Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.Daz 3D is part of
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
Alternatively, you can use the tone mapping controls for lower luminance and color burn if you don't want to mess around with intensity scale or Exposure value in your lights.
The blur settings might be an oversight.
I've changed the exposure settings for the environment sphere shader. No more diffuse + specular influence, it's either diffuse or specular on the AWE Environment light. If you're using the controls in the shader itself its for camera+specular+diffuse, specular (for the ones labeled with offset) and then diffuse (with the diffuse labels).
Tks wowie!
Here's the result of me starting a new scene/character with two of your default light planes with 7EV light intensity, the one in front with poly blur enabled, and the env. sphere (almost black), and some bounce planes off camera. I loaded a skin preset and started tweaking;)) AWE Hair on the hair (transmapped), fibermesh brows and beard (no maps). Sideline: Love these skin textures...if only I had the energy to remove part of those painted on wrinkles
..
Question, if you see this, wowie? When adjusting SS strength, is it better to leave it at 100% and reduce phase, or should I leave phase at (normally for me ) 0,75 or so, and then reduce SS strength? I'm trying both methods on a female skin now..
Didn't spend much time on the sim, sorry bout that;)
Edit:
Oh forgot to mention, while playing with the above dude, I noticed I got some black spots on the skin when having "multiply spec. with roughness on 100% and reducing "use specular maps" to around 0 - 50%. IIRC at 0% the head turned almost completely black.
And, regarding contact shadows, like the shadows of his shirt on the chest, I get a feeling they were better in the previous version. Could be tighter in my honest opinion;)
I replaced the diffuse maps and played a bit more with SS
And an attempt at female skin
Yah, still a bit lost but having fun
Still not sure how to set up SS. Here I tried to play with diffuse gamma (0.85) and saturation (1.1) to see what happens to SS, as I feel it is a tad too washed out for my taste. I guess it's about finding the right balance, but I wish there were a SS saturation control of sorts..
Hm, I reset diffuse saturation to default and raised SS phase from 0.74 to 0.75. Looks like he got a firefly on his left hand. SS-related?
Literature suggest 0.8 for skin. It's probably best to dial down subsurface strength instead of fiddling with subsurface phase. As for saturation, you can play with diffuse saturation a bit, but there's a limit to how much you can push saturation if the texture is already quite saturated. Or do it like photographers do by adjusting color temperature of your lights and the image (via the Environment light) and / or saturation.
The goal for the new build is to have the skin look as close as possible to what you see in the GL viewport when everything is at default.
The older builds did use 5600K for default, but I changed it to 6500K for a more neutral render. Saturation is already set at 0.5 though.
The specular stuff is unfortunate. You can play around with specular map strength to see if that helps. The firefly is likely because of specular not SSS.
Ok got it! Yes I always thought 6500K was the appropriate default value and always start with everything zeroed out (doge burn and saturation). I find it easier to adjust those things in the end;)
Wowie did you actually solve the HDRI diffuse sampling thingy? I just tried turning off everything but the environment, zeroed everything and loaded an HDRI. Looks clean enough to me
. Even Kettu's shadowcatcher seems to work much better? Only one HDRI tested though...
...so pure HDRI light, same skin...
Not sure I follow. Could you open it up a tiny bit?
I checked the specular maps I used here (randomly picked M4) and they were extremely dark. So my guess is that your normalize function did its best to make sense of it, which also could explain the firefly.
.
Not, really. The cleaner HDRI render is probably due to better brute force sampling. I do have an idea about doing it the old fashioned way, which is an environment map look up, for that kind of specific scenes.
As noted, AWE Environment sphere gain/gamma/exposure and saturation settings have been changed so the main control affect camera, specular and diffuse rays. The same settings with the added label 'Offset' will only affect specular rays, not camera and diffuse. In previous builds the offset settings affect both specular and diffuse rays. In hinsight, it would've been better to use this scheme in the start.
Yes. Very dim maps is still problematic. You can try using both 'Normalize Specular Map' and Specular Map strength to boost the levels.
Ok but it sure looks better to me;) Yes I checked and replaced the bad controlmaps and while browsing I found these cool vein maps (color and masks) so used them in the coat layer(transmission and thickness). Here's a testrender with neutral awe light and default samples. I reworked the whole skin the way I used to set them up with SS phase 0.8, and with the proper maps in place I think things are starting to make sense:)
Tks, got it!
20x20 ps, neutral default awe light (256 hair samples), slightly tonemapped in post
I'm guessing the car is a model? Looks quite good from what is visible.
