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
So today's experiment, trying to make a Garibaldi hair converted to mesh look good + testing a skin with a sunny HDRI. Using 4096 Irradiance samples on the hair and rendering with 12x12 pixelsamples and setting specular bounce depth to 16 seemed to clear the noise. Used 4096 Irradiance- and 2048 SS-samples on the skin, looks like I could have used 4096 SS samples, still a bit of noise. There's so much I don't fully understand about using HDRIs. The first thing I do after loading the aweEnvironment is setting exposure to max, anything else is just ridiculous and way underexposed. Then I go to the surface settings for the sphere and set gamma to about 1.5 and exposure to 1. Then I do a testrender. Is every HDRI I own of bad quality or is my monitor not properly calibrated or are renders supposed to be dark and washed out? I don't have these issues with area lights, only HDRIs, so wondering if it's something with the shader or 3DL or the raytracer or DS or if it's simply supposed to be this way=) After the latest update I thought maybe the camerabased exposure would make things easier, but it looks like trying to boost light intensity that way increases noise. Postwork is of course and option.
The first render is raw, no postwork. Env sphere exposure 1, gamma 1.5
After just stretching contrast and normalizing:
And played around with curves and levels to try to achieve the kind of exposure I would like to have:
Would really appreciate some helpful hints
Edit:
So I made one more (unfinished, had to abort and run) render with exactly the same settings, only loaded another HDRI
So is the first HDRI just really bad? With a totally f*cked up white balance? I get that the hotspot is far more concentrated than in the other one but still baffled by the difference and that specular outburst.
Goodnight
lurking but appreciating. those SS renders are amazing.
thanks for persisting on this already amazing shader.
--ms
I second that:) And very eager to try the new updated SS.
Well, working with HDRI is kinda a pain in the ass, as your experiment shows.
If the HDRI has a decent amount of range, raising up exposure will extend that range, but also slightly pushes up low brightness values. To compensate, you can raise the gamma which will darken low values and push up high values, almost like dialing up contrast. If you HDRI is well exposed, playing with the gamma can help avoid that uniform ambient lighting look, since with higher gamma values, you get more difference between the darker/lighter values.
Now, the big catch is that sometimes (or most of the time), you want to use the same map but use different settings between what you see in the backdrop/camera/reflections and what global illumination uses for indirect diffuse. For instance, you want to have reflections have a more dynamic, filmic look, but a more even diffuse lighting.
With the current AWE Environment sphere shader, you have to use two environment sphere to achieve this. I've already written the additional controls in my dev build, so far it works as expected. You can use sphere probes with specular/reflection to dial in values you want for the reflection, then enable the diffuse only correction and fine tune the IBL diffuse separately. The separate saturation controls should also help with some HDRI that has too much of a tint.
Right now, the additional controls are exposed as offsets and defaults to zero values. So, if you raised the exposure/gamma on the main controls, you'll need to use negative values to offset that for the IBL diffuse.
For consistency's sake, the shader will defaults to the old SSS. You will need to set 'Use Diffuse Texture with SSS' to 1 for the shader to take into account the diffuse texture's saturation into absorption values.
Left Default, Right Enabled.
Tks wowie! I'll do some more experimenting with double HDRIs. I'm confident that the next build will make it easier to manage HDRI lighting:) As for my (not so)artistic renders, I put my trust in the arealights more often than not. Really love to work with them and the results are much more predictable. Another question about the udated SS...clearly there was something else going on under the hood? You mentioned SS taking into account IoR now?
Sweet! And since it works with IPR, users can finally do their best in terms of colour correction. Hopefully.
Soooo assuming 3Delight works the same way, basically when we run a generic gamma-corrected render, we're not getting the full dynamic range into the imager. If the imager is to work as a genuine tonemapper, it may be worth it trying to start with a linear render without quantisation, same when rendering to EXR, and then having the imager convert everything to sRGB8 and feeding it to DS display - not sure what DS is going to think of it all, though LOL
Let's wait until it's ready yet and available. I'd say it's more robust and powerful, but there's likely a learning curve. I actually was slightly bewildered when I tried it out for the first time. But then I remember that I used a non zero value as the default in the offsets.
Yes. Since SSS in basically a layer underneath the base layer, the energy that goes into SSS will follow Snell's law. Before, I simply used 1 - Diffuse, but now it's a little more advanced than that.
