The disk-integration technique won't be any use unless nick implements the point-based method for mutliple-scattering, which I don't really see how well it would fit with MLT to be honest.

The various parts of the model are controlled via texture maps, i.e. there is a diffuse map, and then a reflection map and transmittance map for each layer to define the amount of reflection/transmittance, so this isn't an SSS procedural it's SSS using maps. Actually I don't think that's correct at...

... and I think it was implemented in finalRender long before fryRender too

Do you need to understand the file for that? Anyway, I found the information somewhere on the internet (can't remember where now). Basically it goes like this. The first line contains information about the data in the file. All the other lines are the n and k coefficients, pairwise. on the first lin...

The paper uses an approximation to unbiased SSS, but the approximation is reasonably accurate, so the rendering methods are actually very similar. I was talking about the way you handle light transmission at the air/skin interface more than the scattering behaviour once light has entered the skin :)

About 1.4 i think. Greasy water basically

Remember that the paper's using a very different rendering method. In theory everything should work the same (just mlt will be a lot slower), but there will be some differences in look due to the way the materials are set up. ryjo, it might be a better idea to use a 3-point lighting setup with a hot...

What would be really cool is if you had direct rendering to that gui with an image catalogue and some other basic image processing operations (exposure, contrast etc)... oh and do it on the gpu too!

Multiply by 1000 to put in units of m^-1 which indigo uses. by 'g', i mean the asymmetry parameter of the Henyey-Greenstein phase function, e.g. amount of forwards or back scattering. But perhaps these are all uniformly scattering media? Jensen states in the paper that the dipole approximation assu...

I believe they're using the technique described in GPU Gems II. Pixar recently implemented this in prman for production-quality rendering, and it is very, very fast. Not accurate at all tho :) That second video is from a paper at last year's siggraph. It's not really GI, more a clever way of doing a...

Could this not be done trivially in a shader within indigo? It would save the hassle of having to preconvert your textures all the time.

Fireflies are just an artifact of monte carlo sampling. Where you get a high value from a direction with a low pdf you'll get a firefly... There was a solution posted here a while ago: render two images, then combine them with a "min" operation, i.e. for each pixel take the minimum value from the tw...

An even better solution would be to have a display driver interface. Then each platform could have its own implementation that would be loaded as a plugin...

If you want any help with implementing that or general os-independent porting, nick, just ask....

Have you modelled the windshield as it is in the real world? i.e. with thickness? I believe it's the double refractio that gives you a "normal" image through the glass.

Metropolis sampling is a bit of a pig.

If you want a good grounding learn path tracing first. The best reference I know of is this book: www.pbrt.org