It's the Spy Car V12, good model with reasonable surface zones and good topology;) Deriving from a Jag, maybe? I used a dark diffuse with a low-ish IoR and medium specular strength. Dark blue coat layer with IoR 1.55 and thickness 0.08 + some coat reflection. (from memory)
So I spent yesterday not touching the dude's skin but adjusting the env sphere and awe light settings. Decided to bring general specular level down by 2 EV and adjust various offsets to try to minimize specular noise and so on. Did nothing else to the car but add 0.5 EV specular exposure to compensate. I think the result is quite descent for pure HDRI light, this is after all with 1024 Irradiance- and 256 Hair samples, diffuse depth 3 and specular depth 2, 20x20 pixel samples. Oh and I used 6500K, and your default sat, dodge and burn settings with Aces tonemapping. Rendertime with these settings just over an hour..
I'll put HDRI aside for now, good to know it's a viable option but I prefer working with area lights:) But yah I love brute force, always did
So here there's no environmental light, just 20 area light planes, positioned under those light armatures. Set them to double sided with front intensity scale 8.7, back intensity 3. Specular contribution 0.5 and temp 8400K. The fluorescents themselves use the environment spere shader with camera only visibility.
I loaded 6 different skins and tried to set them up in a similar fashion, basically doing gamma diffuse adjustments. And of course SS.
Awe light is using unreal tonemapping, otherwise zeroed, no diffuse/spec overrides, 256 Hair samples, 1024 Irradiance samples, non progressive 24x24 pixelsamples;) 2400pixelwidth, Total Rendering Time: 2 hours 11 minutes 56.3 seconds
The raw render and one with a bit of bloom:
The car is based of the Aston Martin DB5 Vanquish I think. The one used in the last Brosnan's Bond flick.
You probably want to apply selective bloom there to limit bloom only to visible light and not reflected ones.
Any problem with fireflies or noise so far (other than on the previous test shot)?
Ahh thanks, I had a brain fart;)
Yeah a bit sloppy work there..
No nothing I haven't been able to solve so far. Overall look is again much cleaner than previous build, especially the hair shader feels robust now. I used to use min 1024 samples, now 256 seems enough for most scenarios. Adaptive sampling and/or blackbody is improved? 1024 Irradiance samples also seem to cut it most of the times. I think there is still noise issues with GGX if using high roughness, so I tend to avoid that.
And, as long as I limit the arealight shader specular contribution to 0.4 to 0.6, depending a bit on light intensities, I have no problem with specular noise on metals. With default settings I run into all kind of problems (this has been the case probably always, and I started setting up all materials with spec contribution 0.5 maybe a year or two ago). What happens,( if you allow me to speculate), is the metal highlights hit some sort of limit, after which the pixelfilter glitches and you get jaggies that no pixelfilter setting or DoF can deal with. The only thing that helps seems to be reducing metal specular exposure or specular strength or brightness or specular contribution. I like the last option because it gives me headroom;)
I made another outdoor test using two arealight emitters. An HDRI for camera visibility and env. reflections. Same HDRI converted to 8bit .png plugged into sky emitter color with reduced saturation and increased gamma. Specular contribution for sky emitter 0.05, so basically diffuse. Sun emitter specular contribution 0.5. No changes to the character skin from previous render. I experimented a bit with the car glass, combining opacity with base thin film and coat layer with some transmission. AWE light is again neutral (unreal tonemapping) with default samples and diffuse/spec override 2/1. Noise levels look just fine to me.
Raw render:
Edit: I'm taking a closer look at the offending highlight on the left headllight, will post examples...that's from the sun emitter, btw.
Re-render with a tad more sun and sky intensity at 24x24 ps, no overrides,default samples
Untouched render:
So here is what I'm talking about, the issue got a bit worse with increased sun intensity. Everything else look to be pretty well balanced for a quick scene setup?
I made spotrenders of the headlight with various sun emitter specular contribution:
0.01
0.1
0.2
0.5
1(default)
Also note how the metal gets noisier with increased specular level.
So yes there are numerous work-arounds but still wondering if something could be done under the hood of the PT arealight shader? Or am I still missing something, maybe in my workflow or surface/light set-ups? If I'd use this as a reference I'd end up using sun spec contribution in the range of 0.05 or so..
Sun emitter set up:
Hm, if specular rays don't have the same falloff as diffuse rays, that would explain why an emitter placed very far away behaves like this, and the PBR solution would be adjusting specular contribution like I did, or use a non- PBR falloff for diffuse rays
...
Edit: After testing it's safe to say falloff does not affect the diffuse-specular ratio at all. And I noticed with a falloff of 0 or 1 there is no light. Add to any of them a fractal of a decimal and the light starts working. But last time I played with arealight falloff was probably several awe builds ago, and then light intensity was a problem. Now you can apparently change falloff without having to adjust intensities much.
