Photo-real characters. A different approach.
Ooze3d
Posts: 10
I've got a question for whoever wants to give an opinion on this matter.
I've been involved in 3d stuff for a very long time. Mostly as a hobby and a side job, but I hadn't done any human related work till I started messing around with Daz Studio about a year ago. From that moment I've been a bit obsessed, like many others, with skin shading and mostly what makes an image appear like a real photo. Something that's so simple with objects now that we have lots of different Global Illumination and Unbiased render engines, always seems to have something off when it comes to simulating a human character, as we all know.
I've been reading tons of articles, endless forum threads, posts, tutorials... all of them about creating the perfect skin (and hair, eyes... everything). All of them go on and on about technical stuff on the lights, the behaviour of all the different layers of the skin and a huge list of everything you need to do to make your character look alive and realistic. Most threads have high res pictures showing incredibly detailed renders and explaining how you can't have your photo-real look without:
- A good lighting setup.
- The allmighty displacement map. Some people swear by this one.
- 357 different maps for all the different layers of the skin. Dermal, subdermal, blood vessels, coating, bump, translucency, reflectance, glossy roughness, sss... all of them absolutely essential if you don't want to fail miserably.
- Imperfections. Moles, creases, freckles, marks, pores... if you don't have them, it looks too perfect and reeks of CGI.
- A proper, natural pose and expression.
...and on and on. But now I've got a question:
If I need a proper lighting setup, bump, coating, gloss... how is that my eyes can look at a badly lit, low res, screen cap from a heavily compressed phone video showing a flat face with a vacant expression and not doubt for a second that I'm looking at a real person?
Or if I need all the fine detail, the displacement, the imperfections and all that, how is that I can see a high res photo of a model with heavy makeup and photoshop work on top to the point where I can see no pores, no creases, no moles, no freckles, again with the kind of vacant, lifeless expression that most models seem to love... and still be 100% sure that this woman is human and was alive at the moment they took that picture?
And how is that I can take 99% of the best Daz Studio renders DeviantArt has to offer, no matter the resolution, distance to the subject, amazingly detailed maps, lighting conditions, render engine (iRay, Octane or Reality), extra photoshop work... after all these years and having literally thousands of people working hard to make their renders look as realistic as possible, I can't find a single one that manages to fool my eye and makes me think I could be looking at a real person (I'm specifically leaving apart examples rendered in Arnold. I’m only talking about Daz Studio).
Note: Don't get me wrong. I'm only talking about photorealism. No one denies the hard work and the artistic value of the images I'm talking about.
Now, if we don't really need to see all the details. If we don't need the perfect lighting setup. If we don't need to see natural expressions or the skin shining in a certain way to tell apart real people from a render, then what's the key? What's the secret ingredient that's missing?
I'm asking because I'm far from sure, although I'd say my highest bet has something to do with the eyes.
What do you think?
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
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
I agree that it could be the eyes. I've been working with Daz models for couple of years, and I'm still not happy about my eye shaders. Also, most characters are very symmetrical, so that could be it also. Most of us have little asymmetry here and there, and that makes it look normal.
The eyes are the first thing you look at on a person's face. They're the things you're looking at when you're speaking to someone. They are the one thing that fails miserably in every image I've ever done and I'm thinking in images others have done. Because your brain is wired to recognize the human face, you're going to notice even the smallest of differences.
Laurie
I think the problem is that we, as people that work with 3D models as a hobby, know they aren't real and fail to put ourselves in the perspective of the average person that sees our renders when we post them to sites like deviantArt. Those folks, unless they themseves work in a field where they deal with CGI on a daily basis, often can't tell the difference between a render and a photograph unless we say something to indicate it isn't a photo.
+1
I think you're right about the eyes. It's not just the eyes though, I think. Real humans have a TON of asymmetry in just about every part of their faces and bodies. Most 3D characters have too much symmetry and the brain picks up on that even if we aren't aware of it consciously.
People have so much asymmetry that most people if you copy half their face and flip it/mirror it, they often look like a different person.
Original face - without a doubt a real person:
But once you mirror one side of the face, we can tell something's not quite right.
Left side mirrored:
Right side mirrored:
If you take just about any face, and mirror it, your brain tells you that something's not quite right.
I think too much symmetry plays a big role in how we're able to spot a render vs a photograph.
This!
I show quality images to my coworkers that have never done anything with 3D and they are like 'how did you/they get this person to pose for this?" or "I didn't know you/they were into photography" types of comments
The biggest issue I see with figures is most of them look derived from a DAZ figure and not far enough removed to think of the mesh as it's own identity like what you see when you look at a quality zbrush type of sculpt.
