IRAY Photorealism?

1323335373867

Comments

  • PadonePadone Posts: 3,481
    edited June 2020

    @isidorekeeghan Volume transmission as well as scattering is better tested with strong backlights on thin objects. For example a human ear.

    In your settings you use a semi-white transmission with a quite high volume density. From a pbr point of view this means that the inner color of the volume is white and you only get the scattering color mixed with the translucency layer.

    I'm not sure how this simplification may affect the sss "realism". But given that the chromatic mode seems buggy and not documented anything may happen.

     

    edit. This is obvious but of course a 0.9 translucency is not good for hair and dust. The G8 makeups use a diffuse overlay over the translucency layer so this may work fine, but it is very likely that some adjustements may be needed to match the skin color. As an alternative a translucency map may be used in the weight channel.

     

    edit. There's also something very odd with the natural spectral. From what I understand it should be something with slight color variations from faithful, while it seems to lose any light falloff at all, other than light energy. That's not good for studio lighting or any situation where some dramatic lighting is needed.

    https://raytracing-docs.nvidia.com/iray/manual/concept/spectral_rendering.html

    Below an example with your settings on G8F. First rgb, then faithful, then natural. Scene included.

     

    edit. Just out of curiosity this is how she gets in cycles with the diffeomorphic plugin. That's of course similar to iray rgb since cycles doesn't support spectral rendering.

    rgb.jpg
    480 x 270 - 20K
    faithful.jpg
    480 x 270 - 21K
    natural.jpg
    480 x 270 - 11K
    duf
    duf
    test.duf
    132K
    cycles.jpg
    480 x 270 - 11K
    Post edited by Padone on
  • isidorekeeghanisidorekeeghan Posts: 28
    edited June 2020
    Padone said:

    @isidorekeeghan Volume transmission as well as scattering is better tested with strong backlights on thin objects. For example a human ear.

    In your settings you use a semi-white transmission with a quite high volume density. From a pbr point of view this means that the inner color of the volume is white and you only get the scattering color mixed with the translucency layer.

    I'm not sure how this simplification may affect the sss "realism". But given that the chromatic mode seems buggy and not documented anything may happen.

     

    edit. This is obvious but of course a 0.9 translucency is not good for hair and dust. The G8 makeups use a diffuse overlay over the translucency layer so this may work fine, but it is very likely that some adjustements may be needed to match the skin color. As an alternative a translucency map may be used in the weight channel.

     

    edit. There's also something very odd with the natural spectral. From what I understand it should be something with slight color variations from faithful, while it seems to lose any light falloff at all, other than light energy. That's not good for studio lighting or any situation where some dramatic lighting is needed.

    https://raytracing-docs.nvidia.com/iray/manual/concept/spectral_rendering.html

    Below an example with your settings on G8F. First rgb, then faithful, then natural. Scene included.

     

    edit. Just out of curiosity this is how she gets in cycles with the diffeomorphic plugin. That's of course similar to iray rgb since cycles doesn't support spectral rendering.

    The problem is your body light, its too big it washes out other shadows, and you have to increase all of the lights intensity for spectral rendering on, i highly recommend setting everything for spectral rendering on including the lights then swich to spectral rendering off if you like ,also you can change her sss color hue to 35 for more redness ,and thank you for the nice and interesting setup . Edit: also ive seen others pick a color from skin texture and put it in transmitted color or put the entire map in there(which is pointless because transmitted color doesnt even read the maps colos, go ahead and make translucency color mapless and white then put any colored map in transmitted color and see for yourself, spoiler alert its colorless) its unneccecery to put any color in there because the transmitted color is controlled by translucency color.
    TEST LIGHT.png
    735 x 413 - 843K
    Post edited by isidorekeeghan on
  • magaremotomagaremoto Posts: 1,226
    edited June 2020

    my latest attempts to get natural skin via glossiness/specular mode

    wh1.jpg
    720 x 960 - 256K
    wh4.jpg
    720 x 960 - 269K
    wh2.jpg
    720 x 960 - 304K
    wh5.jpg
    720 x 960 - 366K
    Post edited by magaremoto on
  • PadonePadone Posts: 3,481
    edited June 2020

    The problem is your body light, its too big it washes out other shadows, and you have to increase all of the lights intensity for spectral rendering on

    I don't know .. faithful works just fine and it's spectral. It's natural alone that doesn't work. But I'm not a spectral expert I'd really like to hear from @RayDAnt, or anyone with some experience about this. I feel natural just doesn't work right. From a pbr point of view saying "you have to fit your lights" is nonsense no offence intended. Lights are just what they are they don't need to be fitted to a rendering method.

    Post edited by Padone on
  • isidorekeeghanisidorekeeghan Posts: 28
    edited June 2020
    Padone said:

    The problem is your body light, its too big it washes out other shadows, and you have to increase all of the lights intensity for spectral rendering on

    I don't know .. faithful works just fine and it's spectral. It's natural alone that doesn't work. But I'm not a spectral expert I'd really like to hear from @RayDAnt, or anyone with some experience about this. I feel natural just doesn't work right. From a pbr point of view saying "you have to fit your lights" is nonsense no offence intended. Lights are just what they are they don't need to be fitted to a rendering method.