...so one last version where I solved the problem the usual way by adjusting everything one more time:)) Just kidding, I actually went through the car surfaces and reduced a bit of specular on the offending glass, used a sun emitter falloff of 0.00001, and specular contribution 0.3, 9 EV intensity scale. Frankly, I seem to get the same result with physically correct falloff (provided the emitter is very distant) but atleast now I've tested it;)
Raw render with default samples, 2/1 diff/spec override and 10x10 ps:
Well I really tried to create a scenario here asking for trouble.;) A Closed environment lit by two external emitters, and translucent curtains with opacity maps. Using GGX BRDF on the skin coat layer with about 50% roughness (map-driven). All I can say is the low overall noise levels surprised me, since these were rendered with 1024 Irradiance samples and diff/spec overrides 2/1. First I tried 256 hair samples but, as expected not quite enough, rerendered with 512 and that looks ok.
I think the new translucency is fabulous:) Don't know what you did there @wowie but it sure works beautifully with multiple layers + you can almost make things lookthrough much more convincingly now:)) I figure the forum rules won't allow me to raise the level further on that top;) With this type of scenes you just need to forget fast renders, though, but I don't mind. Forgot to check the log but in the range of 2.5h with 20x20 pixelsamples. And that hair has a million layers btw, really slow rendering (even after fixing those mid gray opacity maps that vendors still to this day love to use to make the hair appear soft, just saying).
256 hair samples
Hair spotrendered with 512 samples
A bit of postwork
Well, GGX is always going to be noisier in general, which is why I set the default to my custom Ashikhmin-Shirley ramp. I know Kettu prefers and recommends GGX. I don't share that sentiment though. Even with less than 1% roughness, it still can get too noisy for my liking.
The default should be close enough to GGX for most cases so please use that instead. I never actually tested GGX throughly with the new build.
As for the noise, I think your mask might be a problem there. As far as I recall, using that kind of blurriness on mask tends to produce some noise in renders. Even with my procedurally generated mask. For all practical purposes, the sun is completely round. We don't really see the corona and flares this far out anyway, so most of the feathering comes from the roughness or tail end of the specular lobe, not the emitter.
Translucency is definitely better.
Yes, I think the default is fine for most cases and use it for most things, this was a test and I can't say I notice any offending specular noise on her skin, really. But usually I save GGX for surfaces with 0 or close to 0 roughness, water in particular.
In my experience you can get diffuse noise with these kinds of masks, however raising Irradiance samples usually helps with that. I don't see them contributing much to specular noise? And I'd rather have some noise than a perfectly sharp and round sun reflection;) In that car render it's impossible to adjust surface roughness for a nice highlight without losing the sharp reflections. Btw the lighting was supposed to be "partly cloudy" with the sun covered by a thin cloud layer.
Anyways, the revised environment sphere shader may have played a part here, it is easy to get specular noise on metals and such if abusing the new exposure slider:) I think I know a lot more by now on how to use it;)
...some more experimenting with exposure values, masks and stuff...converted to .jpg due to size...
Re-rendered and re-postworked version with adjusted light levels and stronger volumetrics, default awe samples, diffuse/reflection overrides 3/2, 24x24 ps...having fun:)
Easy enough solution. Use two specular lobes or use the base and the coat. Turn off the reflection on one of the lobe or layer. Keep the reflection sharp and adjust the roughness of the extra specular lobe.
Got to try that, tks!
Wowie I think you need to check the aweSurface Options/Specular settings. Especially the Specular to Roughness thingy. It seems to produce fireflies a bit too easily. I just spent some time trying to eliminate the problem, using a simple two emitter setup. No env light. The emitters are limited in every way (color 200,200,200 and spec contr.50%) with around 7EV intensity, and scaled fairly large. And some bounce planes.
Ajusting SS parameters has little effect, I tried phase 0,7, 0,5, strength 50% etc. Turning off SS makes them go away, though. What really helps is reducing "multiply specular with roughness".
no SSS
normalize spec. off
roughness multiply off
roughness multiply 80
roughness multiply 90
And yeah, this is a very pale skin, I did nothing to gamma etc. saturation;)
Yes. I'm aware of that problem from your first examples. It's only reproducible in 4.8 onwards with 3Delight 12 and some specular maps. For now, I suggest not using multiply specular with roughness. The fireflies is on the specular lobe, so playing around with subsurface settings wouldn't change anything.
Minor problem, as long as one is aware of it;)
Except, as I mentioned, turning SS off also turns off the issue, so it must be related to SS on some level?
All in all, the update turned out to be a pleasant experience, the job you've done is incredible (considering you even apparently lost the source code at some point
) so hats off to you, sire
