Fiddling with Iray skin settings...

AndyGrimm

and by the way - a glossiness map is painted ONCE for a UV set...  the next one changes just slightly...10 shades of grey.. that is all what is needed... 

the BUmp is the real Workload.. people expect now CLOSE UPS....

Albedo can be easy made from a 3d delight texture...   smootjing and bluring ....
Translucency is more or less the albedo... whitout blurring...

and there is not much more for PBR.. all the rest.. of maps (specualr, metalness and what else i got in iray produtcs.. are well - at best a sign that the author did not know that he works on skin 

j cade

@jag11 Nice! I can't decide between 1 & 2. I also owe you a big thanks, After you mentioned using white for translucency color I started expirimenting and my method now involves very very pale pinks in translucency color it was the only way I was getting any of the effect at all. You're method is probably more easily controlable than mine, I've got it but its very hard changing how strong it is in relation to other things. Mine does have the advantage of no extra geometry though which means its more likely to fit on my puny gpu. But like I said more than one way to skin a cat.

j cade

AndyGrimm said:

@j cade @jag11

the idea is great.... BUT you try now to create a soft shadow whith a light which WILL draw a hard shadiow... 

there is no light other then a small flashlight with a very strong LED which can produce this light...

use a normal sized lamp source instead... a photolamp or led pattern....   10 cm circle.

for best simulation use a IES profile of a real lamp, 

And we need to know the distance from the light to the nose.. this has a BIG impact too.
 

I think you and I might be misunderstanding each other a little, the point of the tiny light is to isolate a very specific feature of realistic skin. the very tiny size of the light is to isolate all the other variables, to show the SSS caused by the light entering into the skin and bouncing back out (rather than the sss of the ears which is mostly caused by light passing all the way through the ears from behind).

The problem is that since the effect I'm trying to work on mostly involves skin with light hitting it its not directly evident like the backscattering of the ears is. The best wat to detect it is to have a relatively hard shadow and look for the red bleed into the shadow. In your image the shadow is to soft and small so I can't tell if there is any bleed.

 

(for some realworld testing I stuck my camera flashlight in a such that the side of my thumb cast a shadow on my palm,  and then looked at the edge of the shadow, the light was bleeding under the skin yellow for about 1 mm and then red for about 2 before turning to neutral shadow (if that makes any sense) /

 

I did your second image settings with the light at 10 diameter and 10000 lumens. I couldnt make heads or tail of the location you put, it didnt include the z axis and was pointing at her knees so I placed it such that the light was pointing at her head and looking through the light the boundry was at her nipple.

Arnold C

Rashad Carter said:

3. Iray is perhaps limited, but it is most certainly convoluted compared to Octane. Why not simply name the Slider Absorption if that is what it does, instead of naming it Transmitted Color?  Why not call Scattering what it is instead of naming it....uh oh, I still do not know what the Iray equivalent of Scattering is. Help please? They should rename things in a simpler manner.

... That said, over the past few pages you guys are beginning to find the proper slots and the materials you are setting up are begining to look more and more like proper skin set-ups in Octane. So bravo! Assuming this "convergence" will continue, I will make a few predictions and observations of what is yet to come.

Here is what you guys are probably still getting wrong.

1. Do not waste your time trying to model each layer of skin individually. That is quite silly to my thinking as I've seen it done in Octane forums too and the results do not warrant such headache. The simple fact is that the individual layers of skin are so thin... it is perhaps naive to assume that any one effect such as backscatter only occurs at a certain layer of skin. Absorptions and scattering occur at all depths. One needs to average the inputs of the three layers and threat them as a single transmissive substance which has increasing absorption with depth. I will likely explain this more if people are curious.

2. You do not need Geoshells. Sorry, but no way am I doubling my polygon overhead. As much improved as these currect settings from Jag are, they are still missing important elements if geoshells are needed.

Conclusion:

We're getting there. And like most things in physics, when you get it right your solutions will be elegant, and robust, and surprising how much complexity can arise from such simplicity.

I'm off to test more skin in Iray. When I start getting results that rival current results from Octane, I will surely post them and the settings i've used. Until then, please do keep up the testing and discussions. Keep going around in circles and I'll keep yelling at the screen. For all the hair pulling, this is still quite a good bit of fun to discuss and decipher. Don't you agree?

