Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
a) EJ's Billie looks like that is a matte rubber
b)EJ's Celine has freckles much too dense, look symmetric, size too uniform and too regular, and too brown-grey
c) EJ's Celine with Make-up, well, the make-up looks good (to this non-makeup apply observer) and it makes the unrealistic freckles look more realistic, for instance that have the light red-green color that is more natural (the color they appear depends on the light, surrounding skin-tone, and such. Mine have appeared green (that is light greenish-burnt ochre) to me and they've appeared red (that is light orangish color) and when really multiplied, grown, and darker from the sun they will appear orangish-brown. They fade away, especially on the face, after I get out of the sun and are mostly not on the face except faintly compared to childhood. I imagine different people's behave dfferently, this is something you should look up references to like you did for the eyes.
A) Could you be more specific about the rubber-y look? Is it the color, the shine, or what? The SSS, specularity, and bump are all exactly the same as for my lighter skinned characters, so it's probably due to the heightened contrast between the specularity and the base color. I was told years ago by Bagginsbill, a Poser shader guru, that the specular reflection on human skin remains constant and does not change proportionally with the underlying complexion. That's really the only reaosn I don't dampen the specularity in proportion to the procedural darkening of the base texture.
B & C) I actually have looked at reference photos for the freckles, and there are at least a few where they actually appear that dense, or at least quite nearly so. For example...
Of course, some of individual spots are a bit larger in that second, one which may make them look a bit denser than they actually are. Still, if I'm outside the range of normal human variation, I'm not so sure if it's by as wide a margin as you might think. Nevertheless, I'll revisit my procedural freckle parameters.
Then again, maybe at the time, I was so focused on trying to get the color right (particularly the saturation) that I accidentally ended up exaggerating some aspects.
What do you think of this?
The 1st one it is both the color & the shine. The skin looks actually dark grey or black which in real life it isn't really.
The freckles:
a) Notice that they are all different sizes and shapes.
b) Notice they overlap
c) Notice overlapping ones get bigger
d) Notice the color is more orange-red then the color you used earlier.
Freckles will take a lot of patience to do like those on those pictures you posted.
a) The boy has been run though a post filter - his white skin with freckles that dark would be red with sunburn. The freckles look darkened. The eyes made bluer. And his eyebrows & skin were made browner.
b) The girl picture looks unretouched except for the eyebrows.
Well the color is better but with that many freckles you'd expect some of them to overlap and be darker and bigger than other freckles. They all look very close to uniform in size and distribution on the face is oddly nearly equidistant.
Really? If anything, it looks ever-so-slightly purple-ish to my eye, which I think I've actually seen in some photos, such as this one.
OK, I looked again, there is the slightest red-brown tint in the skin on your earlier post but the problem is that that dark skin accentuates the very regular pattern of the model's skin. Now I am not sure what those bumps are supposed to be, sweat pores probably but such pores are not so regular on a person's skin and the pattern changes and it changes according to forehead, chin, nose, cheeks, eyelids, and so on. They being so regular helps make the skin look like something man-made like rubber. So, in a way it's sort of the same 'man-like too uniformly distributed' look I mentioned that you had a problem with on the freckles too. You need skin references to use for different ethnicities, genders, ages, and natural varience within those groups too as there is much overlap between them similarity wise.
Those pore don't show up so easily on the pale characters because the light reflecting off your model isn't as significantly different in contrast as on your dark model.
Interesting. I see quite a difference between, for instance, the pattern on the cheeks and the pattern on the forehead, but my focus on such details and immersive exposure to my own work may just be warping my perception again. In any case, I'm not sure where that apparent uniformity is coming from, since my bump maps are almost entirely derived from photo-based diffuse maps via a procedure that brings out and accentuates the bumps and creases that are already there, so theoretically at least, they should only be about as uniform as the merchant resource's original references were.
It might help to satisfy my curiosity as to whether the bump map itself looks too uniform.
I can see it there easily but it sure doesn't look that way in the render using it. Notice the difference between your render and that picture of the man you posted. Maybe the problem is the 'bumps' are too deep / too high?
Also, and it's a problem with your pale characters too but less noticable, the color of the dark skin in the render is exactly the same everywhere, even the lips, except for a few shiney areas & that contributes to the matte rubber look of that render.
...actually looking more closely at the picture of the man he is pretty even toned but the translucency is much higher in his skin. For a valid comparison you need to render your ealier render to make the light in the photo as much as possible and adjust the skin material settings from there. It looks like the picture was taken about 11 AM (or maybe 1 PM) in a sandy area (so much light reflected off the ground) in summer somewhere in west Africa when you set up Sun & Sky in the DAZ renderer.
