Adding to Cart…
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.You currently have no notifications.
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
@RayDAnt Do you know the exact settings in Iray to copy Random Walk or Burley Christensen SSS? I tried for a long time to replicate it but couldn't.
Sorry to say I don't. One of the downsides of being seriously into the theory behind things is that you tend not to have the time to fully explore the practice of it.
Thank you for this info, which is instructive and interesting, always prefer to have a primary source for something like this. However, can't help feeling a little misled by the marketing hype into thinking Iray was giving us 'proper' SSS rather than an approximation, notwithstanding it's a good approximation and that fact was documented (in highly technical docs which hardly any casual user will ever find, let alone read). Nvidia could be, and could have been, more up front in saying something like 'This is not 'real' SSS, but it's so close you'll hardly ever notice the difference (and, whisper it, the coding is easier us too - so, bonus!)'.
It's the opposite of what you're thinking. Iray has true SSS. But approximation SSS methods look better than Iray SSS in my opinion. That's because they are trying to simulate specific materials like flesh, and Iray's brute force method doesn't look as good out of the box (but probably can, just need a lot of adjustments)
There is no such thing as 'proper' SSS. Subsurface scattering is an abstract real-world lighting effect whose emulation is key to getting certain things like human skin to look real in computer-based simulations. It isn't a specific method of emulation. The way in which Iray differs from most other PBR renderers regarding SSS is in the method of emulation used (volumetrics rather than depth mapping or texture space diffusion.) Not whether it has 'real' SSS (there is no such thing as real SSS in computer simulations. Just subjectively more or less convincing methods of emulating it.)
Wow, reading this thread was pretty interesting although I have not understood big chunks of it as I am new to the rendering world.
I am a user of both Blender and Daz.
Until now I have been using Blender just for modeling. I love how Blender renders look but since a lot of my models come from Daz as I don't have the time to do it all myself and since just exporting from Daz and importing obj into Blender doesn't look that good I've kept away from rendering anything serious in Blender. I do not believe there is a real need for Iray in Blender. I mean it would be convenient for the people who use Daz as well but I don't think there are that many. Blender is already taking advantage of the Cuda cores for faster rendering so I think people are already tempted to buy Nvidia for faster renders. They don't need the Iray incentive as well.
I know this is kind of only half on topic but I would like to ask it here since it appears we have enough experienced people. Are there any other renders engines that could render out Daz content without spending time to individually convert all materials?
I know Octane render has a free plugin for Daz and I've seen some amazing results but I think you still have to convert materials, right?
Second question would be: which render endinges do you guys think are capable of rendering the most realistic skin (and hair)?
Thanks
Some engines might be easier to use than others ... and for some you have better documentation than others. For instance there seems to be more tutorials available on skin settings in Arnold and Cycles than there is for Vray but that might just be due to the fact Arnold is a Autodesk product and ships with their apps and Blender is free while you have to pay extra for Vray so Cycles and Arnold are more accessible and more People ist them. The more people use them the more tutorials and discussions will be available. Not that you can't do amazing skin and hair with Vray.
I am not so sure about that.
I have seen some amazing renders (I don't know in what render engine they were made) that blow my mind. I have not seen anything close to that from Iray.
Here are some examples:
https://www.cgtrader.com/blog/portraits-of-the-21st-century-the-most-photorealistic-3d-renderings-of-human-beings
https://cgelves.com/10-most-realistic-human-3d-models-that-will-wow-you/?v=f5b15f58caba
https://area.autodesk.com/life-in-3d/photo-real-portraits/
I wouldn't say the Iray Emily looks too far from the Vray version, or the Arnold replications. Now I won't deny if you claim that Arnold and Vray have generally better materials to reproduce skin. Many users have argued that the Iray Uber base isn't ideal for reproducing human skin results.
But anyway, don't look at the final renderer only if you want to understand how it is done. First of all, the source stuff of professional photorealistic renders cannot be compared to source stuff of DAZ renders. If you don't believe me, go up and look at Emily's textures alone, or go to textures.xyz or documentation of Maya, Houdini or 3ds artists how they create their models. Look how many details these models have and compare them to Victoria8 even with HD morph turned on at subdiv4. If you import these to DAZ and render, will they look the same like the Arnold equivalent? Probably not. But way better and more realistic than your average Victoria8. Now imagine if DAZ Iray had a better skin material than the DAZ uber base.
