Uncanny Valley?
Nyghtfall3D
Posts: 764
During the last few years, there's been an enormous spike in artist output using PBR's. It's become all the rage, and I don't see it slowing down anytime soon. That's a good thing.
There's just one tiny little issue with it that I can't seem to ignore from an aesthetic point of view, regardless of the otherwise excellent quality of so much of the work I've seen posted. The moment you add a human figure to a scene, the separation between realism and caricature becomes undeniably noticeable, and it looks like two different forms of artistic expression vying for attention.
I have yet to see a still life or architectural render that didn't look technically impressive. On the other hand, as beautiful as V7 looks, even she appears noticeably disjointed from every environment I've seen her rendered in so far. I don't know whether it's a MAT issue, posing, or something else, but I believe the most appropriate term is, "uncanny valley". If so, I finally understand what that means.
On the opposite end of the spectrum, where biased renderers like 3Delight and Firefly reside, objects and light reveal imperfections that the limitations of those engines kept the artist from resolving, or that the artist lacked the skill to improve on. But the result is that the imperfect environment and the imperfect human look like they're part of the same world.
I am really excited about V7 - I just bought the HD add-on for her. I don't want to go back to biased rendering, but I think human figures have a long way to go before achieving a level of realism that makes them appear more like part of whatever PBR environment they're put in.
Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
I definitely agree. I think it just depends on some things. When I show my friends my renders of the humnan figures one thing I always hear is everything from the head down looks like a photograph but as soon as you get to the face that's when things get uncanny. I dont know if it's an expression thing or what.
I tend to think most 3d females just don't look real. Too little bodyfat, big eyes, and mannequin features just stare deadly off into space. 3d females are often idealized in ways we just don't see in RL. Also the effect of gravity is difficult for many to reproduce, not to mention the skins which may look real, but the clothing which looks like plastic or dough, and the helmet-like hairstyles. More realistic engines often make this more pronounced. It is no good to see every pore if the clothing looks like thick plastic.
3d females are lovely to behold often. But quite uncanny valley without a rreal amount of postwork.
The talent of the artist and the amount of work done has a lot to do with how uncanny it may look. Here's Chris Jones' work that looks like he has almost or even pretty much left the uncanny valley behind.
There's always the subtle visual cues we use to establish what looks real, and we require a lot more detail for the living things.
But another reason that sticks out for me is the posing, which includes facial expressions. I recently went through dozens of pose sets collected over the years, and threw most of them out. Even the "commercial" ones bought here, Rendo, and other places are filled with really bad poses, and I'd rather remake them from scratch than be tempted to use these even as starters. Many poses are unbalanced and unnatural, completely negating the realism of the render. Quite a few are even anatomically impossible, such as twisting the hand rarther than forearm.
I was looking at some paintings by Canaletto the other day and marveled how "real" the people looked, though obviously it's all in a drawing (and done in the 1700s). I noticed its the non-static poses that makes a big difference. They look they like they are captured in time, like a film frame, rather than a tableau of people who are standing still for an hour.
I don't think their drinking choice matters that much. :)
And the disjointment between real and unreal is NOT the Uncanny valley, unless it's causing revulsion or unease to you. It's a concept that came from the creation of artifiical limbs....a feeling that what you're looking at is off..wrong in some un vocalized way. It was from studies of how people reacted to seeing an artificial arm. It was later applied to animation...and then became an internet buzz phrase :)
(The term "valley" is because of how a graph of responses looks on a grid of Emotional Response vs antropomorphism)
For myself, I like the meeting of the real/unreal. 25 years ago, I coined the term "Interface Animation" in an animation class to describe films that had life and animation meet..(like Pete's Dragon, Gerite the Dinosaur, Roger Rabbit)... these days with CGI, that kind of thing isn't even a thing any more, but I like the effect.
PBR?
Pabst blue ribbon (cheap beer)
my avatar used pbr, but she wasn't trying to be real. Realism is overrated :)
as as scavenger pointed out. The valley is narrow
I take it considering the abbreviation madness that this is an american thing?
That was .......... Oh My GOD! WOW!
Everything you always wanted to know about the uncanny valley, explained in 55 seconds in the now-famous 'Succession' episode (2008) of 30 Rock (Frank explains the valley to Tracy, who wants to create a porn video game):
So famous that people outside of the US never heard of the term to start with
Bah-dum ching... SLAP!
For the uninformed, PBR in 3D art terms means Physically-Based Renderer.
Nope, no revulsion here, just something I keep noticing...
Ah. Thank you for clarifying.
No...Pabst Blue Ribbon is basically the dregs from a good beer, filtered through the kidneys of a weasel...then watered down and pacakaged in the cheapest cans known to mankind. They'll spring a leak if you look at them cross-eyed.
If you live in Europe, the cheapest, nastiest beer you can find is probably a couple of orders of magnitude BETTER than PBR.
It won a blue ribbon at a rigged competition over 100 yrs ago...the judges were too far gone on the 'good stuff' to realize what they were drinking....
Which? Pabst Blue Ribbon or the Uncanny Valley? If the latter, it's a term coined by a Japenese cybernetisist in the 70's.
As for Pabst Blue Ribbon
That was surprisingly well done.
The one, and only, time I tried Heineken (my beer-less wife grabbed the cheapest bottle on the shelf when I asked her to pick me up something), the first thought in my mind was "How much whisky did they pour into the horse that [******] this?" I'm a Guinness man, myself (or Killian's), but have not subjected myself to PBR. I don't even touch Bud.