I use Poser, so there's no Daz renderer involved, but your suggestions are easily translated. Here's what Billie looks like after a couple of texture/shader tweaks and under a rough approximation of the ambient lighting seen in the photo of the man.
Also, I think Billie as pictured above is perhaps a bit darker than the man pictured in the photo, so in case it helps the comparison, I spun the skin color dial a bit towards the lighter side to make her skin tone more directly comparable.
Much improved but that light is not near bright enough or reflective enough. Not really your fault, materials are still set up to default for clean studio style 3-point lighting because that's the way the are most often rendered and so you get no freebie material presets for sweat and oil and such, which besides there are a lot of variable to account for even with those two things and most attempts I see to do wet in DAZ Studio wind up looking oily instead because of the intensive labor needed to do water on skin surfaces. You probably need to account for oil & sweat and a lot more light both in the sky & being reflected from the ground.
For the moment at least, I'm okay with that (unless the skin is too far to the other extreme and looks unnaturally dry). I was never aiming for a particularly oily look anyway.
While I have this ambient outdoor scene setup, I decided to see how Celine looks in it with her revised freckles.
Improved but the distribution on both sides of the nose are too similar, even though I can see they aren't symmetric if I look closely. I think it's because you have quite a few 'run together freckles' drawn but I see no evidence of the overlap than makes the 'run together freckles' runtogether except how dark they are and how big they are.
And for your sub-Saharan African skin it does look very dry for some reason as if powder was applied. And disrequarding the oil & sweat completely the translucency looks way too low (as in non-transluncent). Even in very dark skin there should be evidence of translucense I think otherwise the skin looks dry & artificial.
The pale skin has the same problem but it's hidden somewhat because the skin is so pale. It looks too dry. Are these using the settings from your work on the model I think looks similar to 'Poser Miki'?
No, but the revised settings are alot of what made the improvement in the darker skin. The problem I'm faced with is striking a balance in the specularity settings between making the skin look moist and simultaneously not washing out the underlying color in darker tones, the latter of which I suspect was the main cause of the rubber-y look you saw earlier.
OKay, this version should at least look more translucent, if not also a bit more moist. I used Billie because lighter skin like Celine's does seem to mask the problem that we're targeting.
Honestly, I can't help but think I should try to get a few other participants involved in this thread before going too much further with any potential revisions. It's nothing personal, but you and I have become the only people here, and basing one's work solely on any single person's opinion, however well-thought-out they may be, is probably not a good idea. Where did Rashad go, anyway?
That's fine by me. Ciao
LOL! I didn't mean for you to leave the thread. I'm still interested in your feedback! I just may not act on that feedback much (if at all) until I get a few other people involved in the discussion.
Anyway, I think I've managed to find a way to have my cake and eat it too! A custom-designed compound shader node enables me to restore those specularity settings that you seemed to like best (i.e. the Miki-like ones) without having the loss of color in the reflection terminator regions (i.e. those areas where the reflection is weak but still present). You see, with those slightly older Blinn specularity settings, the shine was not only brighter than the surrounding skin but also less saturated. This is fine for the really bright reflections, which are supposed to look desaturated. The problem was that the desaturation wasn't attenuating proportionately enough with the brightness, so even those areas with only partial or weak shine were getting substantially washed out tones. Since surprisingly little of the surface is absolutely 100% diffuse/SSS and 0% specular shine, this was having a serious effect, though it was hardly noticeable with lighter skin. With my compound node, as the skin gets darker, more and more of the base hue gets procedurally added back in! Returning to the original single-spot style, Billie now looks like this. What do you think?
Well if you're not going to act on my feedback that's not much sense of my giving it as I've told you everthing I though was most noticably inaccurate.
And do those critiques still apply to the latest sample image? Your last critiques suggest essentially tthat the skin needs more SSS to make it look more translucent and better specularity to make it look more moist. Does the above render at least look markedly better than the original "matte rubber" look? That will suggest a direction for any minor experimental tinkering I do while I try to drum up some more diverse comments.
Just for the record, here's how the current shaders look on a lighter skin tone under studio lighting.
Greg,
I haven't abandoned the thread!!
Everything you're doing with this project is centered around this particular merchant resource. While it has been fun focusing mostly on what you have done with it, I think it bears mentioning at this point some of the issues with this source texture that could require remedy before you can move forward.
To help you break out of a possible selective blindness, I'll point out a couple of issues with the texture source itself.
1. No human iris looks like that. Just becuase you purchase a merchant resource it doesnt mean that all of the tissue samples provided are genuine. If indeed these are your own pupil creations I'd suggest a couple changes.
