Skin different in Blender

Issue #555 resolved
Former user created an issue

Hello, I’ve exported my model directly from Daz to Blender without changing much but for some reason, the shader looks off in Blender, way off… It happens with most skins by Bluejaunte and I can’t get it to look right again

Comments (31)

  1. Alessandro Padovani

    Please be sure to use the bsdf option for materials and cycles for rendering, otherwise the skin will be approximated and it is expected for the user to fix it if needed.

    Then if the issue stills there I need a reference to look at, if you can point to a figure having this issue I’ll look at it as I get some time. Unfortunately it may not be so soon since lately I’m quite busy.

  2. Thomas Larsson repo owner

    Personally I prefer Eevee and the principled method, since I get nice renders quickly, even if the materials are not exactly the same as in DS (the BSDF method usually doesn’t work with Eevee because translucency blurs all features). Recently I have also encountered problems with white skin, which happened because the translucency channel suddenly did not have a texture slot. There is a gamma correction node which expects that texture as an input, and if there is none the subsurface color is white.

    Nice skin, with translucency texture:

    White skin, no translucency texture possible:

    In the last commit the diffuse texture is used as input to the gamma node if there is no translucency texture. Hm. The same trick should probably be used in the BSDF case as well.

  3. Thomas Larsson repo owner

    Now the same trick has been implemented in the BSDF method as well. Here are renders of the same character, imported before and after the fix:

  4. Alessandro Padovani

    daz studio 4.15, blender 2.93, commit fafb981

    Thomas, you may want to use the test scene in #179 when dealing with skin materials, so to get a clear view and comparison of what you’re doing. Also the volume settings are important, not only the diffuse settings, so you may want to consider them too in your tests.

    Specifically, on a quick test, when translucency is white as in your example, the skin color seems to come from the transmitted color in the volume settings. So it gets nothing to do with the diffuse texture.

    edit. important. On a quick test this seems to also work for the volumetric skin. Again better extensive tests would be required.

    # NOTE this is for principled and eevee bsdf (sss skin), not for cycles bsdf (volumetric skin)
    if translucency color == white
        translucency color = volume transmitted color
    

    Below an example and test scene that you can use skin-test.duf. First iray, then commit fafb981, then the fix where the volume transmitted color is used. Please note that this is a quick fix I didn't have time to test it extensively.

    As a side note, on a first look it doesn't seem this is the issue pointed out by Willfre. But I may be wrong we need a reference figure to understand what we need to test.

  5. Thomas Larsson repo owner

    The weird thing is that in my second example, not only is the translucency texture gone, but also the possibility to add one. There is no arrow to the left of the translucency color. Probably something in my database is corrupt. When I open the file, DS wants to reinstall the starter essentials.

  6. Alessandro Padovani

    I always use manual install and work offline, dim and connect are a mess. Yes you probably may need to reinstall the essentials, including the default resources.

  7. Former user Account Deleted reporter

    I have the same issue with the arrow on the left. I noticed that it doesn’t happen every time but for some reason out of nowhere I noticed that it disappeared when I switched skins while doing tests. Looks like something goes wrong in Daz.
    And when I imported the character where the arrow doesn’t show up, this is what happened:

    Also, just in case, I’ve used the same skin before and it used to work perfectly, it’s the skin from Miriam HD by Bluejaunte.

  8. Former user Account Deleted reporter

    With the new commit and a skin that works (with an arrow next to the translucency color), it looks way better already! The only thing that I can’t get right is the diffuse texture, it’s not exactly the same as in Daz, details of the face and just the textures in general looks faded

  9. Former user Account Deleted reporter

    So… I’ve been able to get the arrow back. I copied all of the surfaces, applied another shader (PBRSkin here) and reapplied the Iray Uber shader, then I pasted the surfaces back. The arrow reappeared and I added back the skin. It seems to work better now and it’s definitely something to do with Daz.

    Here is the character after this solution, no longer missing textures:

    It’s a whole other issue than the one I’ve reported in this topic but it seems to be fixed, or at least has a workaround.