From my own experiments and what I see of other artists, if you want Genesis8 to look better with another renderer you have to do a lot of additional work yourself. If you only plug in G8 materials into Arnold they won't look more realistic than in DAZ Iray, mostly probably worse.
I relink all my materials after importing, but it doesnt take long for me at least. There are no render engines i know that takes Daz material to be converted correctly. Since Daz material is unique to its own engine. Any render is capable of rendering skin or hair if it has the functions for it and how much effort you put in learning the application and how you use it. Heres some examples of my "realistic" attempts in Blender.
FYI at least ten (including the very first one in the first link) of the portraits in the gallery links you posted were created using Iray (Mental Ray is what Iray was called prior to Nvidia buying it in the mid-2000s and adapting it for dual CPU/Cuda use.)
That looks amazing. thanks dude
It's done. In the volumetric options you can choose volumetric, translucency only, or sss.
1) Volumetric will only use volumetric properties the same as iray does, so it will match fine. This is also slower though so it doesn't fit well to animation. That's why the translucency only and sss options are provided.
2) The translucency only option will ignore the volume properties and convert only the iray translucency. This is useful since with many characters there's not much difference with thin walled on.
3) Finally the sss option will try to translate the main volumetric properties to sss, but of course will not match with iray since sss can't do what volumes do and vice-versa.
Also dual lobe specularity is improved and now it works much better. So dig in and have fun. Just get the development version.
http://diffeomorphic.blogspot.com/p/daz-importer-version-14.html
As a side note the cycles dual lobe seems to work even better than the iray one. That is, the skin reacts better to lights with white reflections as a real skin does. Below an example with V8, first iray then cycles with the default hdri.
Please also note the ear translucency that seems to work better in cycles as well. That is, iray tends to scatter the translucency even outside of thin zones while cycles doesn’t.
@RayDAnt I'm not sure I follow you entirely there. As for volumes being better than a dedicated sss solution of course I can't agree. Simulating sss with volumes is purely academic and doesn't make sense in production. Anyway thank you for your comments I do always find them interesting.
@davidtriune Again, surfaces and volumes are two different things, I'm afraid you make some confusion. The daz characters use volumes, not sss.
Thanks Padone for your work on this. It's awesome. However do you know if it can import models with Daz's OpenSubD already applied, instead of importing at base resolution and applying Blender's Catmull-Clark? The result is pretty different.
Actually the plugin tries to import the daz subd level as well. But there's a bug I'm investigating that tends to screw up close geometries such as eyes. I'm going to report it properly to Thomas as soon as I have some time to better define it.
In the current version the best option is to export the model at base level. Then apply the subd in blender. This avoids artifacts.
These look great! Very realistic. What do you mean exactly when you say you relink your materials? You don't configure them with nodes as Blender materials are structured?
Wow, I didn't know that. I saw Mental Ray mentioned on some of them but didn't know it was a precursor of Iray. I now think that Iray is not the problem in and of itself but mostly the uber shader and the textures themselves.
@Padone That looks really nice, I will give it a try and post some results here as soon as I finish my current render.
Subsurface Scattering (SSS) is the name of the actual visual effect you see in real life when light hits skin. It isn't a method of processing computer graphics. In order to achive SSS-like effects in a render you need to choose a particular graphics processing method to do that. Currently the two most popular methods for simulating SSS effects (because they are fast and very tweakable, although significantly less physically accurate) are depth mapping and texture space diffusion. A third less common method of simulating SSS (because it is slower and less tweakable, although technically much more physically accurate) is volumetric (click at your own peril... it was co-authored by Jensen himself.)
Subsurface scattering is the lighting effect. In this context, depth mapping, texture space diffusion, and volumetric are just different methods of achieving it.
Here is the list of currently supported methods for computing SSS in Blender according to its shader model documentation:
Properties
As you can see from above, not only does Blender's Cycles engine have two built-in volumetrics based methods of computing SSS (one a true implementation, the other an approximation), it's own documentation highlights the fact that volumetric SSS renders more accurate visual results.