A. Human irises are basically cross sections of an exposed muscle tissue. Think of it as a thin cross section of a tubular muscle. There should appear to be distinct fibrous layers,more or less suspended within a liquid medium. One outer layer and one inner just surrounding thre pupil. Your current model looks as if it is a random set of noise pixels that vanish to a center point. But there doesnt appear to be any muscular structure, and the colors you are using are also completely imaginary. Ironically the color of an eye has more to do with the way light is refracted by that unique medium of aqueous humour, scattering blue toward the edge and often more greenish to yellow along the inner. There's really nothing blue within a blue eye at all, its all due to the way the medium scatters. so a heavilay painted iris will not be convincing. Also know where to blue and where to yellow and iris. Rarely will you see a normal human eye that is brown or black on the edge and yet blue at the center just around the pupil, but the other way around is quite common.
2. This texture has baked in shadow along the nose on each side, and this has an unfortunate impact when the texture is rerendered. Yes there are some freckels there, but some of that darkening is also a baked in shadow from the original texture source. This is one of thos eexamples I;ve been pointing out where the source artists decided to increase the contrast on these particular details, a choice that if it were up to uyou to make on your own you might jave likely decided against. Remove it I say, at least for a test. Allow the render engine to find all the depths and crevicesof the 3d model for you. You want to carefully remove all those directional details out of the albedo map if possible.
3. Looking at these women makes me hungry for chocolate brownies. Due to the almost matte appearance their faces look more to me like a sweet baked treat than like a true human. The good thing about rendering varying skin tones is that problems with areas like specular and roughness that might be difficult to discern on one tonal extreme can be quite obvious along other tonal extremes. Case in point. Your roughness settings are too high, so the sheen never rises as much as it should. One good rule of thumb is that the darker the skin tone of an individual the more magic the specular highlights will play in the final impression. If this woman is indeed outdoors and being lit with direct sunlight, and has typical to oily or sweaty human skin, that the areas of full specualr highlight should appear to be alomost full white in comparison. Look carefully at the specular highlights on the picture of the gentleman. I would venture to guess you couold lower the roughness on all of you rskin tones, and doing so would have another benefit...
4. Your bump/nomrals are set too high. Let me be frank and alert you that the height map you;ve derived while impressive, isnt truly to the level of detail it needs to be to accomplish what you are looking for. In fact I don;t think you could ever get such a map from conventional methods. Hopefully i will have a chance to paint out a few leather grains, pores, and other surface detail maps that you can enable and disable as needed. There are huge areas of near blankness on that bump map as well as deep dark impressions from the few details that it does have. its as if someone tried to bake away all except those details that really were profound. However in doing so a lot of the minor details were lost and cannot be recovered. I had hoped that adding in some basic noise might help fill in some of that lost info, but no.
When you truly get the proper resolution and detail from your height map, it will do a lot of the roughness for you, allowing you again to fix.
5. Also I would suggest that if you are not already doing so, try using true Reflection instead of mere Specular. In typical biased rendering engines Specular relfections by definition only respond to the illumination of direct point based light sources. Reflection however, regardless of the rendering engine, captures the entire environment. Ideally Reflection is trickier to utilize, but the richness that is gained from allowing the skin to fully interact with all of the light and color information in the scene will make the final impression closer to realism.
Unfortunately, this texture file has limits. You may be seeking to push it into areas it was not designed to go, beyond its reliable functionality. This isnt to say to stop, just to know that it is around this point that some degree of true innovation and consideration on your part will be necessary to get to the next step.
Thanks for your ever-detailed feedback, Rashad! A few responses:
The sclera is really the only part of the eye that comes from the original merchant resource. The rest is mostly painted with a set of brushes specifically designed to create irises. Also, I have the shader set up to use a higher SSS radius wherever the texture has more blue/cyan in it in order to at least roughly imitate the optical effect you're talking about. I briefly experimented with using a purely white iris and generating the blue color purely with scattering, but I'm not sure if Poser is even capable of that. I'll probably try again a bit later.
The baked-in shadows on the diffuse map will be a fairly easy fix, I think. I just always thought it was too subtle to really make a difference. Sometimes when I look at it, it's not quite as easy to know how much of it is genuine shadow and how much of it is actual tone variation.
Where are those near blank areas on the bump map? I ask because none really jump out at me when I look at the bump map itself. I would expect the bump to be more obvious in some areas than in others just naturally due to the lighting, but that's clearly not what you're talking about. Also, could you refresh my memory as to your recommended parameters for adding noise to the maps? I must confess, with all the other experiments and revisions I carried out, I never got around to giving that particular suggestion of yours a fair shot, so there may still be some hope there.
