IRAY Photorealism?

davidtriune

are you talking about this one? https://www.daz3d.com/forums/discussion/237931/photo-real-characters-a-different-approach/p1

davidtriune

bleach2290 said:

Stumbled onto this thread and figured i'd drop one of mine in here for some feed back myself.

First one is Ungraded with no post work done inside daz with nothing but the iray shader with a hdr for lighting ,the others are just Luts that i thoguht would look half decent.

Let me know what you think and any feedbackw ould be nice always looking to improve.

detail wise I don't know what to tell you, it's hard to impress with default-looking skins

but if you want feedback on the tone, I like the third one maybe because he blends into the bg better.

davidtriune

shavonnew said:

Does she look photo real? Somehow I think she looks a little more real here in the studio setting...but still not real enough!
Attached a half-assed render cause I got bored of waiting.

My guess is the first one has nice eye reflections so it makes her "feel" more real than the second, that happens to me sometimes.

Her eyes probably look unnaturally shaped. But I don't know. I have to open the model myself to find out.

Masterstroke

NylonGirl said:

I remember some time ago, in another thread, there were people posting images said to have a high degree of realism. And there were other people who would list all of the flaws in the images they could think of. The person who created the thread, which I can no longer find, observed that for all of the things they listed as obvious reasons images were fake (poor lighting, perfect lighting, lack of emotion, etc), one could go onto the internet and easily find an example of an actual person who was poorly lit, perfectly lit, or lacked emotion, or any number of other things, yet was clearly a real person. So it became clear that whatever was the key to realism, was not obvious.

I also noticed a recurring theme was people saying there was something wrong with the eyes of all DAZ figures. There was also a meme on the social networks that stated the Disney Movie, Polar Express, had CG characters with “lifeless” eyes. I also tried showing images to my dad, who has no interest in computer generated images as a hobby, and nothing to do with DAZ or Poser. With the very first image I showed him, he seemed surprised when I told him it was computer generated. It was a small image so it would be hard to look closely at it, and it was not the main focus of our conversation. But with every other image I showed him, he said it was obvious it wasn’t a real person. When I told him many people would say the images were not realistic enough yet nobody could seem to say exactly why, he said the eyes were not right.

I am not one to try to create my own characters and textures and whatnot. I leave that to people like @jeff_someone and @Bluejaunte. I use products that I think produce good results. And I don’t have any special tricks. But I do have two working theories based on the things I mentioned above. One is that a key factor in people believing an image is a photograph of a real person, is to not tip them off that the image is a computer generated person. They could find flaws in a person standing in front of them if they were convinced it wasn't a real person. And the other theory is the eyes have issues. But I don’t know what kind of issues they have. I think if I want to convince people my image is a photograph then the figure had better be wearing glasses or not looking at the camera.

@Soldato mentioned a desire to create images like jeff_someone except with a “2020 look”. Also, there was some conversation about HDRI backgrounds not matching the figures in the image. I noticed the current trend in high end phone cameras is a blur effect, and some large lenses on SLR camera have an inherent “bokeh” effect. So it seems the trend in the high end cameras of 2019 and 2020 is to have a background that is unnaturally more blurred than the foreground image. Perhaps we’ve reached a point that the mismatched HDRI backgrounds will reasonably simulate the blur/bokeh effects of the current high-end cameras and produce the desired “2020 look”. I am not going to spellcheck all of this.

I see your point, and you're right. We've learnt to look closer. In times of earley CG in movies, we were less critical. We were more likeley to accept those special effects as real. Movies, that were "WOW!" in the old days, look sometimes quite ridicolous now. E.g. I watched the original Jurrasic Park recentley, and CG did not appear as amazing, as it did when I first saw it.

bluejaunte

We humans like to find the fault in one thing, it's easier to grasp. The truth is very likely that everything need to be perfect to look photoreal. And that is awfully hard to do.

Masterstroke

bluejaunte said:

We humans like to find the fault in one thing, it's easier to grasp. The truth is very likely that everything need to be perfect to look photoreal. And that is awfully hard to do.

