. The intensity of the scattering effect is influenced by the nose thickness. If I increase the scale more, it suddenly vanishes completely.

The issue without some transmission map is the unwanted scattering of the brows, changing their color from black to red/brown.

Best regards,
Michael

[Michael314]

Ni Michael

Yes, scale * scattering is the absolute number - however I found the absorption is also important.

For your render - the scale looks too small - since there is no SSS around her nostrils or fingertips.

Hi Paul,
I don't understand this. You say to increase the SSS effect, but if my scale is too small and I increase it, wouldn't that reduce the effect?

Also, gamma 0.8, exposure 1.2.

are these settings for the "imager" or a specific texture? For a test, I changed this in the imager settings, and indeed the picture looked better.
(Unfortunately I currently cannot upload images, I'm getting error messages from the server).

Best regards,
Michael

[face_off]

I don't understand this. You say to increase the SSS effect, but if my scale is too small and I increase it, wouldn't that reduce the effect?

Mmmmm, I have a slightly different setup to Uwe, where I have an equal or higher absorption to scattering - so increasing the scale from 5 to 10 has the effect of a net increase in absorption, which means the SSS effect is reduced.

Whilst my method looks good - IMHO it seems to go against what other people have done. So listening to both me and Uwe will confuse you. I suggest listening to Uwe, since his method is based more on fact than intuition

I derived my setup using a cube with a lighting moved to various angles - so I could see the SSS effect I wanted at the edges of the cube. My conclusion was that it was the ratio of the scattering verses absorption that was key - but these goes against what others are saying - so more than happy to be proven wrong.

[face_off]

I just had a few mins to play with this again.

Absorption is absolutely key - it needs to be non-zero, otherwise the sss leaks to all parts of the mesh. Attached image shows 0 absorption - you can see the whole box is tinted red (undesirable for skin).

[face_off]

Now add some absorption, and you still get the sss effect at the edges, but a neutral colour elsewhere, which is what you want for skin.

NOTE: Both these render give the same result, but one has sss and absorption halved and scale doubled. So you can see how I came to the conclusion that it was the ratio of the scattering to absorption * scale that dictated the sss effect.