    Okey you do whatever is more suitable for you , but spectral rendering natural is more physically accurate than other methods. Edit: also you should know that there is nothing physically based about point lights or other lights that come with daz studio, and if something was wrong with spectral rendering natural then how come it doesnt effect the brighness of hdri maps?
    Post edited by isidorekeeghan on
  • nonesuch00nonesuch00 Posts: 17,929
    Padone said:

    @isidorekeeghan Volume transmission as well as scattering is better tested with strong backlights on thin objects. For example a human ear.

    In your settings you use a semi-white transmission with a quite high volume density. From a pbr point of view this means that the inner color of the volume is white and you only get the scattering color mixed with the translucency layer.

    I'm not sure how this simplification may affect the sss "realism". But given that the chromatic mode seems buggy and not documented anything may happen.

     

    edit. This is obvious but of course a 0.9 translucency is not good for hair and dust. The G8 makeups use a diffuse overlay over the translucency layer so this may work fine, but it is very likely that some adjustements may be needed to match the skin color. As an alternative a translucency map may be used in the weight channel.

     

    edit. There's also something very odd with the natural spectral. From what I understand it should be something with slight color variations from faithful, while it seems to lose any light falloff at all, other than light energy. That's not good for studio lighting or any situation where some dramatic lighting is needed.

    https://raytracing-docs.nvidia.com/iray/manual/concept/spectral_rendering.html

    Below an example with your settings on G8F. First rgb, then faithful, then natural. Scene included.

     

    edit. Just out of curiosity this is how she gets in cycles with the diffeomorphic plugin. That's of course similar to iray rgb since cycles doesn't support spectral rendering.

    I reported that bug last September I think it was. I don't remember the response as most of my bug reports and feature requests are accepted and then get fixed consequently as peripheral changes of their other work and I don't get notified.

    Also, if you ever change your DAZ 3D account email address you loose access to your old bug reports.

    I've yet to see any feature request I've made been implemented but those things take time and for some of them a bit of work though not excessive work.

  • PadonePadone Posts: 3,481
    edited June 2020
    Padone said:

    There's also something very odd with the natural spectral

    I reported that bug last September I think it was. I don't remember the response as most of my bug reports and feature requests are accepted and then get fixed consequently as peripheral changes of their other work and I don't get notified.

    So you do agree it's a bug thank you I was just wondering. I submitted a bug report myself may be this helps them to get it better.

     

    spectral rendering natural is more physically accurate than other methods .. also you should know that there is nothing physically based about point lights or other lights that come with daz studio, and if something was wrong with spectral rendering natural then how come it doesnt effect the brighness of hdri maps?

    Well the iray docs tell exactly the opposite that faithful is more accurate while natural sacrifices some tints to be more "smooth". Then tristimulus to wavelenght conversion is not physically accurate by definition that's why we have different conversion intents. The truth is unless we get spectral data for materials and lights it's all a "best guess". Also I'm not using point lights but sphere lights that have photometric properties so they are intended to be physically correct, though using ies would be better. I agree that point lights are to be avoided. I also agree hdris seem to work fine so the bug is limited to scene lights.

    https://raytracing-docs.nvidia.com/iray/manual/concept/spectral_rendering.html

    Indeed there's something odd with photometric properties in daz studio since I need 15000 lumens in daz studio to make 1 watt in blender that's nonsense, it should be 15 lumens = 1 watt for incandescent bulbs. But that apart they seem to work fine.

     

    edit. Of course until they fix the natural conversion I'd advise to stay with faithful.

    Post edited by Padone on
  • nonesuch00nonesuch00 Posts: 17,929
    edited June 2020
    Padone said:
    Padone said:

    There's also something very odd with the natural spectral

    I reported that bug last September I think it was. I don't remember the response as most of my bug reports and feature requests are accepted and then get fixed consequently as peripheral changes of their other work and I don't get notified.

    So you do agree it's a bug thank you I was just wondering it. I submitted a bug report myself may be this helps them to get it better.

     

    spectral rendering natural is more physically accurate than other methods .. also you should know that there is nothing physically based about point lights or other lights that come with daz studio, and if something was wrong with spectral rendering natural then how come it doesnt effect the brighness of hdri maps?

    Well the iray docs tell exactly the opposite that faithful is more accurate while natural sacrifices some tints to be more "smooth". Then tristimulus to wavelenght conversion is not physically accurate by definition that's why we have different conversion intents. The truth is unless we get spectral data for materials and lights it's all a "best guess". Also I'm not using point lights but sphere lights that have photometric properties so they are intended to be physically correct, though using ies would be better. I agree that point lights are to be avoided. I also agree hdris seem to work fine so the bug is limited to scene lights.

    https://raytracing-docs.nvidia.com/iray/manual/concept/spectral_rendering.html

    Indeed there's something odd with photometric properties in daz studio since I need 15000 lumens in daz studio to make 1 watt in blender that's nonsense, it should be 15 lumens = 1 watt for incandescent bulbs. But that apart they seem to work fine.

     

    edit. Of course until they fix the natural conversion I'd advise to stay with faithful.

    What I noticed is that I had both Sun-Sky & photometric lights in my scene and that when I used the spectral the Sun-Sky stayed normal but all photometric lights seemed to stop working. I think that might be intentional and why they call it 'natural. 'Natural' because it works with the 'Natural' Sun-Sky lighting. I'm not sure though as they didn't tell me anything that I remember when I submitted the bug report.

    I posted an example of both back in Sep/Oct 2019 in another thread but I'll never find that post.