I think this is probably one of the biggest reasons. My brain thinks the last two are "not real", even though the eyes and everything are from a real photograph. True symmetry is just not natural, at least not on people.
When you flip one side of a picture to the other it can cause other telltale signs that its wrong that have nothing to do with the actual physical face of the model being too symetrical:
The symetrical physical face doesnt cause me to think somethings wrong. Its mainly her looking dead straight at me like an android, and the symetrical hair (hair is impossible in real life) that does it.
Notice in the real picture she isnt looking dead straight at you (she is turned slightly towards your left ear), but in the folded pictures she is. Thats because in real life people rarely turn themselves exactly straight at you.
Regarding the original question - I think it is lighting. In real life we are lit from all directions unevenly due to light bouncing off objects. Even in a dark scene the lighting is very complex in real life.
I definetely DONT think its things like moles as I dont know a single person with a mole on their face in real life (we dont get much sun here), so whenever I see a daz model with a mole it alarms me.
Thanks a lot for all the answers!
First of all, the eyes do seem to be a problem. I was running tests yesterday (ended up going to bed near 4 am when I had to get up at 7) because I found a product I hadn't noticed before. It's called Macro Eyes from Parris and, so far, it's the very best eye shader I've seen. I just tried a couple of presets and I was shocked to see a character I'm working on "looking back at me" for the first time. It still needs some tweaking here and there, but it's promising. I'll post some before/after pics as soon as I can.
DivaMakeup, good call on the symmetry! It's absolutely true. I'll try to mess around with the symmetry of my character to see if I can get some "didn't look asymmetric but it is" look and see if it improves the feeling of a real face.
Also, yesterday (lots of things happened yesterday) I got the feeling that I had posted this thread way too soon. I thought I had seen most examples of what people can do in terms of skin shading in Iray, and I have to say I knew about AS-Dimension-Z and his Anagenessis 1 and 2. To be honest, it did feel better but I couldn't manage to get very good results with it. Some of his promo renders were very good but I could still see it was a render. That is, until I saw that he's currently working on Anagenessis 3, and oh boy did he take a spectacular step forward. More of a jump, really. I was checking his DeviantArt page and I found myself staring at some pics for minutes, amazed by the feeling of looking at real skin. I stumbled across it because I started testing something I hadn't tried before. The famous "Dual Lobe". I know I basically said skin reflections are overrated in my first post but it's true the Dual Lobe feature does improve the feeling that there's something you can really "touch" there. Again, I also saw an Anagenessis 3 test of a face that looked like it had a good amount of base makeup on, so it didn't shine at all, and it still looked very real. I'm really excited!! We're getting there!
I'll try to post examples as soon as I can.
Thanks again!!
You also have to bear in mind that a lousy screen cap from a compressed video will hide alot of faults.
If you were to do a render of the same quality, it could easily looks are real as the screen cap.
But a big part of getting images to look life-like is in the poseing. Alot of the pose sets here are great for giving a basic pose, but for any situation, there are subtleties to posing which can turn a manequin into a person, or vice verse. Things like eye position, where the person is looking, and how that relates to their hands, or their stance, etc. All these elements have to work together, and when they do, you can have a very real looking shot, rather than a stiff render. (provided you do have good skin and lighting, etc as you already pointed out)
Rawn
I very much agree that the pose is critical, particularly for non portraits, but even here it is important, not just the expression, but the subtleties that can make a render real. When someone is standing, gravity has a big effect on their pose, as their body compresses in all sorts of ways as their weight is distributed to various parts. Gravity is also responsible for subtle skin/flesh sag, and it is this effect that is missing. Obviously applying the effect of gravity on a 3D model in a realistic way is incredibly difficult.
Clothing fit is another area that kills a realistic render. I have kicked off a personal crusade to fix this, starting here.
Too often, the clothing looks more painted on or a shell over the actor. There are no contact points showing tension and friction - think waist band of pants while worn. Sure, this is nullified if you only work with nudes, but then everything mentioned by @RawArt and @Havos becomes apparent. I have scenes with so many D-Formers in place just to move clothing items around into more natural positions that, when all set to viewable, it looks like I found a new element for the periodic table rather than a 3D actor.
Getting hair to lay naturally is another easy to spot fault, but one Barking Beast hunt at a time.
Thanks!
Pellinore, er... Omen
Hair and clothing movements are big giveaways when you're attempting photorealism. In photography, even with the use of a tripod, timer, and external flash units, you are going to get a small degree of motion blur in hair and clothing as the model moves. I usually do not see that blur in renders at the usual online galleries. That absence of blur, to me, causes me to suspend my disbelief. This is true even when a render is otherwise technically superlative, and it is far less of an issue when ultrarealism is not the goal.