Hey, it's up to them to do what they want. Been actually thinking about allowing overlay/underlay layers to do some quick and dirty compositing. The funny thing is since it's done in 3delight, compositing is faster than using DS Layered Image Editor.
Test shots.
Nice
So I continued to work on this character, trying to understand the grain problem. Same pose and camera, loaded a 4 arealight setup with a simple gray backdrop, no aweEnvironment light. Tweaked the skinsettings, basically reduced diffuse roughness, used both specular lobes with AshikhminShirley BRDF and made a roughness map for the face. Reduced SS absorption strength. Reduced bump strength. Hair and clothes Irradiance samples at 512, for the skin 1024 Irradiance- and 512 SS samples. Non progressive render at 10x10 PS, 20 min. Looks pretty clean IMO.
And this is what it looks like with the new skin settings and that HDRI
Rendertime 30 min
So if I want to use that HDRI the question would be tweak the skin or the lighting, right? Guess I have to try the double HDRI trick next;)
Hmm I wonder if GGX would work better with HDRI lighting
, another thing to test...
And indeed GGX seems to work better. Highlights are better with a more natural fall-off. Turned off one specular lobe, switched to GGX, slightly raised spec strength to compensate. Still looks pretty bad but maybe with some more tweaking
Rendertime slightly longer than with two lobes of AshikhminShirley. And now it looks like I have to make some changes to SS to get rid of the reddish glow. Maybe it wouldn't be a bad idea to save some shader presets for arealights and alternative ones for HDRIs
EDIT: Have a new render going with some tweaks and Irradiance samples/SS samples at 8192/4096, grainy as hell, I think I'm done trying to use this particular HDRI
Guess it was worth a try...but GGX seems to be worth some more testing;)
Ok so it really seems like that HDRI is bad. I tested her with an HDRI set I made just for skin testing. I've calibrated this one pretty carefully for brightness, gamma and whitebalance, and am happy to announce there were no issues whatsoever:) Pheeeew
Samples as low as 1024/512 and looks clean enough. 10x10 PS non progressive render, tested it with auto levels in GIMP and that changed virtually nothing. This is with dual lobe AshikhminShirley BRDF.
I played around with that HDRI as well. Weird as hell! It works as advertised in Marmoset Toolbag, but trying to use it either with awe-stuff or my stuff yielded super strange results. Like, with an aweSurface setup everything is tinted bright green (and I tested on a neutralised setup, no tonemapping, saturation or colour temperature). With my stuff (reading the map from the trace() call, not from geometry), it's just a bit cyan-ish, but it doesn't look like the sun is really all that bright somehow. A similar HDRI here -
https://hdrihaven.com/hdri/?c=sunrise-sunset&h=small_harbor_01
- is listed as having lower dynamic range, but the tones are what you'd expect, and the sun appears at least as bright.
I suspect the 2015 tdlmake we're using with DS might be playing up. I plan to convert the map to tiled EXR using oiiotool and hook it up to a RIB manually. Or maybe that built-in tdlmake can understand that it's a preprocessed file already and leave it in peace...
// I also wanted to see what it renders like in 3DfM, but Maya keeps crashing //
Tks Kettu, good to know I'm not the only one having trouble with it! It's a bit puzzzling, would love to know exactly what's so special about it
I DL:ed the small harbour one, haven't tested it yet.
In general most HDRI is fine for specular/reflections. You only see problems with diffuse lighting. The general solution to these problems (though not ideal) is to create a blurred version and use that for diffuse IBL instead. If you do the blur in a proper image editor that supports HDR (more than 8 bit floats per channel), there's less risk of cutting/culling/clipping the original luminance of the HDRI. You can check for this via the luminance histogram (which every image editor should have).
I have no idea how to fix the tint though. At least, not without messing with the texture too much. In theory, the tint comes from parts of the image that is not a primary light source, but they still have way too much luminance in the texture. For instance, blue tint comes from a sun/sky HDRI. What you want to do is retain the color, but filter out the luminance or scale it down a bit just for those areas (ie a gamma correction).