    Post edited by nonesuch00 on
  • PadonePadone Posts: 3,481
    edited June 2020

    @nonesuch00 Natural has nothing to do with the sun sky light. It is a method to convert rgb to wavelenghts. Again see the iray docs and/or google for "spectral rendering". I myself knew nothing about spectral until some days ago, but I have a strong technical background so luckily it's easy for me to find and understand tech notes.

    Since the faithful conversion works fine with scene lights there's no reason why natural shouldn't. I mean if spectral wouldn't include scene lights then faithful wouldn't work as well.

    https://raytracing-docs.nvidia.com/iray/manual/concept/spectral_rendering.html

    Post edited by Padone on
  • glennblackphotosglennblackphotos Posts: 160
    edited June 2020

     

    Padone said:

    @Jeff Again, as for lights and camera lens your renderings are not "realistic" at all. This last advice about not using depth of field is nonsense if you want to mimic a real camera. Even human eyes have focus. Then again your technique is good enough to resemble old 70s cameras with a strong flash in a dimmed environment, but that's all. And I do love your renderings especially the characters are very good.

    In professional photography the depth of field (focal lenght) is used to isolate the subject from the environment and the same is done in movies.

    As a photographer I totally agree with you, even the iphones have a portrait mode that creates a shallow dof.

    Post edited by Chohole on
  • isidorekeeghanisidorekeeghan Posts: 28
    edited June 2020
    Spectral rendering defines how colors and lights behave, in real life when light hits an object first it changes how the colors are seen second it changes how the light it self is seen there for colors change to a more palish color and light it self losses value to absorbtion and becomes darker when it scatters back, spectral rendering faithful only changes the color, and spectral rendering natural changes both color and light scatter, just like the real life light would, spotlight and other scene lights work fine in spectral rendering natural we just have to increase their intensity or luminance, comparison of spectral rendering methods attached
    spectral rendering comparison.png
    3840 x 720 - 3M
    Post edited by isidorekeeghan on
  • PadonePadone Posts: 3,481
    edited June 2020

    @isidorekeeghan If you use a single light the test fails because you can't compare falloffs among different scene lights. Below you can see the natural conversion where I used a 10x energy factor in tone mapping. Falloffs are lost as it was already clear in the darker picture in my first post above. That is, increasing the light energy doesn't fix anything, apart it being wrong anyway.

    10x.jpg
    499 x 101 - 16K
    natural.jpg
    480 x 270 - 19K
    Post edited by Padone on
  • Padone said:

    @isidorekeeghan If you use a single light the test fails because you can't compare falloffs among different scene lights. Below you can see the natural conversion where I used a 10x energy factor in tone mapping. Falloffs are lost as it was already clear in the darker picture in my first post above. That is, increasing the light energy doesn't fix anything, apart it being wrong anyway.

    In my test i used sunsky only and everything worked as normal , there is somtehing wrong with daz lights not iray, as i told you before one of your lights is too big thats whay you cant see sharp shadows, no matter how much you increase the intensity its still too big, i dont understant what the problem is here, if you want actuall physically based renders then you have to deal with the hassle of changing one light. No offence but because you dont understand it or dont know how to use it doesnt mean its broken or wrong. Again no offence
  • Theme3DTheme3D Posts: 27

    That's the most I could get with a Daz character. That's Mousso character btw. Hes characters are very realistc overall. And from there is definetelly possible to improve a lot of things to make this more realistc. Like peach fuzz, textures etc. The volume SSS definitely play against realism, it's unintuitive, the values are very confusing. And there is a strange bug with the gamma of some textures of the characters skins.

    The normal map of most characters doesn't really affect the final result like it should, the transmitance seems to wash all the detail. My computer crawls if I set subd more than 4, a good displacement is key to get a realistic result. All and all it's very possible, but very hard on daz studio.

    If you check https://www.youtube.com/watch?v=s3fDXZ-c6u8. There is really nothing that daz have different, just the renderer. And he maskes it look realistc so easy, because Arnold is a lot simpler and predictible.

     

     

    testrender.png
    1000 x 1000 - 2M
  • nonesuch00nonesuch00 Posts: 17,929

    For what it's worth as far as 'realism' goes, I definately, out of all the Genesis 8, PA or DO, characters I've bought, think the shaders settings for Dain 8 are the best. They look good on Dain 8's materials images used with those shader settings and they look good applied to other characters whereby you keep the original images used in the skin materials but apply Dain 8's shader settings.

    The other newest DO Pro Bundle characters, at least the ones the bundle was named after, likely use Dain 8's shader settings too. You'll know that's the case if a baby blue is one of the colors used in the surface settings. Dain't the newest Pro Bundle non-PA character I've bought so I ccan't check.

  • magaremotomagaremoto Posts: 1,226

    testing SSS and translucency response to different size

    test.jpg
    1093 x 897 - 107K
  • j cadej cade Posts: 2,310
    edited June 2020
    Padone said:

    @isidorekeeghan I don't feel your "proper" way is really "proper".

    What is with you people? His settings are gorgeous, they work, and he shared them with us for free. If you have better chromatic settings, I'm sure he'd love to see them. There's a lot of weird "this isn't proper PBR" and "this render engine is better" baloney in this thread. Isn't the point to make the most photoreal images as possible with iray?

     

     j cade that is a beautiful and, I daresay, photoreal render! Iray or Blender?

    Iray!

     

     

    I'm finding it increasibly frustrating that even when you get the perfect skin settings, they seem to only be perfect for ONLY certain lighting set ups. If I change things too drastically, the character ends up waaaay too dark and red.