Nice speach, Rashad.

In answer to 3.: Well, the naming of that parameter "Transmitted Color" for the Transmission group is quite correct/more precise. The value set in each RGB channel is the amount of light transmitted through the the volume per distance unit defined with Transmitted Measurement Distance (TMD), only the remainder, the difference to 1.0, is being absorbed.

Scattering is named "SSS Amount", and that, too, is also quite correct. The value set for SSS Amount is the amount of light getting scattered before it reaches the distance defined with Scattered Measurement Distance (SMD).

A big drawback is the lack of a proper documentation. Why doesn't it say for example that the "Refraction Index" doesn't do anything unless you set "Refraction Weight" to >0?! So many skin shader setups, even the new Generation 7 "DAZ Originals'" have one set. Makes me argue, beside other glitches, that even the people at DAZ don't exactly kow, how their new renderer works.

I totally agree. Over the last few days/weeks we discovered a lot of the tiny bits and what works and what not... and how it works. That's the good thing when more than just one head is involved, different ideas, different questions asked, different solutions presented...

I somewhat disagree, but I see your point. If a certain effect in a certain layer is somewhat very faint for the human eye, the effort you had put in to get there really wouldn't be worth it. And with the Iray Uber you can't model each skin layer seperately anyways. You'd need to construct a completely new shader setup from scratch. And then it's talking again about the effort... And regarding "backscatter", on human skin it really only occurs at a certain layer of skin, if it wouldn't be for the hypodermis, sometimes called subcutis, especially UV light would travel deeper into the body.

The geoshell approach sounded interesting, but it really is hard to make it work right/without glitches. On the most figures I tested it on, you have poke throughs in some places with certain expressions, eyes closed... not critical for mid and far range renders, but at close-ups it really can drive you mad. Turning the offset screw doesn't help, the difference is choosing between bad and really worse. I guess it simply haven't been designed to be used for someting inside a figure in the first place...

I agree...

j cade

God, If I could make heads or tails of the shader mixer for Iray, but man it is not easy to navigate. I have poked around, but  AAAHHHH!

For instance I could only find one block for fresnel, but it only connected to one specific type of material. And all the st's bf's st's ef's :(((. All in all it makes me run to the comforts of blender's material nodes.

Arnold C

 

AndyGrimm said:

translucency is not just set with transluceny weight... the transmitted color in the sss affects translucency way more ....

Absolutely. "Translucency Weight" simply determines, what amount of your "Transmitted Color" is layered (laid) over your "Base Color". They don't mix, as one might assume. At lower values the transmission tints the base color, and at TW 1.0 it replaces the base color completely.

It's the other way around, Andy. The translucency color value for each RGB affects the corresponding transmission RGB color value. It's a modifier, a multiplicator, or, simply put, it determines the "strength" of a transmission color RGB value. What we see in our renders isn't the Translucency Color, we see the endresult of our Transmitted Color when the Translucency Color is done meddling with it.

AndyGrimm said:

we can not get blue out of pure red...  .. the translucency color is best set to a brown/beige (or map) which MUST include R, G and B... 

Affirmative. A 1.0, 0.0, 0.0 for "Translucency Color" will kill any value in the GB color channel of the "Transmitted Color". All what's left is the red portion, and that's the reason all skin settings using this method turn out to look too reddish. And then they use SSS Tint to pull the too much red. *Sense... this method doesn't make any!*

Viewing some translucency and transmission color settings in the Aux Viewport using the NVIDIA renderer preview, transmission only, scattering and reflections are set to off. Test object is a geoshell I had already in the scene:

1. Yellow Base Color, pure red translucency color, purpleish transmission. Translucency Weight at 1.0: modified transmission kills (replaces) the diffuse (Base Color).

2. Same colors, Weight at 0.75: modified transmission tints the diffuse.

3.  Same colors, Weight at 0.50: modified transmission tints the diffuse, less strong.

4.  Same colors except Translucency Color (pure white), Weight at 1.0: unmodified transmission replaces the diffuse.

5. White Base Color on Face, Lips, Nostrils, everywhere else 0.5 grey, pure white Translucency Color + diffuse texture map, pure white transmission. Translucency Weight at 0.50: modified transmission (by the texture) tints the diffuse.