There's also the lack of virtual muscles emulating the tensions and relaxaions that occur whenever a living creature moves. Some muscles contract, and others expand for every move we make.
There's a pretty big caveat to keep in mind though. When you mirror a photo like that, you also mirror the lighting and the perspective. This makes it look weirder than it really would be if her face was actually symmetrical.
Eyes are a big giveaway but we are also pattern recognition creatures, evolved to excel at picking up the micro details that make up faces and assessing them at a glance, and when something is off, even though we may not quite know what it is thats wrong, we are aware of it. Patterns, patterns and more patterns...
Interesting topic. The mirroring of images is a very good technique in for us artists that also draw. Drawings really appear different when mirrored and you can fix issues.
I think adding natural flaws would help in making images appear even more real, I think BlueJaunte models have a very relasitc look and certain added naturalness in the models. At least it appears to me, and why they are sometimes more appealing to use over the perfect asymm models.
Greetings,
I've more than once considered postworking my images to have the flaws that a cellphone camera in sketchy light adds in, just to see if it increased the versimilitude of the images. Making the white balance worse, and stuff like that.
Also almost no human holds a camera perfectly flat; in studio shots you can get away with that, but everywhere else there's a hint of tilt or angle. Also some hair is (again) too perfect, typically...not enough flyaways, arches or bridges, details.
But it's all part of the same thing; asymmetry and imperfection, eyes that seem real, picture quality, subtle inconsistencies everywhere that cue us to what's real and what's not. It's kinda cool that we're at the point where that's what it's going to take to get there, having gotten past the body/face form itself, the intricacies of sub-surface details, bumps, hairs (both body and atop the head) and more. Now we're down to the really subtle, hard to pin down things.
That's one of the reasons a _lower_ quality image can seem realer; its own imperfections hide the too-perfect rendering.
-- Morgan
That’s definitely one of the most difficult parts of the whole process. Getting the “weight” of the character right so it doesn’t look stiff takes time and lots of observation on real subjects. When mastered, that’s a “trick” to make characters look realistic even if the render itself is not. The work of DB Spencer comes to mind. Her characters are not exactly photorealistic. More like comic characters, but they look much better than others thanks to the natural feeling of their poses and expressions, both of them top notch. Also, the DOF she gets using Reality helps with that photographic/artistic look.
That's another problem. Hair now is way better than years ago when it was basically a piece of PlayDoh on top of the head, but it still requires lots of post work to look like real hair. Talking about the “imperfections” factor, I love using the Everyday Updo piece because it adds that extra touch of randomness to the character, but since it has so many loose hairs here and there, it can look very fake depending on the style and the angle. Again, it’s all a matter of practice.
I've had a moderate breakthrough (or what certainly seems to me like one, anyway) when I decided to experiment with triple-layered SSS. I looked up a couple of Blender shaders that looked really impressive and, after a bit of research, came up with a decent adaptation of it for Poser 11 SuperFly. Regardless of whether it truly tricks the eye, it seems clear that it's at least a big improvement over single-layered SSS. So I think multi-layered SSS may be very important for realism. In my opinion, the trick with imperfections is subtlety. I think you need blemishes to be one of those things that are more conspicuous when absent than when present, if that makes any sense. They may hardly be noticed when they're there, but when they're not there, people definitely notice. You don't want your model to look like a doll, but at the same time, you don't want him/her to look diseased either.
For whatever it's worth, here's a close-up sample of my textures/shaders that fetched some very kind words at the Smith Micro forums.
And here's a more mid-range shot of the same character.
True. I saw some renders a while ago (in the DAZ gallery I think) which were intentionally made so they looked like low quality photos from an old cheap phone. I'd never believed they weren't photos if I hadn't known.
This is a mega interesting topic!...
I think the photo mirror example put by Divamakeup is a huge step to breaking this down.
I think if you are to theoretically search for the item that triggers your acceptance of what is real/fake, it will involve you systematically deducting the face piece by piece until it looks fake.
So, ie- you take a real photo, and then put on fake eyes. Does your brain then say,
a) "it looks fake altogether", or
b) "it's a real photo, but with fake eyes".
Then move on to lips, then hair, skin, etc, until you find the 'trigger'. Maybe it's a cumulative thing, or maybe something that can't be added or hidden, like ambient occlusion, which likely is never captured properly by a rendered image. But it's something....and this something is subtle that the brain is seeing, and we can not yet definitively articulate with 100% clarity what it is.
Perhaps it's the pupil size with respect to surrounding light?....Maybe something like that, something where you can't just put it into words to an accurate metric with the snap of a finger, but your brain can compute it instinctively.