    This seems to be an issue with chromatic skins.

    There are a couple comments here with a similar thing and generally speaking, If your skin changes relative to other objects depending on lighting then that is a clue something is up with your skin settings.

    Take this skin. Looks pretty good non?

    Right up until shes strongly backlit and we realize shes actually a glowing jello monster disguised as a human!

    Now this is obviously an extreme example as generally you arent rendering chaacters just backlit and nothing else, but even in less extreme versions, any lighting setup with backlighting the chest neck etc will be unnaturally brighter relative to a setup without backlighting, (not to mention you likely get some weird shadowing around the mouth and eyes)

    Strong side lighting is also likely to create rudolph nose

    This is basically my skin testing setup for the record.

    even neutral lighting for over all look and tone

    strongly backlit and raise transmitted distance as high as I can get it without the neck and torso starting to glow

    rembrandty key light to make sure the nose doesn't glow and also that shadows on the skin soften. (see how the edhes of the shadows are red? also if you so a reder close up there isnt an ege at all but a transition)

    this for me is a major key to making skin look soft and imo is subconsiously noticable in most lighting conditions, even though its rarely as distinct as here (its also great for a sort of painterly realism even if you dont hit photo real)

    FWIW This is the area where I'm having the most difficulty matching what I see in reality. While where the nose and head cast relatively crisp shadows is close to matching reality, the crisp shadow terminator edge is very much not. This should be much softer,  though I'm not sure is this is all down to sss or an effect of skin in reality being a much less smooth surface. If anyone has some suggestions for softening this up I am all ears

     

     

    TLDR PSA: when setting up your skin settings don't do it with just even lighting

    skin testing1.jpg
    700 x 1050 - 361K
    skin testing2.jpg
    700 x 1050 - 608K
    skin testing3.jpg
    700 x 1050 - 316K
    skin testingb1.jpg
    700 x 1050 - 362K
    skin testingb3.jpg
    700 x 1050 - 357K
    skin testingb2.jpg
    700 x 1050 - 303K
    Post edited by j cade on
  •  

     

    I totally agree that normal maps of most of the characters are useless, they dont have small details like pores and stuff, only one that has great details(that i know of) is underbelly8, displacement map is unnecessary if normal map and model itself are detailed enough, i highley recommend using auto face enhancer, its great ,it gives great details to the face and keeps polygon count low but it also doesnt work well on all characters, in the image below i used elithia 8, auto face enhancer and its hd details, multiple hd details from other characters and everything was set to subd level 2 and ofcourse my own sss method also used only sunsky and didnt do any post work, i realy dont like how most characters ears look , they look too basic and unnatural.

    Bluejaunte's characters have those kind of normal maps with pores and wrincles.
    I agree, it doesn't make sense to just grey scale a diffuse map in order to convert it to a bump map, as it makes no sense either to convert it to a normal map.
     

    Oh i love Bluejaunte's characters, theire very good and detailed , im sure there are others who make great characters i just cant remember the names , also the whole point of normal map is to replace the need for high subdevision level, if it doesnt have high level details then its useless, also i agree bump maps have the same function as normal maps , they must be made separately not made from diffuse maps
    j cade said:
    Padone said:

    @isidorekeeghan I don't feel your "proper" way is really "proper".

    What is with you people? His settings are gorgeous, they work, and he shared them with us for free. If you have better chromatic settings, I'm sure he'd love to see them. There's a lot of weird "this isn't proper PBR" and "this render engine is better" baloney in this thread. Isn't the point to make the most photoreal images as possible with iray?

     

     j cade that is a beautiful and, I daresay, photoreal render! Iray or Blender?

    Iray!

     

     

    I'm finding it increasibly frustrating that even when you get the perfect skin settings, they seem to only be perfect for ONLY certain lighting set ups. If I change things too drastically, the character ends up waaaay too dark and red.

    This seems to be an issue with chromatic skins.

    There are a couple comments here with a similar thing and generally speaking, If your skin changes relative to other objects depending on lighting then that is a clue something is up with your skin settings.

    Take this skin. Looks pretty good non?

    Right up until shes strongly backlit and we realize shes actually a glowing jello monster disguised as a human!

    Now this is obviously an extreme example as generally you arent rendering chaacters just backlit and nothing else, but even in less extreme versions, any lighting setup with backlighting the chest neck etc will be unnaturally brighter relative to a setup without backlighting, (not to mention you likely get some weird shadowing around the mouth and eyes)

    Strong side lighting is also likely to create rudolph nose

    This is basically my skin testing setup for the record.

    even neutral lighting for over all look and tone

    strongly backlit and raise transmitted distance as high as I can get it without the neck and torso starting to glow

    rembrandty key light to make sure the nose doesn't glow and also that shadows on the skin soften. (see how the edhes of the shadows are red? also if you so a reder close up there isnt an ege at all but a transition)

    this for me is a major key to making skin look soft and imo is subconsiously noticable in most lighting conditions, even though its rarely as distinct as here (its also great for a sort of painterly realism even if you dont hit photo real)

    FWIW This is the area where I'm having the most difficulty matching what I see in reality. While where the nose and head cast relatively crisp shadows is close to matching reality, the crisp shadow terminator edge is very much not. This should be much softer,  though I'm not sure is this is all down to sss or an effect of skin in reality being a much less smooth surface. If anyone has some suggestions for softening this up I am all ears