6. Same setup, Weight at 1.0: modified transmission (by the texture) replaces the diffuse, texture details are visible, but kinda blurry. If it would be the Translucency Color in the render we see, the texture would appear to be more sharp, like if it would have been put into the Base Color.

evilded777

Khory said:

mjc1016 said:

Rashad Carter said:

Khory said:

Actually Mjc if you look at current ideals for the base map you will see that they are vastly different than the sort of map that is used for 3dl. So different that they really make photoreferences pointless since they are often blured and the colors are altered from the initial resource.

I could not agree more. Photo-references are over-rated. PEople are tricked into thinking that a skin that looks "real" beore rendering will still look realistic after rendering. They are wrong. The fact is the diffuse map shopuld be somewhat incomplete...this leaves room for the render engine to complete the task. If the skin source is already too "realistic from baked in AO, Specular, and other effects, then the rendering engine will only detract from that. Skin should not look "real" until the render has completed.

And I said specifically PROPER maps are renderer neutral...who said anything about the photoreference maps actually being proper for 3DL any more than they are for Iray?

 

PA's don't sell to themselves so they spend a fair bit of time meeting the expectations of the customers. That does not always mean what they produce is "perfect" or "correct" or even realistic. Why? Because people want their expectations and their beliefs met. Look at the sort of flooring that people love best. It is almost always has over the top reflection. The PA may know that reflection is too strong but they also have to be aware of what the customer will perceive as correct because in their head that is the expectation they want met. So "proper" maps may not in fact create a product that will sell. At this point if I said I had carefully hand painted a texture for V7 that looked lovely and natural there is a fairly good chance it would be over looked because hand painting is textures is considered "old school" in this market. Simply put, customers look down their noses at anything not based on photo resources.

Once you are locked into using photo resources by the market then your not going to have the opportunity to do maps that are directly driven by the texture your painting but rather that your going to to have to create all the support maps to match the photo rather than creating them as the skin texture is painted. Is it possible that there are right thinking customers out there who would buy a painted rather than photo based figure? Sure. At least a couple of us would. But I don't know a single character creator who can afford to take risks financially. Going with what most customers would perceive as "old" methodology would be such a risk.

Don't forget that most people will never do any real research themselves as far as PBR rendering. They don't have time, or are not information geeks, or simply wouldn't know where to start looking. And that is alright. We don't all need to be research geeks. Being here just for the art is always an acceptable position to take. Because of that their expectations may not shift away from photo resourced characters for a while. At some point that will start to change I am sure but I don't think we are at that point yet. On top of that learning to paint a full texture in zbrush is going to add education time to any one considering doing characters that way. More over since map creation would be heavily driven by PBR concepts then all the core knowledge for 3dl would also have to be adjusted and tested. Yet more risk for the content creators. And on top of that then everything that the customers know about what will happen when they render a character in 3dl will be subtly shifted. That might or might not be a "thing". But again it holds potential risk for the content creators.

It is always easy for us to armchair quarter back on subjects that we don't have a financial interest in. If you asked my friends who are PA's you would know that I am one of the pickiest most difficult people about well.. details. And I confess that I question surface settings more than any normal person should. In my defense, I almost always question my surface settings at some point after I do the product so I am not just picking on everyone else. But I am also aware that there are many factor that influence how products are created. I'm also aware that there is not really any perfection because people often have a different ideal about what perfection really is.

I'm gonna be quiet and do my best to retract my earlier remarks because they were uncalled for and well, perhaps unfair. Khory makes very solid points here and I have the utmost respet for her and a significant number of vendors. Lets be careful how we say what we say, lest we be unintentionally insulting to people who work very hard.

Richard Haseltine

Let's keep this to a discussion of skin settings and not have it dengenerate into insults please.

fastbike1

Take this for what it's worth, but I feel V3Digitimes offered some real, practical wisdom that seems to have been overlooked by the "rush to science". That was "Anyway what is important to understand is that Scattering in DS cannot be computed "like in real world". Scattering computation will use models which will try to mimic the best reality, and even more, these models will be simplified in term of sampling (color / spatial / type of scattering), to allow tolerable rendering durations."

Since my profession requires complex logic models to model real world situations, I have often felt it necessary to remind less experieced colleagues of a simple truth similar to V3Digitimes statement; No matter how complex the model, the real world is more complex still.