As I always point out: I am talking of "Photo-real" not "Eye-real. "Eye-real" is nearly impossible to achieve. Photo-real should be doable. We recognize the worst photo-qualities as Photo-Real - even post colored black and white photos from the 1940s.
... And still our best Iray renders look, no matter how amazing they are, like renders.

shavonnew

It seems a lot easier to achieve realism with dark skin than it is with light skin. Did not even bother with SSS for dark skin. 
Light skin is such a pain though.

notiusweb

I'm going to follow through on what NylonGirl said, because we are starting to progress here.  You have to identify literal things that the brain sees.  You know, if you never did Ballroom dancing you can watch Dancing with the Stars and be like "Wow, that was so good when so-and-so danced, right!???!!!!", but then when you know how to dance, you look at it and see every point where that contestant messed up and didn't hold there frame, was off timing, positioned the head incorrectly, was slouching their mid section.  And you can do this even if uyou yourself are susceptable to these errors.  But it almost becomes an art in itself to troubleshoot what's wrong with art.  But it seems that is what reveals truth...not blanket praise, but focussed insight.  The point of this thread is Iray Photorealism. 

The biggest thing I found in this thread is that you see the difference with poster's icons, as in this forum poster's icon pics.  Like mine looks like a render, easy, and Masterstroke's does not.  But which has more color.  JCade's eye looks like a render, Rashad Carter's face icon looks real.  Again, look at the fine detail discrepancy.  This difference is what, though, because at the pixel resolution level it is likely uniform.  And like I can tell NylonGirl's icon is a render.  But Shavonnew, I really am not sure, looks real.  But even if there was no background, like a black background, my brain gives Shavvonew's icon the edge in realism, to the point where it looks real.  I think it is real, right?  

 

 

shavonnew

Yeah my icon is me lol
I've definitely found a lot of photo references where the skin barely has any details but the model is clearly real. It's quite frustrating trying to figure out the difference.

Gordig

I’m sure you’re all at least vaguely familiar with the concept of the Uncanny Valley, but I’d encourage you to read about it in more depth. Long story short, our brains are hard-wired to seek out and recognize human faces, especially eyes. The more lifelike a human face is, the harder our brains work to determine whether or not it is real. BlueJaunte was also right in that realism is an all-or-nothing proposition: you could absolutely nail a person’s face aside from one detail, but that one detail is all your brain needs to determine that the face it’s seeing isn’t human.

shavonnew

I think what's frustrating is you can see all kinds of features and face characteristics on people around the world. Big noses, small eyes, wide jawlines etc. Disfigured people don't suddenly appear fake. They just look... disfigured.
 

davidtriune

 

Masterstroke said:

bluejaunte said:

We humans like to find the fault in one thing, it's easier to grasp. The truth is very likely that everything need to be perfect to look photoreal. And that is awfully hard to do.

As I always point out: I am talking of "Photo-real" not "Eye-real. "Eye-real" is nearly impossible to achieve. Photo-real should be doable. We recognize the worst photo-qualities as Photo-Real - even post colored black and white photos from the 1940s.
... And still our best Iray renders look, no matter how amazing they are, like renders.

True. But we are sort of getting there. A guy on this thread refused to believe Jeff's render was not a photo until he showed the wireframe lol

I think if we take it one step at a time, we can eventually get there. First photoreal then eyereal. 

davidtriune

shavonnew said:

It seems a lot easier to achieve realism with dark skin than it is with light skin. Did not even bother with SSS for dark skin. 
Light skin is such a pain though.

dark skins happen to have a lot less SSS than light skin. thats why you see veins and purplish colors and stuff underneath white skinned people but not black skinned people

notiusweb

Gordig said:

I’m sure you’re all at least vaguely familiar with the concept of the Uncanny Valley, but I’d encourage you to read about it in more depth. Long story short, our brains are hard-wired to seek out and recognize human faces, especially eyes. The more lifelike a human face is, the harder our brains work to determine whether or not it is real. BlueJaunte was also right in that realism is an all-or-nothing proposition: you could absolutely nail a person’s face aside from one detail, but that one detail is all your brain needs to determine that the face it’s seeing isn’t human.

