Anonymous at Thu, 3 Mar 2022 19:37:46 UTC No. 885466

it's hard to replicate human skin and its interaction with light, even with current shader and lighting technologies

Anonymous at Thu, 3 Mar 2022 19:55:13 UTC No. 885467

>>885462
well, for one, the guy has no eyelashes. #4 I think gets closer to reality, since you're removing another source of unrealism - the colouring.

Anonymous at Thu, 3 Mar 2022 20:03:56 UTC No. 885472

>>885466
what do you think is the hard part? Even with cross polarized scans they are uncanny

Anonymous at Thu, 3 Mar 2022 20:14:45 UTC No. 885475

>>885462

simulated light sources are too pure. most light sources have rough peaky spectrums. real world materials absorb and reflect different wavelengths differently. shining a pink light onto human skin might not give back such a pinkish reflection, it might shift the hue a little as some specific wavelengths within the pink light are absorbed more than others. forcing all the calculations in to RGB channels makes things come out plastic looking.

another layer of missed realism is simulating how actual physical cameras capture light. our brains are familiar with many different photographic traits. we can tell the difference between a color photograph from 1970 and a digital photograph from a modern $2000 DSLR, even if the end result of each is a jpeg, but we recognize both as photographs. 3D renders have an unfamiliar feeling to their appearance.

Anonymous at Thu, 3 Mar 2022 20:20:03 UTC No. 885477

When eyes and teeth are hidden and it STILL fails (while having the bone, fat and muscle structure spot on) then it can only be the lighting and skin materials. Especially how light it scattering inside the skin is one of the most difficult things to get right.
Many just crank up the skinpores normalmaps and SSS to the max and call it a day

Anonymous at Thu, 3 Mar 2022 20:22:28 UTC No. 885478

>>885462
notice how it all falls apart in your mind where you expect the skin to be thin even when the image is only gray scale

Anonymous at Thu, 3 Mar 2022 21:46:20 UTC No. 885490

>>885475
This is why good & professional renderers incorporate spectral rendering

Anonymous at Thu, 3 Mar 2022 22:06:01 UTC No. 885495

>>885462
It doesn't look real? Could have fooled me, looks like one of my former managers.

Anonymous at Thu, 3 Mar 2022 23:26:05 UTC No. 885505

>>885472
>what do you think is the hard part?
Not him but skin, or rather flesh (because it's more than just the outer-most dermis) is very complex:

>skin is actually translucent, light passes through it
>dead bodies (of Caucasians) are greenish blue (assuming the blood is pooling in the back or extremities)
>redish pink color of skin is from blood
>skin is multiple layers of different kinds of material: translucent tissue, fat, vascularity, then all the defects that can occur on the outermost layer (age spots, moles, etc)
>the physical structures within the flesh, such as veins, musculature, fat deposits all redirect the light bounces in addition to the color
>skin is naturally oily, which means it's naturally got some gloss but not too much and not necessarily uniformly
>for extreme closeups, the pores on skin are actually little divots or well-like shapes, again this changes how light bounces
>fine hairs on the epidermis, which all humans have even women even if they shave, are yet another significant factor
And remember it's all these little things coming together that form a total perception that reads as "wrong".
Humans, unlike many other mammals, rely very heavily on vision and while we don't have necessarily the best eyes, we have extremely developed visual cortexes.
This part of our brain does a huge amount of interpolation, extrapolation, and processing of visual information (which can be exploited by visual illusions).
It may not tell you what exactly "looks wrong" but it will tell you something "looks wrong"; the ability to spot tiny differences between things is an extremely powerful skill that we, as primarily scavengers and opportunistic hunters, have honed over a long time.

Semen is also super complex as a substance, as is blood which is why they often look really bad and are very hard to do right

Anonymous at Fri, 4 Mar 2022 00:24:27 UTC No. 885512

>>885462
>entire thread is wrong
the lighting isn't interacting with the bone, muscles, layers on skin, veins etc. it's essentially just a balloon with a skin texture on it

Anonymous at Fri, 4 Mar 2022 06:23:57 UTC No. 885532

>>885478
that one specifically looks completely real, exactly like a studio photo

>>885475
>>885505
>>885512
>all this minor bs that doesn't matter at all
it's probably mostly the ambient lighting which is wrong