  10. Alessandro Padovani

    The fix by Thomas, apart being incorrect, does not apply to Miriam, because in her case there’s a translucency texture and color. So if you see any difference with the last commit it’s just because you solved the issue you were having with daz studio.

    Then there’s indeed an issue with the Miriam skin but it’s because of the dual lobe effect that’s too extreme. Specifically the fresnel effect is not right. I’ll see if I can find a solution to this.

  11. Former user Account Deleted reporter

    I’ve also tried to give it a go, it’s getting closer to what I had in Daz, it can’t exactly be the same obviously but it’s better already. I feel like changing the weight is not the solution but it’s better, the skin details are more visible.

  12. Thomas Larsson repo owner

    Using the diffuse texture instead of a missing translucency texture is now optional, called Fake Translation Textures. It may be incorrect (in my case the skin is white in DS too), but it is useful to me when the texture has gone missing. Note that this problem cannot be fixed with the material editor, which never changes the topology of the node tree.

  13. Alessandro Padovani

    @Willfre judging from the eye reflections in your images you’re not using the same lights in blender and daz, this together with tone mapping does affect the skin tone.

    @Thomas I’ll do some more test to include both the volume trasmitted color and a better fresnel effect, this may take a few days since I get a limited free time.

  14. Former user Account Deleted reporter

    @Alessandro Padovani I didn’t get much difference with the same lighting as in Daz, the results were mostly darker and I can’t get the tone mapping right. Even with better lighting, the skin seems more grayish while it is closer to orange/red with Iray

  15. Former user Account Deleted reporter

    It will need some work on the dual lobe, the fresnel and the volume but with better tone mapping and the same lighting as in Daz, this is how it looks like right now.

  16. Alessandro Padovani

    For tone mapping, if you want to match the colors, I’d suggest standard with medium or high contrast. While filmic is “better” to avoid overexposure but may “flat” colors as well, especially on skins.

  17. Former user Account Deleted reporter

    I did tweak these settings to get closer to the original but I definitely need to fix the skin materials as well, it seems to have an impact on the skin. Changing the tone mapping didn’t really help with the skin details, the texture still seems to be grayish and the freckles are not as visible.

  18. Former user Account Deleted reporter

    So… I’ve noticed that some textures were not imported, especially temp maps (layered images) created in Daz, for this skin at least. A translucency map was created but not imported in the process. Also, the SSS color was not the same as in Daz but I’m not sure if that’s important in this case with the maps. The bump map disappeared as well since the last commit, it didn’t get imported properly. It’s also a temp map so I guess it has something to do with it.

    From what I’ve noticed, temp maps are actually replaced by node groups, which doesn’t seem to take the final image from Daz but does the layering in Blender directly, it forgot the bump and translucency maps.

    As I said, it still needs a few tweaks because the skin is not as glossy as in Daz, the dual lobe is way off and I’m still trying to tweak it the right way (I’ve managed to make the freckles more visible this way, I’ll see what happens once @Alessandro Padovani manages to get somewhere with his tests, tell me if you need something).

    The tone mapping was also tweaked, it’s not perfect but I don’t want it to be exact since it’s way too different in Blender. I’ll get it as close as I can. I guess this is why it’s so yellowish in Daz.

    Here is what I managed to get with all of this (I only changed the face/lips/ears materials, notice the difference with the torso material and the back of the head with the original dual lobe, the missing maps, etc):

  19. Alessandro Padovani

    Yes please post your material settings it will be interesting to see what you tweaked, supposing we use Miriam as reference. As for temp maps we don’t use them, also it wouldn’t make much sense in my opinion. Actually we rebuild the materials with nodes as you noticed, using the original textures.

    As for the bump and translucency maps they seem correctly imported. I don’t get what you mean.

  20. Former user Account Deleted reporter

    As indicated in these screenshots (it’s only the face material, the others use the basic textures and not the layered/generated textures), the bump was not imported for the face Material. The translucency map was actually imported. These changes definitely does not seem to be the right way to go, it’s just the tests I did to get as close as possible to the Daz image.

