Update v1.4

-----------------------------------

Rayleigh Scattering

We've reparameterized it and we're using it to introduce a red tint super usefull for skin rendering for example. You can use this for simulation or to just tweak your SSS color without altering its albedo map or even as an extra parameter to tweak overall SSS scattering.

Mie Scattering

The Mie phase function describes the complex shapes that arise when light is scattered by water droplets. We use a blend of HG and Draine's phase function that can accurately match 95% of the Mie phase function. This is way more physically correct than the double HG phase function introduced lately in Arnold7. It is engaged with high anisotropy and an ad-hoc parameter called 'Mie Scattering' and work just great with the latest 'Similarity Theory' addition you can read just below. Works also with romboSSSRefract.

Similarity Theory

here is used to reduce the overall time spent sampling highly anisotropic SSS. It is completely automatic and does not require any user invervention. It works only when the Anisotropy parameter in the SSS randomwalk shader is tweaked. It can reduce overal render time by 20% while having the same visual results as before.

Albedo Remapping

The one that Arnold and our fist SSS incarnation use is based on the Disney model. However this remap does not support Anisotropy... meaning that as soon as we increase the Anisotropy parameter we start losing the intuitive feedback we had before with just Color and Radius. With latest rombotools update we introduce a couple more models that are able to keep anisotropy involved when converting Color/Radius to the underlying volumetric parameters needed for subsurface scattering. 

Diffuse Contrast

is a new parameter to directly control the SSS lambertian light terminator to better match real world materials like plastic. It stays on the SSS lobe. With the default 1 we have the same exact behaviour as before. 

Indirect Subsurface

With this latest rombotools update we introduce a consistent stochastic strategy that smoothly degrades SSS to a lambertian lookup that is perceptually equivalent to the reference full SSS. For a fraction of time. This will lead also way better solution quality, much more uniform over the whole image without patches of raw SSS where SSS is indirectly sampled.

Continuation Rays

We borrow the 'continuation rays' terminology from PRMAN where rays that hit a front-facing poly or a different geometry than the one from where the rays originate are continued like no intersect was found effectively minimizing the troubles of needing an enclosed manifold for proper volumetric random walk. Below the cones are simply added into the volume and not 'intersected' with the volume.. so we have extraneous geometry in the volume generating that dark silhouette.

Overall Speedup

Now beside all the nice additions we're also way faster than stock Arnold SSS. For the same exact settings we're already 2x faster. Then look at indirect SSS solution.. you see there we've the same quality we've on the direct SSS, while Arnold is struggling to get convergence on those zones. Also take a look at the left angel wing in direct light.. it's thin and light passes side by side.. Arnold is full of outliers, rombo is smooth.

....

New Scenes

Houdini

Cinema4D

Maya

3ds Max