Oooh...You guys are right, there are 2 independant things maybe we cross at times

(1) Photo-real vs render

(2) Human vs virtual human (AI, robot)

See, because we could make a photo-real version of an Uncanny Valley mannequin, and likewise, separately render a lifelike human.  The 2 interactions are not the same, albeit they can overlap.

I think we are ultimately talking about the 1st, photo-real vs render.  We can see a rendered model with eyes closed and klnow it is a render. Also, a person with dark sunglasses can appear photo-real.  Also, Bluejaunte can't explain photo-real eye vs render eye - What is the difference if "all-or-none".  

bluejaunte

I'm talking about photoreal, as in what we see in photos or movies. I don't see how we can get eye-real (if I understand correctly that is what our eyes see directly?) unless we plug a source directly into our visual cortex.

Leonides02

bluejaunte said:

I'm talking about photoreal, as in what we see in photos or movies. I don't see how we can get eye-real (if I understand correctly that is what our eyes see directly?) unless we plug a source directly into our visual cortex.

Go on.... 

notiusweb

bluejaunte said:

I'm talking about photoreal, as in what we see in photos or movies. I don't see how we can get eye-real (if I understand correctly that is what our eyes see directly?) unless we plug a source directly into our visual cortex.

Let's even take this further then.  What is the difference between the human real eye and Uncanny valley eye in a photo.  Anyone use a desription if you can.  What do we see as different between the 2.  I say, for this topic, we see no difference because our brain knows it is a real photo, as in we know it existed at some time in some place, where it was captured by a photo.  You know, why is this...because our brain will say, "Tht's not a *real* human, but at the same time, "That's a real life capture..."

But in any event, Uncanny vs real-human is not so important to this topic as much as it is a different interesting discussion in its own right.  We want the difference between photo vs render, and then from there we address how to get that from Iray.

bluejaunte

notiusweb said:

bluejaunte said:

I'm talking about photoreal, as in what we see in photos or movies. I don't see how we can get eye-real (if I understand correctly that is what our eyes see directly?) unless we plug a source directly into our visual cortex.

Let's even take this further then.  What is the difference between the human real eye and Uncanny valley eye in a photo.  Anyone use a desription if you can.  What do we see as different between the 2.  I say, for this topic, we see no difference because our brain knows it is a real photo, as in we know it existed at some time in some place, where it was captured by a photo.  You know, why is this...because our brain will say, "Tht's not a *real* human, but at the same time, "That's a real life capture..."

But in any event, Uncanny vs real-human is not so important to this topic as much as it is a different interesting discussion in its own right.  We want the difference between photo vs render, and then from there we address how to get that from Iray.

Don't think there is one honestly. Your reaction to this

is going to be mostly the same wether it's a photo or you'd be standing next to it. I mean, obviously the latter might be even creepier but that would just be the normal effect of seeing it "in the flesh", just like a photo of a spider isn't going to invoke the same horror as having it sitting on the table in front of you. But the instinct that something is very wrong here is going to happen in both cases.

It still looks like a photo to me though, not a render. If that was your point?

davidtriune

notiusweb said:

bluejaunte said:

 

Let's even take this further then.  What is the difference between the human real eye and Uncanny valley eye in a photo.  Anyone use a desription if you can.  What do we see as different between the 2.  

The photo above is obviously not anatomically correct. So I guess being anatomically correct should avoid the uncanny valley? 

I guess that means we should study up on anatomy? 

notiusweb

bluejaunte said:

notiusweb said:

bluejaunte said:

It still looks like a photo to me though, not a render. If that was your point?

"It still looks like a photo to me though, not a render. If that was your point?"

 

Okay, exactly, yes, that was the idea I was driving at.  That is a perfect example you gave.  I can almost peripherally summize in my mind what approxiamtely the environment was around this faux-human.  And instantly I do not see it as a render, you know?  I did look at the eyes and determine it was not a real living person, but just as fast I knew it was 'real' in the universe.  Which seems to then redefine 'photo-real' as not neccessarily having much to do with *accurate* detail on a model.   Like, I have never made a render, myself, where I said, "That's not a living-breathing person for sure...but wow it sure looks photo real!"  

