Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
OK. I think I've been able to reproduce the fireflies on skin. It has to do with translucency boost with SSS. If you dial it down or set it to 0 (zero), they should be gone. Working on a possible fix, but it does slightly dim the boost.
Ok tks, will keep that in mind!
Test render of Rybolt Mechanical, I just love all those Nightshift 3D environments;) https://www.daz3d.com/rybolt-mechanical
10x10 ps progressive render
...re rendered at higher resolution with 12x12 ps (non progressive) and diffuse bounce depth at 6. Seemed to fix the slight specular noise...and the concrete floor looks smoother.
Tuscan Villa reworked for the current awe build...testing some gravel textures, better than the original but too rough...oh well, got to search the web some more...
...aaand testing another set of gravel textures & adjusted lighting...
...started working on interior + lights, and night lighting...quick testrender...
...
A night time update. Used adaptive sampling with 1024 samples on the exterior, and, as this scene renders extremely fast, used 16x16 ps, diffuse bounce depth 6, specular- and raytrace depth 16. Also started working on a more "physically plausible" moon...since the enviroment shader doesn't support displacement, looks like I'll have to use aweSurface and light it with something, I have a nice moon diffuse- and displacement texture, but it's 16k so a little overkill
in this context.
Here's the 16k moon texture in a quick test. Will need to play a little more with it...the displacement map obviously uses black as zero displacement, so the min- max values need some further tweaking. (Thank you NASA
https://svs.gsfc.nasa.gov/4720)
2k moon...with built in lighting
Decided to upgrade one of my characters for the newest awe build, and testing the AWE Hair shader, while at it;) Feels so good to get rid of those baked in highlights in the diffuse textures
She might be heading for the Tuscan Villa for a night jam...
Nice work, I do like the new hair shader. Makes a huge difference.
Tks! It does! Makes it a lot easier to salvage some of the old Gen4 props;)
Just an experiment...no postwork
Leyton Hair and Garibaldi brows with AWE Hair...testrender
@wowie
Would it be possible to make a hotfix for aweSurface? I'm having issues with fireflies on skin and clothes, and sometimes on hair with AWE Hair shader too. I've tried to dial down translucency boost but it doesn't seem to help much. I get them even with no environment/HDRI, just area lights. Or do you have some workarounds?
I love your hairshader
Removed fireflies in post;)
I can only fix if I can replicate the problem. Can you do a test render with the problem? Based on your previous comments, I'm inclined to think what you're having are 'just' render noise/speckles and not fireflies.
Plus you do a lot of your renders in progressive. As I noted on previous post(s), progressive will always be noisier than non progressive so there will always be noise (like in that Rybolt scene you posted).
Here's an example of false positive.
Most of the noise in the first render is caused by using too strong bump (100% @ -0.15 and 0.15). Once bump is disabled, most of it is gone. There's still some noise near the armpit/elbow tip though it's very hard to see with the forum compression artifacts.
I'm still working on the subsurface noise, which I did replicate and is due to setting translucency boost enabled (value bigger than 0).
Thanks. For all that tested it, any problems/reports/request?
Yup I'll post an example. When I zoom in on the bright dots they are one pixel size bright squares, so not noise.
I often do testrenders in progressive just to test lighting etc. But the fireflies I get in both modes.
I have had no issues, other than the odd firefly, but not sure that they are related to the hair shader or if it's the skin that triggers them. I may have a rquest or two, let me think about it
Ok, non progressive render, 8x8 PS, skin samples 1024 adaptive/ 128 SS, translucency boost 25%. Look at her right cheek, ear, armpit and inside of her left arm.
Untouched render:
Same settings except 0% translucency boost and 256 SS samples, still fireflies:
Also tried turning off bump:
I see it and I've actually never seen those problems before. Are you using DOF? Just checking to rule things out. Feel free to send the scene to me via mail. Even if I don't have the morph/textures, your settings should carry over.
Yes, DoF was enabled. I used a 3 area light setup + a jpeg loaded into the environment sphere with reflection/refraction enabled. I'll send you the scene file;)
Hmm, I made a version of the scene with just the G1 Base female with 30% youth morphs and the default skin. Applied the material settings to it (saved as a materials preset excluding textures) and now I'm unable to reproduce the issue. Will have to test some more before I send you the scene...