The reason I began doing "noir" or comic-book shaders was to leave the fake-look appearance of de facto 3D renders behind, before I even knew "unbiased" was a thing.
3D models are symmetric to a degree that you will never see in real life. Very few characters address this.
Speaking of Pabst Blue Ribbon (for fans of Blue Velvet and/or David Lynch):
(language warning).
The dead eyes, perfect skin and unrealistic expressions are dead giveaways. I doubt V7 will change this problem based on what I've seen so far. Concerning animation, unless DAZ Studio's animation tools are significantly improved then trying to edit a combinaton of aniblocks and keyframes to create naturalistic movement is virtually impossible. I agree that quite a lot of the commercial poses and aniblocks I've bought are very poor quality. For example, I'm currently animating a project where a commercial pacing aniblock I've used has noticeable glitches. The only way to view what is causing this glitch is to slow the animation down. At certain points the leg will suddenly twist sideways in a very unnatural way. I spent ages editing keyframes in this commercial aniblock to correct the movement. No wonder then that those of us trusting in the accuracy of commercial poses and aniblocks find it very difficult to create naturalistic images and animations. More versatile tools, better products and greater quality control would go a long way to achieving that goal.
Nope. It was from the UK. The guy who came up with the CFA (campaign for acronynms) was given an OBE, and later, a KBE.
How about a Utica Club?
I know what it is...
...It just sounds like a salad dressing for 3d models.
Honestly i dont mind. i think it gives it a bit of an artistic touch to me the figures dont seem creepy with g2 or g3 in my opinion. You may sure be able to tell its not "real" but maybe thats missing the point of daz studio . as if you want real real. might as well just use real models :-p.
to me i think the figures ive seen so far do just fine with the uncanny vally and dont seem unatural and as far as "dead eyes" maybe it depends on what type of eyes you use ;-). it may not be perfect but honestly still one of the best options out there as far as 3d figures go other then making your own from scratch!.
I kind of agree that I'd rather embrace a quality fantasy render than fight for uber-realism. I love what Iray has done for the overall quality of what I can create -- not due to realism, but due to lighting and detail that is just better than what I usually get out of a biased engine. (larsmidnatt's avatar is a great example of what I mean.) This is as close to "realism" as I've been able to get in Iray thus far. It's easier to get realistic renders when you're not up close, I think.
Well that's the thing..the "valley" is personal to each person, with generalaties across people, and likely cultures.
This is the same reason why those celebrity character morphs are more caricatures than realistic. As it's not possible to include the tiny subtle human details in Daz/Poser, the facial features have to be exagerated to make them recognisable.
Celebrity, or even characters based on friends/family, will only ever be caricatures. Not because the render engine is lacking, or the morph is lacking fidelity to the original; but, because we lack the ability to accurately mimic the subtle facial movements, expressions, and affectations that every individual has. Even in real life, one can say, "he looks just like Sean Connery." The truth is, only when "he" strikes a certain pose, or adopts a specific affectation does he actually resemble Sean Connery. Rich Little, a famous impersonator, was a caricaturist. He didn't look like anyone, but was able to, moment-by-moment, recreate a caricature of that person to which the audience would respond, "damn that looks just like Richard Nixon.".
We are, actually, approaching, the uncanny valley. When we as 3D artists actually descend into that valley, we will completely and forever lose our audience. If Rich Little were actually able to recreate the actual image and mannerism of his target, he would have been shunned. In the same way, if we are ever able to recreate, let alone animate, a character with fidelity, we too, will be shunned. IMHO, no artist, from DaVinci to modern times ever captured their subject in total. Artists, whether in paint, photo, or 3D recreations can only ever capture a moment; a snippet, a single facet of the original.
As I look at the PBR renderings of fictitious people, I find that the most "realistic" are those that capture a look (think Mona Lisa), or a mood, or even a memory - unique to the viewer - in which they see true "realism" (not withstanding the stereo-typical "she's hot" response). Yes, asymmetry, specularity, reflection, mat fidelity, and lighting are key; but, in the end any piece of art must strike a chord in the viewers' mind and heart. I have attempted to "recreate" friends and family in 3D. I have come to accept, and state, that I can create a caricature, but I will never attempt to "recreate" a person. I truly believe that if I could, truly, capture the image of a living (or dead) individual, it would be shunned as it sat firmly at the bottom of the uncanny valley.
No portrait artist has EVER presented his/her subject as they truly are. They draw upon their skill to enhance (even pre-Photoshop) their subject. A stronger chin, a slimmer waist, a shapelier leg, a fuller head of hair. PBR rendering will not bring us closer to "realism". What it will do is provide a stronger/better tool set to recreate the environment in which our subject poses. The uncanny valley will be reached when the viewer can no longer distinguish between our renderings and a photo of an individual. I believe, when we achieve that, we will be shunned. Humans like to live in a fantasy, the popularity of the anime world is evidence, for it showcases the "preferred" view, rather than the reality.
what I often do is setting min. values of expressions for example from 0.0 to 0.2 or whatever is needed. So the figure will have this bit of expression in every other expression that is used.
example on v4:
eye squintR min value set from 0.0 to 0.2. So she will ever have the right eye squint at 0.2 in every other expressions, that means if the right eye is whide open it is also squint at 0.2. The right eye will look a bit different then the left in every other expression also if the values for the eyes normaly equal. That works fine.
I have done this also for a complete face expression so she will ever have this bit strong expression also if she is smiling or laughing. The same I have done on the g3f character I am working on.