Trying to work backwards, here is photo that I found that looks like a really really good render - but no - it is in fact a human...there are other pics on the webpage of her:

 

 

bluejaunte

So you're trying to differentiate between a photoreal looking photo and a photoreal looking human? Yeah I guess that is an interesting aspect.

j cade

Per Rashad's suggestion I've put a character in a backdrop rather than a windowless cube, Its my third or so attempt. Characters full scenes is pretty much exponentially more work for me, especially since as soon as I start adding in scenery I lose what little sense of restraint I have. At least this one is otdoors and not primarily lit by light coming through a window. (that would be the first one I started), Or at night (the second). Also apparently I can spend all this effort on hair and skin and then go "ehh close enough" for the pose... untill the render is 3/4ths of the way done and I begin to feel the pangs of regret

Theres still plenty of extraness though, For instance the skin is once again 100% SSS and no diffuse. It can be done! Dark skin in Iray with all the SSS that isnt a dark red mess, nor is there rudolph nose! (Its rll fiddly though lots of minute tweaking of just off white tints so the undertones don't go funky). The hair came very close to completely halting my computer, blender is still definitely more memory efficient at hair than Iray. Also it does curves so I can do hair like this without any pointy bits. Tiny detail that is probably unnoticable to anyone whos not me, but i love the velus hair on the back of the neck. On the other hand I was really unsure about the specular in the shadowed part of the neck, It feels cg, even though while looling at photos of OBJ for hair inspiration I noticed the same thing in several photos, but it still feels off even if I can find similar examples in the real world I think theres been some discussion in this thread espousing similar ideas.

 

 

...And of couse as soon as I was finished rendering I decided I prefered the the scene from a different angle (and improved the pose ofthe left hand). not a final render smallering and using the Iray denoiser so theres some bluring, but look at the pretty backlighting on the hair Thats some #Aestetic there

 

Leonides02

Part of it is the subtle asymmetry in virtually every aspect of the face, I think: Eye size / shape, eye brows, lip contour, nostrils, etc. They're all subtly different.

notiusweb

bluejaunte said:

So you're trying to differentiate between a photoreal looking photo and a photoreal looking human? Yeah I guess that is an interesting aspect.

No, not discerning (1) a 2nd-hand image of an image of a real person, vs (2) an image of the actual person 1st hand.  I think photo-ing a photo only diminishes the detail/resolution, but it still looks real if it was a real photo to begin with.  

Moreso what I am saying now is that a human- and any assocaited details, shapes, poses, etc - is seemingly not an element of the equation of "photo-real" at all.  Because consider this- you can have (a) an image like you posted, a picture of a faux-human, which our brain says "real / photo-real...", but then you at the same time can have (b) an image of a lifelike human, like JCade just posted, which our brain says "Render".  

Now, we know that Iray can render-produce a flat image plain of an actual photo, so we know in fact Iray can render photo-real.  Yet, we also know that not all Iray output looks photo real...Both could be of an apple on a table, it has nothing to do with a human...

So, might there be some "pixel-pattern" found in lighting-computed-renders not found in real-photos, and vice versa a "pixel-pattern" found in real-photos not found in lighting-computed-renders.  And users in this forum have given some observations to that pixel point (blurring, smaller color variation), but we haven't figured out its mechanic yet.  

emoryahlberg

I took an amalgam of everything I learned from this excellent thread and put them together in one render (minus hair, because I'm focusing on skin). I also worked very diligently on teeth / gum shaders and eyes because to me they're a dead give away. 

Thank to you all, I think this is my best effort so far. Would love any feedback.