Remember that if you enable DOF, bump on material/surfaces near the camera's focal point gets 'amplified' a lot with path tracing. I measured something like 8x the strength compared to non DOF renders. If you don't account for that, it's very easy to use too much bump during setup/tests where you're not using DOF, then end up with too strong bump when you finally do a render with DOF enabled.
One workaround - you can somewhat avoid this by placing characters still in focus but slightly off center in your frame.
This behaviour also makes it very difficult to gauge roughness, since bump also affects roughness indirectly. I haven't fully explored the problem yet, so I haven't worked a shader based solution for this.
Anyway, I've fixed the ramp problem on AWE Hair and made other improvements as well. As expected, when fully using 3delight 12 features, it's possible to eek out a bit more performance from the shader. I think in best case scenario, it was a 2 min improvement out of a 9 1/2 min render. Render time for the DS 4.7 friendly code is roughly similar to the preview build.
Noise is very minimal at 4096 samples. By minimal, I mean the leftover noise you can mistake as just variations seen in actual hair strands.
Other notable changes:
After spending tonight testing the original scene, I think we can rule out DoF and bump, they have no impact on these fireflies. Only thing that makes them go away is turning off SS, so I dare say it's definitely an SS related issue.
I have pretty much tried every SS setting but can't get 100% rid of them. Among other things I reset everything SS to default values with t-boost zeroed, also zeroed deep color- and shallow color blend, zeroed use diffuse color with SS, non of which made much difference. So I started decreasing SS phase from the default 0.82 in 0.1 steps and testrendering. The lower I go the less fireflies I get but still one ore two left at 0% phase. Didn't go any lower at this point.
Zeroing the figure shape also didn't solve anything. Maybe something with the diffusemaps triggers this issue? They look normal, nothing extreme:) Will try some other maps tomorrow...
Sounds nice! Yea I found I had to dial down the specular stuff quite a bit to avoid that over shiny look with arealights. Haven't tested much with HDRI yet. I've gotten ok results with Garibaldi brows/body hair, I've just set diffuse strength to 100 and melanine/red melanine to 0, then adjusted color in the GB editor. Although I did some testing with forarms hair and the shadows looked a bit strange or noisy...maybe turn off shadows is a workaround. Not a problem with thicker hair like brows though.
4096 samples?
. Did you try that with some complex OOT hair? 
Another character tested under same light conditions/rendersettings. Hair, GB brows and fibermesh body hair and beard with AWE Hairshader. Skin SS phase 0.82, translucency boost 0, did get one firefly (right armpit), but not as bad as the girl. Go figure...
At least that narrows down the problem. I've never encountered it and haven't been able to reproduce the issue with any of the sets I worked with. Of course, I'm using the dev build so it might just be it somehow got fixed without me knowing it.
No, I was just using regular, old Charm Hair. It's a relatively simple hair prop, but handy since the hair cards are long and relatively straight. Makes it easier to troubleshoot problems especially in closeups, which in this particular case is related to noise.
Most of the time, the noise will be in indirectly lit areas, so that fits what you're seeing. Currently, the only way to get rid of that is to raise the samples. After revising the Marschner hair BRDF weights used internally by the shader, I am getting better results. Rendered hair colors looks generally better and much less prone to noise.
I've also worked out the overbright scalp problem, typically visible in poly hair props. The shader will automatically adjust translucency and secondary lobe strength depending on the proximity of the hair. This sounds pretty much like ambient occlusion, which it really is except it's not ambient, but rather proper GI. Hmm, I probably should rename it accordingly. This is free from a render time perspective. You will still be able to control it, though the default occlusion distance (0.1) should work for most scalps.
The occlusion trick seems to also be handy in adding just a bit of darkening to creases/wrinkles. I've added that to AWE Surface as well though by default its set to zero strength.
The Marschner hair BRDF reweighting also should solve the over shiny specular highlights. It also leads to another coming feature to AWE Area PT light. There will be a new specular and diffuse contribution controls added to the shader. You can basically adjust the specular or diffuse strength of the emitted light. This makes it much easier in simulating a common photography trick of adding a diffuser in front of the light (to tone down specular highlights without adjustime specular exposure or strength of the material. For simplicity's sake, it's currently implemented as a multiplier to emission on both sides.
I'm all PRO more options
Yup this will be very handy indeed!
I did this test in 4.7. Below a render also done in 4.7...no problems. Don't know what that means, will try to re render exactly the same scene in 4.9.
So I opened and rendered that exact scene in 4.9 and I do get fireflies.