     

     

    TLDR PSA: when setting up your skin settings don't do it with just even lighting

    I do the exact same things, i like to compare my method with yours, can i find yours anywhere? , also are you using different values for the ears? Because it kinda shows, i tried that too but the difference is always visible. Attached comparisons of my sss setting on digital emily and a picture of emily herself, again no post work beside croping
    emily comparison.png
    572 x 769 - 759K
    emily comparison full.jpg
    3336 x 2500 - 5M
  • j cadej cade Posts: 2,310
    edited June 2020
     

     

    I totally agree that normal maps of most of the characters are useless, they dont have small details like pores and stuff, only one that has great details(that i know of) is underbelly8, displacement map is unnecessary if normal map and model itself are detailed enough, i highley recommend using auto face enhancer, its great ,it gives great details to the face and keeps polygon count low but it also doesnt work well on all characters, in the image below i used elithia 8, auto face enhancer and its hd details, multiple hd details from other characters and everything was set to subd level 2 and ofcourse my own sss method also used only sunsky and didnt do any post work, i realy dont like how most characters ears look , they look too basic and unnatural.

    Bluejaunte's characters have those kind of normal maps with pores and wrincles.
    I agree, it doesn't make sense to just grey scale a diffuse map in order to convert it to a bump map, as it makes no sense either to convert it to a normal map.
     

     

    Oh i love Bluejaunte's characters, theire very good and detailed , im sure there are others who make great characters i just cant remember the names , also the whole point of normal map is to replace the need for high subdevision level, if it doesnt have high level details then its useless, also i agree bump maps have the same function as normal maps , they must be made separately not made from diffuse maps
    j cade said:
    Padone said:

    @isidorekeeghan I don't feel your "proper" way is really "proper".

    What is with you people? His settings are gorgeous, they work, and he shared them with us for free. If you have better chromatic settings, I'm sure he'd love to see them. There's a lot of weird "this isn't proper PBR" and "this render engine is better" baloney in this thread. Isn't the point to make the most photoreal images as possible with iray?

     

     j cade that is a beautiful and, I daresay, photoreal render! Iray or Blender?

    Iray!

     

     

    I'm finding it increasibly frustrating that even when you get the perfect skin settings, they seem to only be perfect for ONLY certain lighting set ups. If I change things too drastically, the character ends up waaaay too dark and red.

    This seems to be an issue with chromatic skins.

    There are a couple comments here with a similar thing and generally speaking, If your skin changes relative to other objects depending on lighting then that is a clue something is up with your skin settings.

    Take this skin. Looks pretty good non?

     

    Right up until shes strongly backlit and we realize shes actually a glowing jello monster disguised as a human!

     

    Now this is obviously an extreme example as generally you arent rendering chaacters just backlit and nothing else, but even in less extreme versions, any lighting setup with backlighting the chest neck etc will be unnaturally brighter relative to a setup without backlighting, (not to mention you likely get some weird shadowing around the mouth and eyes)

    Strong side lighting is also likely to create rudolph nose

     

    This is basically my skin testing setup for the record.

    even neutral lighting for over all look and tone

     

    strongly backlit and raise transmitted distance as high as I can get it without the neck and torso starting to glow

     

    rembrandty key light to make sure the nose doesn't glow and also that shadows on the skin soften. (see how the edhes of the shadows are red? also if you so a reder close up there isnt an ege at all but a transition)

     

    this for me is a major key to making skin look soft and imo is subconsiously noticable in most lighting conditions, even though its rarely as distinct as here (its also great for a sort of painterly realism even if you dont hit photo real)

    FWIW This is the area where I'm having the most difficulty matching what I see in reality. While where the nose and head cast relatively crisp shadows is close to matching reality, the crisp shadow terminator edge is very much not. This should be much softer,  though I'm not sure is this is all down to sss or an effect of skin in reality being a much less smooth surface. If anyone has some suggestions for softening this up I am all ears

     

     

    TLDR PSA: when setting up your skin settings don't do it with just even lighting

     

    I do the exact same things, i like to compare my method with yours, can i find yours anywhere? , also are you using different values for the ears? Because it kinda shows, i tried that too but the difference is always visible. Attached comparisons of my sss setting on digital emily and a picture of emily herself, again no post work beside croping

    I'm using the same settings on the ears as the rest of her, I always edit as much as a unit as possible, so when editing sss i always select templates 1-4 (I even avoid having different spec settings on the lip from the rest of the face)

     

    I'm using settings pretty close to yours, the render I'm currently working on has

    spectral rendering on

    Translucency strength 1.00

    base colr effect: scatter and transmit intensity

    Transmitted distance .025

    transmitted color .98/.98/.98

    Scattering distance .025

    SSS color .95/.42/.18

    SSS direction .3

     

    also that first image you attaced is the perfect example of the problem thats currently annoying me. The shadow on the neck is spot on but the transition on the cheek is all janky. I'm trying to see if better vellus hair helps, but even if it does thats a pretty resource intensive solution currently edit: its definitely not just vellus, even without any vellus hair there is no sharp transition in blender

     

    grr.jpg
    572 x 769 - 78K
    Post edited by j cade on
  • emoryahlbergemoryahlberg Posts: 133

    isidorekeeghan, I used your excellent settings and I believe my character came out near-photoreal in this HDRI scene. Thank you!

