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
And a switch to use the raytracer without being locked to the "progressive" mode with its quirks and no filters.
How high? =) 2048? 4096?
Higher.
Just in case anyone is wondering, AWE Surface does not have a clamp/limit on irradiance samples. But it's probably best to wait for the update regardless.
We probably never will. From what I see, with iray it perhaps best to over expose the scene in render and readjust exposure in postwork. You could do that with 3delight, but the difference in both speed and noise levels stays. I did dialed back on some of the optimizations I used though. Went too aggressive at some point and cut render times in half.
I guess 1/4 of iray render times will have to do for now. Probably best to just use up to date hardware.
I will have to find out which parameters are limited and which can be "abused"
, the end justifies the means every now and then:) Already knew about the Irradiance samples, I tried 8192 at some point. Also diffuse strength, opacity strength, SSS samples if I'm not wrong, and maybe something else... I know I'm not supposed to do that if I am to stay within the physically plausible, but...well... curiosity killed the cat
I plan on doing it, it's just paying work takes priority. ;)
And yeah, you use the same tricks as real world photography; use more, diffuse light and then adjust in post. If you are going to adjust tone a lot, it helps to have higher color depth, which Iray does easily with Beauty canvas.
A lot of people doing CGI work get weirdly hung up on simulating things exactly rather than using tools to produce effective results.
I hear you:)
This I totally agree with, been there done that, as I said in my previous post, the end justifies the means
Diffuse/specular/reflection strength is actually limited to 1. You can overdrive values from textures, but the actual input will never go pass 1.
Opacity strength is actually a color, but yes it is not clamped when viewed directly. SSS and translucency strengh isn't limited though, but using more than 100% isn't physically plausible.
Irradiance and SSS samples don't have limits, as you've discovered.
Here's the quick and dirty postwork 3delight shot, to get it somewhere close with the iray render with default settings. I used the default AWE emitter prop 1 at EV 7, which seems to closely translates to around 1000 - 1250 luminance in iray. The iray shot is the one without reflections.
You meant higher dynamic range, right?
Which reminds me I need to package one little scary arcane "render script" for my cozy little freebie thread... in the tradition of "DAZ Soon (tm)", I expect this to happen around mid-December. Cuz paid work and all that.
Yes. It's a lot easier to darken/lighten/whatever if the image isn't going to get all jagged because you're trying to expand 3 shades of black into an entire image.
Nice to see progress. Draft settings with default 128 irradiance samples. Still in DS 4.7 with a 3delight build that's now around 3 years out of date.
Current release build - 7 minutes 34.88 seconds
Developmental build - 4 minutes 42.22 seconds
Still amazes me to see 3delight's rendering potential unleashed.
Edit:
Finished more extensive testing.
With 2048 irradiances samples. Old - 32 minutes 25.45 seconds. Dev build - 15 minutes 15.10 seconds.
With 2048 irradiance samples and 8x8 pixel samples. Old - 36 minutes 41.9 seconds. Dev build - 20 minutes 27.12 seconds.
Well, rendertimes cut down by almost half is really impressive!
...meanwhile, playing around with transmission- and SSS-settings...
Ah neat.
The ball's emissive in the second render? Or this the no-extra-darkening in play?
...OK, this is getting me excited.
No. It's just having its global illumination exposure driven up. Both GI and specular exposure on the shader can now be used to over/underexpose the material even more. So I guess it has more extra non-darkening now.
Technically it breaks physical plausibility, but I figured I'd leave it up to users to use and experiment. As the saying goes, it is best to have it and not use it rather than not having it when you need it.
I've figured out how to implement some optimizations I put off with the release build. Technically, it's even more brute force and rely more heavily on multiple importance sampling. In effect, you can have twice the number of samples for equivalent render times. There'll be no change to the irradiance samples dial, but you can always deactivate the limits and push it up as high as you need it.
Right now, I see almost close to 1 : 1 performance between specular and diffuse. The biggest performance hit that's left is due to using more pixel samples, which is necessary to minimize noise from reflections. Have some more crazy ideas around to tackle that and perhaps optionally speed up some of the GI even more, but those are more difficult to work into the shader. Probably have to work on imager shaders now.
Thank you
Take your time
, things are getting better and better!
Here are a couple of experiments, trying to create fog with awe:) Did not let the renders finish.
I used 40 parallell planes with opacity maps, same maps also inserted in the translucency strength channel, both translucency strength and translucency shadows at 100%.
Having some problems getting the back lit version to work, with one plane I get the shadow right, not so much with 40:) Would it be better to go with transmission and tr. roughness? Render times are acceptable as long as I stick to opacity/translucency, did a quick test with transmission (IoR at 1), didn't finish it, naturally takes much longer. I turned off indirect lighting and shadows for the planes. Thoughts?
I confessed that I never looked into fog/interior/atmosphere shaders. I believe they currently don't work with mustakettu's render script.
This is rather handy if you're working with the planes method - https://sites.google.com/site/mcasualsdazscripts/mcjjet
...I wonder if Nerd3D's old Fog Tool Deluxe would work as it uses planes as well.
That's what I used, just modified the opacity maps and converted to awe. There are 40 planes and 2 material zones + a nice bunch of morphs;)
That's what she told me, so decided to try this. Kind of works, doesn't it. For a smoother fog, one probably could instance the whole fog prop. And also thinking of starting a nice forest fire
"Kind of" is very apt.
But that's not using the renderer to its best potential. 3delight can do proper volumetric rendering, even with .vdb, but the host app need to expose that.
https://twitter.com/jcubeinc/status/1056834832037113857
It should be possible to use volumetric/interior/atmosphere shader to do proper fog/volumetric rendering (at scripted renderer speeds) if DAZ have path tracer and ray caching enabled for the standard renderer.
https://twitter.com/pberto/status/756244464817934341
Just goes to show how out of step DAZ is compared to 3delight main branch. We're in 2018 and DAZ Studio still don't use the path tracer by default.
Wow nice render! Yeah can't help it, I feel the same way as that twitter dude:) It's unbelievable really, would very much love to forget all kinds of old school approaches
Thanks to mustakettu, one more bug resolved.
Current public build.
Development build.
Anisotropy with polygon meshes (using a sphere primitive in this example) now renders without jagged artifacts. Obviously, with subD the artifact would be less noticeable.
Very good news!
Found a nice M4 character in fast grab (https://www.daz3d.com/angus), had to try an awe conversion;) Made a G1 character to go with the skin. Rendertime 15 min with 16x16 pixelsamples..
Had to chop his feet off, still no shadowcatcher for awe, and didn't want to use IBLM:)
Out of curiosity, is the scene lit with just a HDRI or are there other lights?
Just HDRI;)
...and a WIP, 5 arealights and HDRI...rendertime with 8x8 pixelsamples 1h 40min. No issues converting this stuff, but need to see what could still be optimized to keep rendertimes down;)
But when I made a couple of pool scenes using RR3 by Marshian I was looking at 12 h IIRC:)