Very interesting Lele!
I’m not sure to get the whole process right, would it be possible to have a few screens of the process you describe.
Mainly for this part : “GanzFeld lighting (Spherical IBL with 1.0F white color, in modern CG lingo)”
Cheers
Stan
A screenshot of the UI, with the render under lighting conditions (the dome settings are shown on the right. rest is at default.)
While the render looks very uninteresting, that’s the whole point of the exercise, as the render elements reveal (diffuse filter, reflection and specular, in order.):
Here is the render with an HDRI texture in the dome.
Looks great, but I’m not sure to understand how this is any different process than loading a HDRI in a VrayDome and do the shaders as usual.
Am I missing something?
Stan
It’s simple, Lele is simply channelling his extrasensory perception to accurately calibrate his shaders. Don’t you do this too?!
https://en.wikipedia.org/?title=Ganzfeld\_experiment
A ganzfeld experiment (from the German for “entire field”) is a technique used in parapsychology which claims to be able to test individuals for extrasensory perception (ESP). The ganzfeld experiments are among the most recent in parapsychology for testing telepathy.[[1]](Ganzfeld experiment - Wikipedia)
Eheh, GanzFeld simply means “total field” in German.
So it’s used in a number of fields, psychology for one.
The substantive however, is “Lighting”.
The difference is very simple: your L term in the shading pipeline is always 1.0, making shader evaluation an exact science, something which is hard to impossible to do with an arbitrary (if calibrated), and varying, HDRI dome.
Render elements do come to the aid, but they won’t return exactly the same results as this method does (for instance, the specular term in the example above contains both the specular amount, and its glossiness, in one, across the whole object domain, and not restricted to lit areas.).
here’s where i got the initial idea from (Ch. 9.1.2):
Lele just uses a dome light in full spherical mode with a white diffuse as his starting test point cos he’s a purist. I think he’s 1/8 swiss? He authors his shaders in a white environment first and then moves them into hdri’s to see how they behave. Only bad point is you can’t shoot reference materials in such light unless you’ve a fancy ganzfield room. You also have the worry that if you do have a pure white room to shoot things in, you might be a figment of Stanley Kubrick’s imagination.
Fair points, John.
However, some reference comes with precise color values and albedo values, however those were measured in the first place.
Then figuring out if the shader is outputting those without the 1/8th swiss in me becomes difficult allright.
In other words, it mimics something impossible to shoot in RL (you’ll always have something suspending your object, and shadowing it), but it provides exceedingly accurate shader measurements.
When those are clear, the dropping of an HDRI into the dome (especially if calibrated and well known) hardly ever presents surprises, tbh.
Yep - I’d much rather measured stuff as a starting point - I’m kind of excited for the quixel megascans database, if not for the actual textures but more as a reference database of albedo and specular values for a range of different surfaces. I’ve got a monitor calibrator here that also works as a spectrophotometer so I can get fairly accurate rgb values for a surface. I suspect it doesn’t work well on bumpy or curved surfaces though and I’m not sure how to estimate the reflectivity that I have to remove from the data but it’s possible better than an educated guess!
Right, so it’s basically go back to what we did 8 years ago, making a shader based on a white dome (ok it’s spherical and not half dome), in stead of a HDRI.
This will prevent you to push reflection or whatever setting too far as a HDRI could miss lead you.
Is this a revolution or am I really missing something here? 
Stan
I’m sorry, this is probably a dumb question. But why does Lele’s screenshot show a texture in the Vray Light if the procedure supposedly uses only a solid white light?
Is it that he toggles the texture on and off?
I guessed it got there by mistake, the fist image should have been a grab of the white environment. might be terribly wrong though.
Yes, it’s to show how easy the process of switching to a textured environment is.
You will notice in both cases intensity for the light is at 1.0
EDIT: oh silly me, Rivoli’s right. it ought to have been off, the render showing the ganzfeld result.
Sorry, they were very busy days.
As for how valid the method is in evaluating surfaces, i’ll just leave this here:
http://gl.ict.usc.edu/Research/DigitalEmily2/images/MixedPattern.jpg
EDIT:
You will notice the first camera to the left of the bottom row has a somewhat passing resemblance to this crop:
Once properly balanced, the shader reacts like this to different lighting conditions:
The file’s still very much work in progress, and I am not yet matched 100% to the provided references (something on the cards today. it’s somewhat convoluted, although i feel the data is all there, albeit with one potential ingredient being amiss and having been established somewhat arbitrarily.), but the shader ranginess is around about correct, i feel.
I second pretty much what lele is saying.
To extend the talk, since we almost never have calibrated refs here for what we usually do, i introduced in the lookdev stage the use of a hdr that tries to simulated a cg overcast environment, for example a scanning room. It’s not “scientifically correct” by any means, but it gives a bit more help from an artistic point of view especially when dialling diffuse and reflection contribution.
Like THIS from ten24, then desaturated and pumped to get a 0,5 diffuse contribution on a 50% grey lambert sphere in the highest part of the hdr. Doing so i found my assets to be quite good in almost any kind of lighting condition afterward.
I’m glad I started this thread really. Very interesting approaches overall.
A bit further down the road of balancing the shader (and textures. and removing color cast in the refs. and slimming shader layers down.).
The rendered SSS matches quite closely the control images (under the same lighting).
Specular needs likely a tiny bit more work, but before you ask, yes, she IS that shiny in the references (my first attempts were in fact quite flattering…).
I guess once the “zero” is prepared, I’ll take artistic liberties and dial back down all that oiliness.
Notice the lighting in the first two images remained the same as the previous attempts, so any change is to be imputed to the shader.
A contact sheet with the references and corresponding renders (latest set) of Emily.
Notice how the one bit that I couldn’t nail down in numbers is the one bit obviously singing outside of the choir (namely, the specular component.).
Renders in more pleasant and interesting lighting conditions will follow shortly, while the max file and documentation will follow later still.
Great stuff, I’ll look forward to digging around in your scene.
This is great Lele! Thanks…
ps. I couldn’t download the full image to view in 1:1 locally with full screen. Anyone knows how to? :???:
umh. can’t be done, unfortunately.
Will upload it along with the EXR version of it on google drive and link it.
Gimme some time as i want to add a few of the bigger sized crops with more interesting lighting conditions, as well.