    (larger attached)

     

    home.png
    927 x 1500 - 2M
  • j cade said:
     

     

    I totally agree that normal maps of most of the characters are useless, they dont have small details like pores and stuff, only one that has great details(that i know of) is underbelly8, displacement map is unnecessary if normal map and model itself are detailed enough, i highley recommend using auto face enhancer, its great ,it gives great details to the face and keeps polygon count low but it also doesnt work well on all characters, in the image below i used elithia 8, auto face enhancer and its hd details, multiple hd details from other characters and everything was set to subd level 2 and ofcourse my own sss method also used only sunsky and didnt do any post work, i realy dont like how most characters ears look , they look too basic and unnatural.

    Bluejaunte's characters have those kind of normal maps with pores and wrincles.
    I agree, it doesn't make sense to just grey scale a diffuse map in order to convert it to a bump map, as it makes no sense either to convert it to a normal map.
     

     

    Oh i love Bluejaunte's characters, theire very good and detailed , im sure there are others who make great characters i just cant remember the names , also the whole point of normal map is to replace the need for high subdevision level, if it doesnt have high level details then its useless, also i agree bump maps have the same function as normal maps , they must be made separately not made from diffuse maps
    j cade said:
    Padone said:

    @isidorekeeghan I don't feel your "proper" way is really "proper".

    What is with you people? His settings are gorgeous, they work, and he shared them with us for free. If you have better chromatic settings, I'm sure he'd love to see them. There's a lot of weird "this isn't proper PBR" and "this render engine is better" baloney in this thread. Isn't the point to make the most photoreal images as possible with iray?

     

     j cade that is a beautiful and, I daresay, photoreal render! Iray or Blender?

    Iray!

     

     

    I'm finding it increasibly frustrating that even when you get the perfect skin settings, they seem to only be perfect for ONLY certain lighting set ups. If I change things too drastically, the character ends up waaaay too dark and red.

    This seems to be an issue with chromatic skins.

    There are a couple comments here with a similar thing and generally speaking, If your skin changes relative to other objects depending on lighting then that is a clue something is up with your skin settings.

    Take this skin. Looks pretty good non?

     

    Right up until shes strongly backlit and we realize shes actually a glowing jello monster disguised as a human!

     

    Now this is obviously an extreme example as generally you arent rendering chaacters just backlit and nothing else, but even in less extreme versions, any lighting setup with backlighting the chest neck etc will be unnaturally brighter relative to a setup without backlighting, (not to mention you likely get some weird shadowing around the mouth and eyes)

    Strong side lighting is also likely to create rudolph nose

     

    This is basically my skin testing setup for the record.

    even neutral lighting for over all look and tone

     

    strongly backlit and raise transmitted distance as high as I can get it without the neck and torso starting to glow

     

    rembrandty key light to make sure the nose doesn't glow and also that shadows on the skin soften. (see how the edhes of the shadows are red? also if you so a reder close up there isnt an ege at all but a transition)

     

    this for me is a major key to making skin look soft and imo is subconsiously noticable in most lighting conditions, even though its rarely as distinct as here (its also great for a sort of painterly realism even if you dont hit photo real)

    FWIW This is the area where I'm having the most difficulty matching what I see in reality. While where the nose and head cast relatively crisp shadows is close to matching reality, the crisp shadow terminator edge is very much not. This should be much softer,  though I'm not sure is this is all down to sss or an effect of skin in reality being a much less smooth surface. If anyone has some suggestions for softening this up I am all ears

     

     

    TLDR PSA: when setting up your skin settings don't do it with just even lighting

     

    I do the exact same things, i like to compare my method with yours, can i find yours anywhere? , also are you using different values for the ears? Because it kinda shows, i tried that too but the difference is always visible. Attached comparisons of my sss setting on digital emily and a picture of emily herself, again no post work beside croping

    I'm using the same settings on the ears as the rest of her, I always edit as much as a unit as possible, so when editing sss i always select templates 1-4 (I even avoid having different spec settings on the lip from the rest of the face)

     

    I'm using settings pretty close to yours, the render I'm currently working on has

    spectral rendering on

    Translucency strength 1.00

    base colr effect: scatter and transmit intensity

    Transmitted distance .025

    transmitted color .98/.98/.98

    Scattering distance .025

    SSS color .95/.42/.18

    SSS direction .3

     

    also that first image you attaced is the perfect example of the problem thats currently annoying me. The shadow on the neck is spot on but the transition on the cheek is all janky. I'm trying to see if better vellus hair helps, but even if it does thats a pretty resource intensive solution currently

     

    oh cool there arent that different, glad we are on the same page. That jankiness happens in two situations , one is low poly count(i didnt add any subd to her thats why it looks like that) second is the bump or normal map values are set too high(this is the most common reason for daz characters)

    isidorekeeghan, I used your excellent settings and I believe my character came out near-photoreal in this HDRI scene. Thank you!

    (larger attached)

     

    Thank you, you are too kind , and nice render , although she sticks out from the background a little bit, notice how the wall behind her is brighter than her, maybe increse her brightness in photoshop or something so she blends better.
  • j cadej cade Posts: 2,310
    edited June 2020

     

     

    j cade said:
     

     

    I totally agree that normal maps of most of the characters are useless, they dont have small details like pores and stuff, only one that has great details(that i know of) is underbelly8, displacement map is unnecessary if normal map and model itself are detailed enough, i highley recommend using auto face enhancer, its great ,it gives great details to the face and keeps polygon count low but it also doesnt work well on all characters, in the image below i used elithia 8, auto face enhancer and its hd details, multiple hd details from other characters and everything was set to subd level 2 and ofcourse my own sss method also used only sunsky and didnt do any post work, i realy dont like how most characters ears look , they look too basic and unnatural.