I would certainly appreciate whatever resources you decide to paint and share with me. If you do, I should advise you that I eventually hope to release this stuff as a for-sale product. Technically, it's already available at Renderosity, but it's a very outdated version, and I'm planning on releasing a "version 2.0" of sorts.
I had considered off-and-on using actual soft reflection rather than Blinn specularity, but let's address the bump issues first.
Are you using an actual photo reference for texturing or are you flood filling a color over a caucasian skin?
The latter.
.
This is an interesting thread, largely beyond me, but definitely interesting.
Hoping you manage to keep on going with this and achieve your goal
Firstly, I stopped trying to make photo-real pictures, as they are only possible under very restricted circumstances (see wolf359's sobering post, hitting the point), and mostly beyond me. But I admire people who get photo-real pictures.
There are some aspects in several of the pictures you posted for me to consider. Look at the eyes in the photos and compare it with the renders. The rendered eyes are "too perfect" sometimes. They look watery and with too many reflections. The sclera in contrast looks a bit too dry and structured. Also the iris/pupil area is sometimes rather (too) small compared to the sclera, in my opinion. There are real people (not few) with small iris areas but they look a bit artificial even on portrait photos. I think it is better to work with a bigger iris in renders.
Freckles, especially strong freckles, look unreal, even with real persons. If you add too much freckles to renders (where frankly said even the very best skin never looks like life skin) it can add the small grain of salt which tells you "rendered", even if it may be very life-like per se.
The girl from the June 9 post looks a bit deformed to me, and as such not real, even if the skin and eyes may work. With smaller slightly differently shaped eyes, a different depth of the middle face (aka less flat) and smaller chin it may get better. Of course it's a lot of bias involved in such an opinion, based on how you imagine people and as such renders.
The hair of the girl has only one advantage, it covers a bit the eyebrows. Both do not look realistic. The fault of the hair is obvious, the eyebrows are not dense enough and the single hair in it looks too stiff.
BTW, freckles again. Here is a character with some (default) freckles. I shaped him (very) losely based on an actor I like from the old movie "Heart Over Head". The skin is mostly out of the box. You immediately see that it is not a real person for several reasons, but the freckles look good, I think.
Well that won't work because the SSS / transluncency model isn't sufficently accurate enough so the flood filling will overwhelm the details of the base texture set. So to improve that you need to look at the character picture of that man's picture you cited as source and choose the lightest skin color present on his body's skin to use.
Or maybe, instead, perhaps you can add 3 new layers of texture maps that is a melanin map for a character and that consists of 3 colors for the 3 types of melanin color scientists know of (if I remember right) and a melanin density map for each of the 3 melanin densities. That to me sounds plausible to do with the current 3D shader technology.
You will need to look up the science of skin pigmentation to create a set of 3 melanin layers using 3D shader materials. Guessing is not going to work. The current model depends on almost all of that color being painted on in the diffuse layer and given shadow, reflection, refraction by the other layers. There is tramitted colors and others material settings to alter those but no really good documentation on how to use them.
Thanks for the continued feedback, guys! Nonesuch00, it's actually a bit more complicated than a simple flood-fill, and even more so now that I've been experimenting with ways to further mitigate the washing-out effect. The source color that I use to tint the skin for darker tones is actually quite bright already (one of the three RGB values is at a full 255), with the darkening that goes along with it taking place later in the node tree. The color and the base texture map are both fed into a custom-made tinting node specifically designed to preserve texture detail through the tinting process (though I can't say definitively just how successful I was in that respect). Also, there are three "layers" of SSS (only one actual SSS node, but three different values for each of a few key settings that are blended together at ratios modulated by the concentration of different color components in the raw texture map), and only the middle "layer" has the edited skin color applied to it.
Even so, after taking a short break, I decided to tackle at least what I consider to be the most pressing of recent critiques, namely the "baked good" look that Rashad was talking about as well as the intricacies of generating a range of believable skin tones from a single base map. I've tweaked the bump map, lowered the SSS radius a bit, and attempted to maintain more detail in the darker complexions by adding some map-modulated variance into the base tint color. As part of the latter initiative, I also incorporated a more nuanced range of tints for darker skin tones based on Luschan's chromatic scale.
For whatever it's worth, here's a test of the current results, this time featuring, just for the sake of not being too repetitive, skin that's not quite as dark as in the previous render of this character but still clearly within what we might call a plausible "African" range. Tomorrow, I'll probably be posting an updated version of the very first render in this thread, though again with somewhat modified skin and eye colors just to keep things interesting and further test the flexibility of my basic shader system.