Anonymous at Fri, 4 Mar 2022 08:15:08 UTC No. 885535

>>885532
nah it looks like a blob of vax

Anonymous at Fri, 4 Mar 2022 08:54:28 UTC No. 885539

All the fags shit talking here couldn't do better lmao

Anonymous at Fri, 4 Mar 2022 14:44:35 UTC No. 885567

>>885535
You look like a blob of wax

Anonymous at Fri, 4 Mar 2022 16:03:19 UTC No. 885582

>>885462
I'd argue it looks 'real' as in it almost could be a real artifact in the world, but the model itself is way too undetailed to read like a real human head.
Esp the quality of the hair and low resolution of skin(compared to detail level of real skin where we can see light return from the edges of skinflakes a few microns across, ~16K+ map for only the face etc)
and just all the missing details like fuzz hairs, dust particles, oil buildup, etc so all in all it looks like a realistic rubber movieprop type head of equivalent detail.

A real head in those light conditions and that degree of focus you'd just see tons of micro-details and specs of highlight from all the skinflakes and fine fuzz hair
breaking up the shading across the entire surface, here we have none of that so it looks like vax/rubber meant to mimic skin.

Anonymous at Fri, 4 Mar 2022 16:32:51 UTC No. 885583

>>885582
>low resolution of skin
bitch it's a 391px image you can't tell that for shit

>real skin where we can see light return from the edges of skinflakes a few microns across
that's for a microfacet model, not a texture

the model is a high resolution scan of a real person, you are an idiot

Anonymous at Fri, 4 Mar 2022 17:26:00 UTC No. 885589

>>885582
N

Anonymous at Fri, 4 Mar 2022 18:23:41 UTC No. 885600

>>885583
>bitch it's a 391px image you can't tell that for shit
That's not how it works anon. Sampling requires a lot of data to set the correct values for each visible pixel.
If all it took to achieve photorealism was to match the pixel resolution of the final image all of our lives would be easy mode.

>that's for a microfacet model, not a texture
You don't even comprehend the problem here, if you want the right return value of a pixel your simulation has to match the real world phenomena you're trying to capture.
The pixel value as captured if you snap a picture of someone with a DSLR depends on all light returned from that point of the person, your digital sampler has to account for this
and capture something that gives the same light return as the real thing, which mean you have to account for how all the microdetail shades if photorealism is the goal.

>the model is a high resolution scan of a real person, you are an idiot

No anon, you are acting the idiot here by stating someone who understands this topic on a magnitude above your current comprehension is an 'idiot'.
It doesn't matter that it's a scan, this kind of scan doesn't sample all data present in the artifact it scans.
It snaps a good picture of it and when you rasterize that for a render you must re-sample your scanned samples and the result is very crude and simplified to the original dataset (real head).
It's kinda like if you snapping a picture of a picture of yourself, you'll lose a lot of data in each of the sample steps.

Anonymous at Fri, 4 Mar 2022 18:53:35 UTC No. 885607

>>885600
these are resampled with 1000x supersampling it could not get more perfect

you can't tell if the texture is low resolution when the picture is low resolution. that is just a fact. shannon-nyquist.

scans are imperfect yes, but not for any of the reasons you think

you are an idiot

Anonymous at Fri, 4 Mar 2022 18:54:47 UTC No. 885608

>>885600
neck yourself

Anonymous at Fri, 4 Mar 2022 19:18:56 UTC No. 885613

>>885490
I didn't know about this. It looks like there are ways to fake spectral rendering with rgb, too. You can program the material shaders to behave differently when the object is near different light sources or you can slap a color grading filter over the whole thing. Maybe that's why console games went through that "orange and blue" phase... it simplifies the problem.

>>885532
Different renders reveal different problems. #5 definitely reveal the problem with colored lights, while #2 and #4 minimizes that problem. But in #2, you can more clearly see the rubbery subsurf. Of course this is all nitpicky, but we are looking for those hardest to reach improvements.