I make no claims to being a Blender expert - much less being knowledgeable about the specifics of Daz Studio > Blender importing. However, it would seem to me that that best way to go about getting Iray SSS to import into Blender for rendering would be to first focus on an Iray volumetric > Random Walk volumetric path (since those are evidently the closest equivalents) and then tackling the trickier task of an Iray volumetric > Christensen-Burley volumetric approximation path (I am presuming from what I've gleaned upthread that Christensen-Burley is the prefered SSS method among Blender users looking to temper rendering quality with speed.)
ETA: Assuming all of the above is accurate, this would lead to that documentation point #3 ideally reading something more like this:
And the importer plugin itself having the featureset to match.
These are good looking! Especially the second one!
@RayDAnt
You are mixing different things. Of course sss uses volumetric methods, otherwise it wouldn't be pbr. But it's for surfaces, not for volumes. That is, in a surface the light scatters along the sub-surface, thus producing sss effects, for example skin rims and translucency desaturation with random walk. In a volume the light scatters through the volume, not along the surface, thus producing different effects. For example random walk doesn't make sense for volumes because it's a volumetric method to solve subsurface simulations. That is, volumes don't need random walk because they don't have to solve the problem to scatter light along the surface.
Now iray doesn't implement a "volumetric sss" in any way, it only implements volumes. That has nothing to do with volumetric methods used for sss. If you look at the cycles implementation it's absolutely clear. The uber volume only uses the volume absorption and volume scatter nodes. These are for volume simulation, not for subsurface simulation. To get subsurface effects with volumes you have to model the three layer skins with geometry. This way the light is forced to scatter along the surface instead of scattering through the volume. But it doesn't make sense to implement sss this way.
P.S. I'm not explaining this just for chatting. My goal is, once it's clear that the uber shader only uses volumes, and what's the difference with sss, that someone could help to better convert volumes (iray) to sss (cycles sss option), since at this time the sss option is quite weak and I feel it can be improved.
I tried importing with your settings and the model is still missing details. I am using Auto Face Enhancer geoshell, I'm not sure if that changes anything. I attached picture of imported with obj vs the plugin and you can see it's missing face bumps.
@Padone, nice work, I can't wait to get home and try this.
But can you describe how to get the plugin to import the daz subd level? I have only seen the base mesh imported with subd applied in Blender, which, as @davidtriune has pointed out, can look quite different. That's the main thing keeping me from using the plugin for everything; Blender subd COMPLETELY changes the look of my protagonist and I would love to get a more vertex-perfect result closer to what, say, Alembic gives.
@Padone, when you talk to the creator, any chance you can ask him to allow like 20 runtimes to be mapped lol
ATTN: @Padone et al,
I just realized I committed a HUGE SNAFU back near the beginning of this thread. You know how I posted this excerpt from the Iray whitepaper seemingly confirmnig the assertion (first mentioned in this post) that Iray "does not feature dedicated approximations for simulating sub-surface scattering effects"? This is not true. Upon rereading the portion of the whitepaper (pages 13-15) dealing with materials evaluation (which is to say, the part dealing specifically with how surface materials like human skin are processed) it has come to my attention that this statement is only valid in reference to how Iray handles volumes - not surface materials. Material evaluation in Iray is handled by an extremely flexible and highly programmable (via MDL) layered BSDF function which includes all of the standard light interaction methods commonly used to achieve SSS effects. Meaning that the only limits on how SSS skin is achieved in Iray are those established by the particular Daz Studio SSS skin shader you are using.
ETA:
As luck would have it, it appears that the MDL source code behind the default Daz Studio shader responsible for SSS skin effects in Iray (the aptly named "Iray Uber Shader") is readily available. Just go to:
"C:\Program Files\DAZ 3D\DAZStudio4\shaders\iray\daz_3d/irayubermaterial.mdl" or
"C:\Program Files\DAZ 3D\DAZStudio4 Public Build\shaders\iray\daz_3d/irayubermaterial.mdl"
Blender users looking to get SSS skins to import properly - I'm pretty sure the secret lies in reverse engineering this MDL code (which, by the way, was written by Daz's developers - not Iray's.)
Sorry for injecting extra confusion into all of this. Hopefully some of this info helps in your quest.
Tried my luck with the development version, skin and eyes rendered pure white in both cycles and evee :(
Forgot to save the renders before I closed, oh well lol.
I thought just to avoid possible confusion since it wasn't clear in the post, it should be noted that AOA's Subsurface Shader Base is for 3DL, not Iray.
Yeah, corrected it above you (and possibly made a very useful discovery in the process.)