One must remember that for any model to be truly useful, practical considerations of time and technology require (for now at least) some simplfying assumptions or choices. This is the Art that goes with Science.

evilded777

fastbike1 said:

Take this for what it's worth, but I feel V3Digitimes offered some real, practical wisdom that seems to have been overlooked by the "rush to science". That was "Anyway what is important to understand is that Scattering in DS cannot be computed "like in real world". Scattering computation will use models which will try to mimic the best reality, and even more, these models will be simplified in term of sampling (color / spatial / type of scattering), to allow tolerable rendering durations."

Since my profession requires complex logic models to model real world situations, I have often felt it necessary to remind less experieced colleagues of a simple truth similar to V3Digitimes statement; No matter how complex the model, the real world is more complex still.

One must remember that for any model to be truly useful, practical considerations of time and technology require (for now at least) some simplfying assumptions or choices. This is the Art that goes with Science.

+1, wish this forum had an actual method for some sort of bumps like this

AndyGrimm

@j cade

well.. yes.. i was not clear enough..... 

what i tried to say... to see how the light reacts and colors the falloff of the shadow redish... you need another light.. because 1500 lumen directly on the nose with a flashligth would color the whole shadow red.... make the spot distance bigger and just the falloff is affected....

Now if you just adjust translucency and sss direction .(to a unkown flashlight distance)... you cheat a real setup...

It must also not be a pocket flaslight - it has to do with the strength of the light and environment brightness... so it makes more sense - to first test the skin.. like i suggested (the skin reacts to a real setup and skyllight and transluency looks "real")....and then fine tune falloff in the same setup
Your 10000 lumen render looks very good! that render says way more to me how well your skinmodel reacts with light and translucency then your test...

now.. just go down or up with lumen and in reality you should see somewhere your black to red fallow (allreay well set in your case)... in a setup which we all can replicate...

Oso3D

I'll point out that a lot of people don't have smooth, luminous skin, and that some of the dry 'bad' skins people have pointed out look very realistic and appropriate, to me.

 

AndyGrimm

@timmins.william

At the end it is about taste....    BUT as a ex partner and owner of a advertising and photostudio agency.. i can tell you.. that i never saw a nose on a overexposured photo with light from a side which is not translucent! (autobracket tests - skin glossines flashbulb tests)... same to glossiness... some skins are dryer.. BUT all are in the range of 0.7 - 0.8 glossiness on the nose in PBR specular...

measured tests which you can read.. sigraph 2006 as example.. include all types and ages of caucasian skin....  

 

V3Digitimes

AndyGrimm said:

@timmins.william

At the end it is about taste....    BUT as a ex partner and owner of a advertising and photostudio agency.. i can tell you.. that i never saw a nose on a overexposured photo with light from a side which is not translucent! (autobracket tests - skin glossines flashbulb tests)... same to glossiness... some skins are dryer.. BUT all are in the range of 0.7 - 0.8 glossiness on the nose in PBR specular...

measured tests which you can read.. sigraph 2006 as example.. include all types and ages of caucasian skin....  

 

For the nose, same at rendering, slightly showing translucency with strong exposures... But by curiosity, 0.7/0.8 what unit? Apples? Potatoes?

AndyGrimm

from a photo - i  did alot postwork for pre-print and digital media... we had to test CMYK... splitting a image in each channel and adjust... there is ALWAYS more red in the shadow of a nose.. actually that was one of the mainpoints what we used to adjust a photo so that it looks real in print and digital space....

AndyGrimm

@V3Digitimes
photoreference and a decade experience starring on noses  - and knowing that the range and effect of different skins is actually very small

the real differences makes a make up artist and post work - not the skin 

KurzonDax

@AndyGrimm

Do you have a link to the siggraph paper?  I looked around on the siggraph 2006 website, but wasn't sure which you might be referring to.  

Edit to say I don't mind paying for it, but want to make sure I'm getting the right one.

V3Digitimes

Ok this is very interesting as an information. I should MP you about a specific need I have.

So 0.7-0.8 is the Lightness or Value of the highlight part of the nose on a photograph, is that what you meant?