 

nonesuch00

j cade said:

Per Rashad's suggestion I've put a character in a backdrop rather than a windowless cube, Its my third or so attempt. Characters full scenes is pretty much exponentially more work for me, especially since as soon as I start adding in scenery I lose what little sense of restraint I have. At least this one is otdoors and not primarily lit by light coming through a window. (that would be the first one I started), Or at night (the second). Also apparently I can spend all this effort on hair and skin and then go "ehh close enough" for the pose... untill the render is 3/4ths of the way done and I begin to feel the pangs of regret

Theres still plenty of extraness though, For instance the skin is once again 100% SSS and no diffuse. It can be done! Dark skin in Iray with all the SSS that isnt a dark red mess, nor is there rudolph nose! (Its rll fiddly though lots of minute tweaking of just off white tints so the undertones don't go funky). The hair came very close to completely halting my computer, blender is still definitely more memory efficient at hair than Iray. Also it does curves so I can do hair like this without any pointy bits. Tiny detail that is probably unnoticable to anyone whos not me, but i love the velus hair on the back of the neck. On the other hand I was really unsure about the specular in the shadowed part of the neck, It feels cg, even though while looling at photos of OBJ for hair inspiration I noticed the same thing in several photos, but it still feels off even if I can find similar examples in the real world I think theres been some discussion in this thread espousing similar ideas.

 

 

...And of couse as soon as I was finished rendering I decided I prefered the the scene from a different angle (and improved the pose ofthe left hand). not a final render smallering and using the Iray denoiser so theres some bluring, but look at the pretty backlighting on the hair Thats some #Aestetic there

 

I like 2nd angle better too

Leonides02

emoryahlberg said:

I took an amalgam of everything I learned from this excellent thread and put them together in one render (minus hair, because I'm focusing on skin). I also worked very diligently on teeth / gum shaders and eyes because to me they're a dead give away. 

Thank to you all, I think this is my best effort so far. Would love any feedback.

 

This looks great!

By the way, everyone: You can easily control gloss / specular by rendering out different canvas passes and then combining diffuse, specular, gloss in Photoshop. 

Masterstroke

emoryahlberg said:

I took an amalgam of everything I learned from this excellent thread and put them together in one render (minus hair, because I'm focusing on skin). I also worked very diligently on teeth / gum shaders and eyes because to me they're a dead give away. 

Thank to you all, I think this is my best effort so far. Would love any feedback.

 

This looks awesome. I especially do like the teeth.

notiusweb

@JCade - I tried my hand at doing the Filmic looking color thing you talked about, but in Gimp, with your image.   When I saw the green pole things I was imagining more shine and less diffuse.  But then it made the hair look cool too, I think?  Really impressed by your art, JCade!

 

 

jeff_someone

nonesuch00 said:

j cade said:

Per Rashad's suggestion I've put a character in a backdrop rather than a windowless cube, Its my third or so attempt. Characters full scenes is pretty much exponentially more work for me, especially since as soon as I start adding in scenery I lose what little sense of restraint I have. At least this one is otdoors and not primarily lit by light coming through a window. (that would be the first one I started), Or at night (the second). Also apparently I can spend all this effort on hair and skin and then go "ehh close enough" for the pose... untill the render is 3/4ths of the way done and I begin to feel the pangs of regret

Theres still plenty of extraness though, For instance the skin is once again 100% SSS and no diffuse. It can be done! Dark skin in Iray with all the SSS that isnt a dark red mess, nor is there rudolph nose! (Its rll fiddly though lots of minute tweaking of just off white tints so the undertones don't go funky). The hair came very close to completely halting my computer, blender is still definitely more memory efficient at hair than Iray. Also it does curves so I can do hair like this without any pointy bits. Tiny detail that is probably unnoticable to anyone whos not me, but i love the velus hair on the back of the neck. On the other hand I was really unsure about the specular in the shadowed part of the neck, It feels cg, even though while looling at photos of OBJ for hair inspiration I noticed the same thing in several photos, but it still feels off even if I can find similar examples in the real world I think theres been some discussion in this thread espousing similar ideas.

 

 

...And of couse as soon as I was finished rendering I decided I prefered the the scene from a different angle (and improved the pose ofthe left hand). not a final render smallering and using the Iray denoiser so theres some bluring, but look at the pretty backlighting on the hair Thats some #Aestetic there

 

I like 2nd angle better too

Nice job... only recommendations would be to lessen the DOF effect.   Also, its rare, but possible, to have DOF in the fore and background as you do (e.g., the bar is out of focus but so is the background).  Can happen... just to me, makes my eyes wonder whats going on.