    Bluejaunte's characters have those kind of normal maps with pores and wrincles.
    I agree, it doesn't make sense to just grey scale a diffuse map in order to convert it to a bump map, as it makes no sense either to convert it to a normal map.
     

     

    Oh i love Bluejaunte's characters, theire very good and detailed , im sure there are others who make great characters i just cant remember the names , also the whole point of normal map is to replace the need for high subdevision level, if it doesnt have high level details then its useless, also i agree bump maps have the same function as normal maps , they must be made separately not made from diffuse maps
    j cade said:
    Padone said:

    @isidorekeeghan I don't feel your "proper" way is really "proper".

    What is with you people? His settings are gorgeous, they work, and he shared them with us for free. If you have better chromatic settings, I'm sure he'd love to see them. There's a lot of weird "this isn't proper PBR" and "this render engine is better" baloney in this thread. Isn't the point to make the most photoreal images as possible with iray?

     

     j cade that is a beautiful and, I daresay, photoreal render! Iray or Blender?

    Iray!

     

     

    I'm finding it increasibly frustrating that even when you get the perfect skin settings, they seem to only be perfect for ONLY certain lighting set ups. If I change things too drastically, the character ends up waaaay too dark and red.

    This seems to be an issue with chromatic skins.

    There are a couple comments here with a similar thing and generally speaking, If your skin changes relative to other objects depending on lighting then that is a clue something is up with your skin settings.

    Take this skin. Looks pretty good non?

     

    Right up until shes strongly backlit and we realize shes actually a glowing jello monster disguised as a human!

     

    Now this is obviously an extreme example as generally you arent rendering chaacters just backlit and nothing else, but even in less extreme versions, any lighting setup with backlighting the chest neck etc will be unnaturally brighter relative to a setup without backlighting, (not to mention you likely get some weird shadowing around the mouth and eyes)

    Strong side lighting is also likely to create rudolph nose

     

    This is basically my skin testing setup for the record.

    even neutral lighting for over all look and tone

     

    strongly backlit and raise transmitted distance as high as I can get it without the neck and torso starting to glow

     

    rembrandty key light to make sure the nose doesn't glow and also that shadows on the skin soften. (see how the edhes of the shadows are red? also if you so a reder close up there isnt an ege at all but a transition)

     

    this for me is a major key to making skin look soft and imo is subconsiously noticable in most lighting conditions, even though its rarely as distinct as here (its also great for a sort of painterly realism even if you dont hit photo real)

    FWIW This is the area where I'm having the most difficulty matching what I see in reality. While where the nose and head cast relatively crisp shadows is close to matching reality, the crisp shadow terminator edge is very much not. This should be much softer,  though I'm not sure is this is all down to sss or an effect of skin in reality being a much less smooth surface. If anyone has some suggestions for softening this up I am all ears

     

     

    TLDR PSA: when setting up your skin settings don't do it with just even lighting

     

    I do the exact same things, i like to compare my method with yours, can i find yours anywhere? , also are you using different values for the ears? Because it kinda shows, i tried that too but the difference is always visible. Attached comparisons of my sss setting on digital emily and a picture of emily herself, again no post work beside croping

    I'm using the same settings on the ears as the rest of her, I always edit as much as a unit as possible, so when editing sss i always select templates 1-4 (I even avoid having different spec settings on the lip from the rest of the face)

     

    I'm using settings pretty close to yours, the render I'm currently working on has

    spectral rendering on

    Translucency strength 1.00

    base colr effect: scatter and transmit intensity

    Transmitted distance .025

    transmitted color .98/.98/.98

    Scattering distance .025

    SSS color .95/.42/.18

    SSS direction .3

     

    also that first image you attaced is the perfect example of the problem thats currently annoying me. The shadow on the neck is spot on but the transition on the cheek is all janky. I'm trying to see if better vellus hair helps, but even if it does thats a pretty resource intensive solution currently

     

     

     

    oh cool there arent that different, glad we are on the same page. That jankiness happens in two situations , one is low poly count(i didnt add any subd to her thats why it looks like that) second is the bump or normal map values are set too high(this is the most common reason for daz characters)

    Even with higher resolution you still get a certain amout of crisp termination that doesn't line up with what I've seen of reality

    like on the cheekbones and jaw here theres a pretty sharp edge to the shadow which doesn't feel real to me (I probably should have turned down the light but it is a quick test render)

    In this image I found on the interwebs the shadow cast by the nose and on the neck is fairly crisp so its clearly not a much bigger light source, but the falloff on the cheek and chin is distinctly softer

     

    I definitely think some of it is vellus hair so I did a super quick test in blender because it renders faster

    vellus hair affecting shadow termination

    no vellus hair

    I still think blender handles this a bit better than DS does in general

    janky vellus that i didnt spend all tht much effort actually making look good

    muuuch softer transition on the cheek and temple at this scale you cant even really see the individual strands (whichh is good as you don't in real life) just the softened transition

    bad news the vellus hair here is made up of 300,000 strands. That is going to be signifigantly memory intensive in Iray

    shadow term.jpg
    700 x 1050 - 650K
    vellus ex.jpg
    666 x 865 - 83K
    vellus ex2.jpg
    666 x 865 - 90K
    Post edited by j cade on
  • Theme3DTheme3D Posts: 27

    Yash, thanks for the tips.