AndyGrimm

i can just point out again - that  how a real or at best as possible simulated "naked" skin - migh not be what you have in mind as final result on your images... makeup powder which you see on ALL model photos including postwork...  changes real skin glossiness and wet(dry look EXTREM...

V3Digitimes

AndyGrimm said:

@V3Digitimes
photoreference and a decade experience starring on noses  - and knowing that the range and effect of different skins is actually very small

the real differences makes a make up artist and post work - not the skin 

Yes Ansiko who is also a movie director told me the same. Powder is the key (with using huuuuuuuuuuge flat diffusive reflector pannels). Except that in the movies the postwork is.. more difficult.

AndyGrimm

V3Digitimes said:

Ok this is very interesting as an information. I should MP you about a specific need I have.

So 0.7-0.8 is the Lightness or Value of the highlight part of the nose on a photograph, is that what you meant?

 this is what i set as the highest glossiness value in Iray (glossiness weight 1, glossiness 0.8).....    for the nose....  the start point for a real glossiness map in iray.... all the rest in the face... is somehwere between 0.6 - 0.8

AndyGrimm

i referenced this document...
http://www.merl.com/publications/docs/TR2006-071.pdf

KurzonDax

AndyGrimm said:

@V3Digitimes
photoreference and a decade experience starring on noses  - and knowing that the range and effect of different skins is actually very small

the real differences makes a make up artist and post work - not the skin 

I think this is where the challenge is for most people that use D|S. Without the deep experience of knowing how untouched skin should look through a camera, most people only know what they see in a magazine or online which has always been postworked and includes fair amount of make-up, and thus that is what they strive for.  Searching for untouched, raw images yeilds some pretty wide ranging results, most of which is not good at all for using as a reference.  

KurzonDax

AndyGrimm said:

i referenced this document...
http://www.merl.com/publications/docs/TR2006-071.pdf

Awesome, thank you very much.

AndyGrimm

KurzonDax said:

AndyGrimm said:

@V3Digitimes
photoreference and a decade experience starring on noses  - and knowing that the range and effect of different skins is actually very small

the real differences makes a make up artist and post work - not the skin 

I think this is where the challenge is for most people that use D|S. Without the deep experience of knowing how untouched skin should look through a camera, most people only know what they see in a magazine or online which has always been postworked and includes fair amount of make-up, and thus that is what they strive for.  Searching for untouched, raw images yeilds some pretty wide ranging results, most of which is not good at all for using as a reference.  

right - you can check daz gallerie and how many people like a render - the real, looking  skin ones.. dont get much likes .. except a very old face with impressive wrinkes 

AndyGrimm

yep - i search since days for a high resolution image of a bautyfull womens natural lips whitout botox and makeup ...  very hard to find - also not on photostock pages 

AndyGrimm

i seriosly started to think that i must hire a model myself and shoot the needed images which i have in mind as reference and for my skintexture of a natural beauty myself  

AndyGrimm

oh - and i just forgotten.... real values work only with a bump which HAS microdetails...   taking the bump frum a photo in 4k - does not work.. because it is way to distorted for bringing it in the render with the needed strenght for affecting the glossines and the pores are to big because of inbaked backscattering, shadows and sss... that's why all sience skin renders work with 8k and higher resolution for normal maps on the face...AND overlay it with a real microstructure which shows the "roughnes" ON THE PORES ....

i get around this problem simulating this with a bump pattern grid also in 4k... not real.. but more real then whitout 

 

AndyGrimm

@V3Digitimes and just in case you have a direct line to Nvida...

tell them the simplest parameter which is missed in Iray is ROUGHNESS SCALE...  that's the missing thing for realistic glossiness control. and would eliminate many other maps.

KurzonDax

AndyGrimm said:

oh - and i just forgotten.... real values work only with a bump which HAS microdetails...   taking the bump frum a photo in 4k - does not work.. because it is way to distorted for bringing it in the render with the needed strenght for affecting the glossines and the pores are to big because of inbaked shadows and sss... that's why all sience skin renders work with 8k and higher resolution for normal maps on the face...AND overlay it with a real microstructure which shows the "roughnes" ON THE PORES ....

i get around this üroblem simulating this with a bump pattern grid also in 4k... not real.. but more real then whitout 

 

And don't forget about the texture compression that D|S/Iray does as well (advanced render settings)