Anonymous at Fri, 4 Mar 2022 19:34:45 UTC No. 885617

>>885589
I

Anonymous at Fri, 4 Mar 2022 20:10:59 UTC No. 885627

>>885617
C

Anonymous at Fri, 4 Mar 2022 20:14:36 UTC No. 885630

>>885627
E

Anonymous at Fri, 4 Mar 2022 20:22:32 UTC No. 885632

>>885630
N

Anonymous at Fri, 4 Mar 2022 20:23:42 UTC No. 885633

>>885632
I

Anonymous at Fri, 4 Mar 2022 20:24:46 UTC No. 885634

>>885633
G

Anonymous at Fri, 4 Mar 2022 20:31:06 UTC No. 885635

>>885634
E

Anonymous at Fri, 4 Mar 2022 20:32:28 UTC No. 885637

>>885635
R

Anonymous at Fri, 4 Mar 2022 20:33:32 UTC No. 885639

>>885637
I

Anonymous at Fri, 4 Mar 2022 20:34:35 UTC No. 885640

>>885639
A

Anonymous at Fri, 4 Mar 2022 20:53:34 UTC No. 885648

>>885640
P

Anonymous at Fri, 4 Mar 2022 20:54:36 UTC No. 885649

>>885648
E

Anonymous at Fri, 4 Mar 2022 21:46:11 UTC No. 885656

wtf is going on here

Anonymous at Fri, 4 Mar 2022 23:01:39 UTC No. 885659

>>885613
they went through the orange and blue phase because of hollywood and they just discovered fullscreen image filters so of course they overused it

Anonymous at Fri, 4 Mar 2022 23:07:01 UTC No. 885661

>>885490
>>885613
again something that doesn't matter in 99.9% of cases. for the very very few materials where it does matter like thin film and sky materials you can do specific spectral calculations without needing a full spectral renderer

realism comes down to tone mapping, lighting and textures
that's why some of the earliest CG is still the most realistic, as they had to match the tone mapping directly to a film plate, manually do all the lighting again matching it directly to a film plate, and textures would be sourced from photographic sources.

now you have generic tone mappers designed for animation, generic auto-lighting solutions designed for animation, and simulated materials from substance painter and whatever. so it never comes together the same way as say godzilla or agent smith

Anonymous at Sat, 5 Mar 2022 14:20:26 UTC No. 885722

>>885462
Why do humans have to be the most advanced entities on this planet, why can't we just be robots to make modeling easier

Anonymous at Sat, 5 Mar 2022 18:03:13 UTC No. 885744

>>885661
>BAAAAAAAAAAAAAAAAAAAAWWW

Anonymous at Sat, 5 Mar 2022 21:39:10 UTC No. 885762

>>885661
>again something that doesn't matter in 99.9% of cases
>now you have generic tone mappers designed for animation, generic auto-lighting solutions designed for animation, and simulated materials from substance painter and whatever. so it never comes together the same way as say godzilla or agent smith
>doesn't matter in 99.9% of cases
>generic x y z never comes together
Not sure I'm following that logic, but your explanation of matching tone mapping to film is interesting. This seems like something hobbyist newbies would miss, if they hadn't been around when practical effects and film photography (and actually filming on scene) were still widespread practices. Formally trained artists would know, assuming their professors were old guys.

Enki at Sat, 5 Mar 2022 23:32:13 UTC No. 885779

add more imperfections around the eyes and face. real faces have alot of imperfections it adds to the realism factor. also sweating and pours study faces deeper.

its not bad. take time
imperfections in hair also help give it more realism. real things are Never perfect for this reason.

Anonymous at Sat, 5 Mar 2022 23:52:58 UTC No. 885784

>>885462
Needs more ugly scanned real world people, especially some mystery meat goblins, then it's perfect and realistic. Current industry hates beautifully characters in games. It makes the femfatties insecure.

Anonymous at Sun, 6 Mar 2022 08:04:29 UTC No. 885816

>>885722
If you were a robot, it would be hard to draw other robot's likenesses.

Anonymous at Mon, 7 Mar 2022 04:30:47 UTC No. 885924

https://youtu.be/3Ws-hl1zt_s

found this video today - I was on page 4 or 5 of an unrelated youtube search, panning for gold in the river, and I see this mother fucker. I said Hey wait a minute

Anonymous at Mon, 7 Mar 2022 04:32:16 UTC No. 885925

>>885924
Uploaded September, 2012

Anonymous at Fri, 11 Mar 2022 10:48:32 UTC No. 886490

>>885462
Too smooth.

Anonymous at Fri, 11 Mar 2022 10:52:59 UTC No. 886491

>>885462
I fix it.

Anonymous at Sun, 13 Mar 2022 02:38:42 UTC No. 886705

>>885613
Pretty sure Battlefield artworks from 3 onwards all used real actors for their characters, it's just really well composited.

Anonymous at Sun, 13 Mar 2022 16:52:29 UTC No. 886753

>>885462
Your brain says it's a corpse.

Anonymous at Sat, 19 Mar 2022 14:23:17 UTC No. 887448

The eyelids are bugging me, but I couldn't tell you why.