    Also, I don’t see where the SSS color is used when Imported, does it have a use anywhere in Blender? Did I get something wrong?

  21. engetudouiti

    Though I usually use custom shader for character skin, but I think current add on bsdf option not use blender offer SSS shader node for bsdf skin. but only use translucency then try to mimic iray absorb and scatter by volume scatter and volume absorption.

    I do not hope to use any volume shader for blender skin, (it is my view) so I usually not import volume shader. (it may work well clear refractive material, but do not expect diffusive absorption and SSS like human skin etc, I do not think blender cycles volume shader is made to work for the purpose well)

    If you check the add on generate Daz translucency you may find , it include bsdf translucency node and SSS shader node. then usually Cycles only use translucency shader. and Eevee use SSS shader.

    then if you do not use volume shader option (global setting), you may need to modify it.

    about blender volume shader (absorption and scatter) , I do not know how add on convert it. but if you choose the option, it may count the SSS color I suppose. (though I do not think it work well for skin as I already said)

  22. engetudouiti

    No I miss informed. current add on not offer option, which choose volume shader or not (it seems auto generate with bsdf option. I might hope if Thomas offer the option as same as displacement. ( try to use SSS shader only to mimic daz volume SSS and absorption for skin) .

    anyway after import character, then if you hope to customize, my recommend way is, once remove (un-socket) current volume shader node temporally. then change setting SSS node which was included in translucency group nodes for cycles. (you may need to plug the SSS node to cycles out put)

    About the case , it not offer volume absorption for skin.

    but SSS shader do, diffusive translucency (which filter color absorbed as second layered skin) +

    Subsurface scatter. you can adjust which color will scatter with tweak SSS RGB color (the unit is meter, so care to change RGB subsurface values)

    if you really do not get reasonable skin, you may try again volume shader setting later . (but I do not think it actually need, without you make ghost or volume glass etc)

  23. Former user Account Deleted reporter

    @engetudouiti Well my thoughts on the SSS color came from the fact that the face looked “redish/orange” in Daz and that the SSS Color was not used. So I wondered if that was the culprit, but after testing other solutions, I didn’t need it. I was just curious as what it was used for. Playing with the tone mapping and using the diffuse as translucency seems to get back to the color I need, I just have to fix the Dual Lobe and it should be good. Thank you very much for your answer though, it might be very useful for a few things I have in mind! For now I think I’ll stick to the volume shader, it seems to get closer to what I need.

  24. Former user Account Deleted reporter

    It’s all good, it’s already working as expected, thank you very much!

    My focus is more on the dual lobe and the glossiness of the face, the fresnel. The translucency has been greatly improved since I used the solution Thomas provided in the beginning, but I’m still stuck on the dual lobe. For now, the glossiness seems wrong, especially when you look at the tip of the nose and the lips. It’s also the reason why the skin is whiter with less details, and I’m sure that it will fix everything once I manage to get it right.

    I’ve reduced the weight to get back the skin texture and reduced the Fac to get more glossiness but it’s too blurry, not sharp enough like in the Daz render (might or might not be possible to get right though). It feels like the fresnel is definitely blending too much over the center of the face. Again, I might be completely wrong and not going in the right direction there, but my goal is to get as close as possible, not to get something perfect, and I’d like to do it the right way if possible.

  25. Former user Account Deleted reporter

    I could send you the file so you can have the model to work with for your tests, if you think it can help

  26. Alessandro Padovani

    No you can’t. The daz eula doesn’t allow to share daz or blender models. I have a bunch of them for testing including Miriam though just want to sort things out for Thomas for the implemetation and to be sure we get a general solution that works fine enough with most figures.

    Thanks anyway.

  27. Former user Account Deleted reporter

    Okay then, great! I’ll keep working on it as well until then, trying to figure out a few things on my side.

  28. Log in to comment