The bump mapping on the lips is not different from the rest of the face. The freckles are nice, but the bump mapping of the lips throws it off. As well, the whites of the eyes are too white, almost luminous.
In the second photo, the hands are the problem. There's no life to them. Hands are the dead giveaway for all CGI to date. Squaresoft (Square-Enix, etc) have always had the best video game CGI, and even their Final Fantasy movies have been great, except for the hands. The hands always ruin it because no one can do realistic hands. Even the latest Marvel movies fell victim to the hands for the CGI stuff.
If you tell a model to pose for a shot, they hold a pose. If you snap a shot of someone performing an action, the result is drastically different.
When you look at a magazine or some other posed shot, you expect that to have used a human model, and you expect that model was made-up and the original photo heavily altered. It's no big secret these days like it used to be, so we expect that.
There's a thread running around here where someone's 3D work was featured in Vanity Fair, and the image looks like a magazine photo. Yes, it's clearly doctored and emphasized in a similar manner, but it looks like the photos they doctor up for magazines. That's an important step in getting 3D to look real.
I’m not sure how it happened, but for lots of reasons I think she does look hyper realistic and has managed to convince people. I think also that because your eye is so trained you’ll never be tricked.
Female faces in particular of 3d characters are pretty stylized. They are easy to spot and it’s not because I use 3d. It’s because of the way they look and the proportions and the way the face fat moves even the shape and size of the head and skull
I think that we notice things others don't. Sometimes I show my more realistic renders to people who know nothing about 3D rendering and they just can't get it through their head that it's not a real person. When I tell them the girl doesn't exist, she's just pixels and polygons, their brain kind of explodes. They think I'm a photographer and I have to keep explaining that no, there never was a real girl, she just doesn't exist. Often making the image black and white or using Instagram type filters make the images look more real than a straight render.
CG Female faces often almost seem symbolic. Rather like someone took a slightly stylized very 2d illustration of a face and then tried to reconstruct a 3d face from it. Like the nose, say, can't stray to far from the platonic ideal of what a nose is or it won't be recognizable.
You add that to the maket pressure for the female characters to be pretty, and...
Of couse the weirdest part from my perspective, is I personally find that there's way more variation that women can have and still be aestetically appealing... I have a way more specific type for men.
Here's one of the few images that made me look twice and had to really LOOK at it before I realized it was a 3D render. On first glance though, it had me fooled for a minute.
The body folds help. And I think the fact that it's kind of an unnatural awkard pose actually helps "sell it" as well - as most real people trying to sit in that pose will look a bit stiff or unnatural; so the fact that it's not really a normal pose that someone would sit in, helps - as we don't often see people sitting like that so we don't have a "quick reference" file in our memories to instantly compare it to other images we've seen with real people in that pose. The fact that the face is partially inverted also helps "sell it", I think, as we don't often see faces that are up-side-down. So our brain doesn't immediately recognise artificialness that would be easier to spot with a face that's vertical or in a position we normally see faces in.
I think the lump on the back definitely doesn't look right to me, and the hair is definite give away. Still, just browsing past it among other "photos" it wouldn't jump out immediately (at least to me) as 3D.
Thanks for your comments! I anticipated a few possible critiques, but that first one about the lips wasn't one of them, perhaps because they stand out at least somewhat better on the bump map itself (attached in case you're curious).
After reading your comments, I Iooked up a real photo of human eyes with reasonably similar lighting, sampled the eyewhite in a well-lit spot, and looked at the HSB values. I then sampled the eyewhite in a well-lit spot on the above render and looked at its HSB values. Strangely enough, the brightness of my rendered eye actually seemed to be lower than that of the photographed eye. The former hovered on average around the high 60s or low 70s, while the latter was consistently in the low 80s. This appeard to be the trend after taking a few samples from a couple different eye photos.
Of course, there's no guarantee that the lighting was sufficiently equivalent, but it seems that one or both of us have skewed expectations for sclera color. To me, the eyes of many 3D characters look a bit too dark, perhaps in a misguided attempt to bring out the reflections better. I also wonder if the contrast with my character's rather tanned skin might be contributing to the discrepancy in our perception.
Although I'm not sure I see what you see with the hands, I don't think there's much I can do about them yet.
I think she looks very convincing, Diva, lump-on-the-back hair and all. Could definitely have me fooled. What gives it away is the fountain, which just looks totally fake.
Annoyingly, I learn from perusing the comments of the DeviantArt page where she's posted, that her hair - WildHair by DigiCalimero apparently - is no longer available. Pity, I could really use a hair like that.