    Here another anoying thing, the hair should be blown out, but it's not.

    You have to adjust every single asset to make it consitent.

     

    TestRender.png
    712 x 634 - 626K
  • j cadej cade Posts: 2,310
    edited June 2020

    And a more draamatic example. Very testy render. this is not "good" lghting and I put minimal effort into actually making the vellus hair realistic

    but to my mind one of these looks infinitely closer to reality even with its flaws

    No vellus

    vellus

    vellus ex3.jpg
    666 x 865 - 95K
    vellus ex4.jpg
    666 x 865 - 103K
    Post edited by j cade on
  • NylonGirlNylonGirl Posts: 1,501

    isidorekeeghan, I used your excellent settings and I believe my character came out near-photoreal in this HDRI scene. Thank you!

    (larger attached)

     

    It's nice to see somebody finally post a picture again.

  • j cadej cade Posts: 2,310

    isidorekeeghan, I used your excellent settings and I believe my character came out near-photoreal in this HDRI scene. Thank you!

    (larger attached)

     

    I agree with what isidorekeegan said she doesnt really seem to be fitting into the hdri for several reasons.

    she seems overall a bit dar for the scene as isadorekeegan mentioned

    the shadow she casts doesn't interact with the wall correctly (this could be fixed by using a matte object there are tutorials floating about)

    perspective of the background is wonky relative to figure. the theoretical plane the figure is standing on seems to be a very different angle from that of the scenery, we're looking at the figure very straight on, but the table appears angled towards us the wall behind her even has a bit of a curve, which gives me flashbacks to the fist pair of glasses I ever owned which were slightly misaligned and turned the whole world expressionistic

    The lighting doesnt seem to fully match up either, strangely enough for hdri, In particular if you compare her to the table she has very strong backlighting that is mostly absent from the table and the shadow she casts at her feet is signifigantly stronger than the shadow we can see the table leg casting. The table is also casting a shadow on the wall behind her but not on her legs

     

    Most of this is honestly not stuff that people will actively notice unless they're looking for it but even without actively noticine something will still feel off

     

    I honestly find HDRI where the background is close to the figure as it is here incredibly difficult to get looking good. HDRI are great for adding sky and distant trees, but really difficult to use as rooms to stick your characters in, at least in my experience

  • j cadej cade Posts: 2,310
    edited June 2020

    and i definitely need to fine tune this, because she looks a bit like a nice young menonnite lad right now, but now the actual Iray comparison

     

    no vellus

    vellus

    Definitely starting to get that softer transition on the cheek , much more in line with how I look when I shine my phones flashlight on my face (well other than the fact that I definitely don't have cheekbones like that)

     

    And while this is definitely just a test scene rather than a more real world lighting setup those harsh transitions are definitely something I've noticed, sunny hdris for instance which are pretty common, can easily create harsh shadow transitions like this

    also irrelevant side note but blender is so much more efficient it hurts. Those blender renders finished before the iray ones even loaded fully and started.. but in ds I can reuse settings, strand hair, materials, etc so much more easily. So I'm not ready to abandon it quite yet

    shadow term iray.jpg
    700 x 1050 - 592K
    Post edited by j cade on
  • emoryahlbergemoryahlberg Posts: 133
    edited June 2020

    Thank you j cade and isidorekeeghan for your comments! I'm continually trying to improve, like all of us, so your feedback is invaluable. smiley

    Those blender renders finished before the iray ones even loaded fully and started.. but in ds I can reuse settings, strand hair, materials, etc so much more easily. So I'm not ready to abandon it quite yet

    Coincidentally, I'm going through tutorial courses now to learn Blender. Is it really very difficult to reuse settings and materials? You'd think this was something Blender would have in the bag! 

    My goal is to transition to Blender completely. I use DAZ to create comic images, so I hope that isn't a pipe dream....

    Post edited by emoryahlberg on
  • j cade said:

    and i definitely need to fine tune this, because she looks a bit like a nice young menonnite lad right now, but now the actual Iray comparison

     

    no vellus

    vellus

    Definitely starting to get that softer transition on the cheek , much more in line with how I look when I shine my phones flashlight on my face (well other than the fact that I definitely don't have cheekbones like that)

     

    And while this is definitely just a test scene rather than a more real world lighting setup those harsh transitions are definitely something I've noticed, sunny hdris for instance which are pretty common, can easily create harsh shadow transitions like this

    also irrelevant side note but blender is so much more efficient it hurts. Those blender renders finished before the iray ones even loaded fully and started.. but in ds I can reuse settings, strand hair, materials, etc so much more easily. So I'm not ready to abandon it quite yet

    Does it still look harsh if you remove bump and normal maps? And also in real life there is always something in the space surrounding us, furniture, walls etc, maybe add some simple plane with some texture on them and see if that helps
  • PadonePadone Posts: 3,481
    edited June 2020
    j cade said:
    If anyone has some suggestions for softening this up I am all ears

    As for light tests the method by @isidorekeeghan doesn't work with a simple three sphere lights setup. You may like to test your method too to see if it fits studio lights. Test scene included.

    It is my advice to avoid natural and use faithful to get a correct light falloff.

    https://www.daz3d.com/forums/discussion/comment/5743051/#Comment_5743051

    Post edited by Padone on
Sign In or Register to comment.