I'm working on a scene from inside a greenhouse atm. Searched through my HDRI library to find a nice sunny one with noon to 3 o'clock clear day lighting and sharp shadows. Tested with maybe ten different HDRIs (all of them from HDRI Haven) and everything just gets very noisy. Not clean and sharp if you know what I mean. Even metal parts have grain. Granted I use bump on the metal to scatter the reflections a bit. The tiled floor (with normal map and displacement) comes out very noisy even with very high sample settings. So just a while ago I gave up, loaded a jpeg into the environment sphere and added a PT arealight as sunlight. The result is like night and day, suddenly every surface works, minimal grain, sharp and nice render. The jpeg giving just the right amount of ambient light with default settings and the aweEnvironment exposure at 5. Also the specular highlights suddenly started to make sense. It's a bit puzzling. The HDRIs with a large hotspot and very soft light seem to work better in general. Anyway, I don't mind using the arealight, I find it much easier to work with, and the changes I make to surfaces make more sense. And using jpegs for ambient/indirect light together with an arealight seems to yield the best results for me, when it comes to outdoor lighting.
The specular output for HDRIs seems to be very different from the PT area light shader? But I admit that 99% of my HDRIs are free stuff, I only own a handful from DAZ, and haven't tested them lately, maybe they have better quality?
Well, this was not a rant, just some thoughts
Arealights rock
Edit to add: The fact that I'm shooting from inside the glass walls may contribute to the grain? I initially used the glass preset 1 and enabled thin glass, and noticed that I got no direct light. Enabling opacity at 0% solved that.
https://free3d.com/3d-model/cannabis-96626.html
Not a very realistic model, well it's free:) Textures need some fixing. Or I'll find something better. Tips are welcome:)
Anyway, here's the WIP of the greenhouse scene with a jpeg plugged into the environment and a PT arealight as sunlight, progressive render at 10x10 PS:
Latest GIMP builds work very well for that kinda stuff. If manual conversion via oiiotool doesn't help, I do plan to compare the maps - study their histogram etc to see where exactly they differ.
I'll also stick that lake map into other renderers I have handy.
It's all about the emitter size - like IRL, a large softbox gives you not only softer shadows, but highlights occupying a larger area (every surface is a bad mirror :)). So if like half the sky is much brighter than the other half, you have half the world washed in a highlight.
There was a recent thread here where someone complained about the HDRI quality in the store, got a link to HDRI Haven and ended up very happy apparently. Greg Zaal-authored maps have been the best out of the material I've tested (not DAZ, various other sources). But I do prefer at least augmenting photo-based HDRI with area lights - if anything, you have more artistic control that way.
If it has 0% opacity, it shouldn't be blocking anything, so probably not an issue.
The greenhouse itself does show it's a random free model, but the plants aren't bad. If the pots are UV-mapped, you could grunge them up nicely. Or if not, hide them and use better pots :)
Okay, having issues with oiiotool (gotta build from the source, the Windows binaries I found are too old). Ran a simple test in Carrara instead, lakes vs smallharbour. The result is basically the same.
I'm certainly glad I didn't spend more than a day (or was it two
) trying to make that one work. The small harbour looks healthy though. Let's see if that girl looks happier with it;) Only I'm on a quest for hunting down some more freebies atm. Lost in cyberspace LOL...
Ok, here it is, the small harbour, 10x10PS progressive render with Irradiance/SS samples at 1024/512, 13 min. No issues as far as I can see. Noise level as one would expect with those settings.
But this one doesn´t have any direct sunlight, so very diffuse. The lakes HDRI has some very brightly lit clouds surrounding the sun, and they seem to fire away laser beams that really mess things up=)
AweEnviroment exposure at 5 as always, exposure for the sphere at 1 with a gamma of 1.6:
I did wrote awhile back why I don't like using HDRI to light a scene. Ambient lighting isn't incorporated in 3delight RSL multiple importance sampling (MIS) framework. If you want to use MIS with an environment light, passing the texture directly either to trace (), indirectdiffuse () or environment () is the proper way. The problem is that 3delight uses Maya's coordinate system, which is different to DS.
Yes. I found highlights have more range and are visible at higher trace depth compared to area/mesh lights.
Same here. I tend to use the freebie Uffizi HDRI a lot, mostly because it's used to test renderers. There's also a free one I found on Corona's forum but I've forgotten how to find it.
That's because glass don't allow diffuse rays to continue. So to get direct light from the outside, you'll need to enable opacity and turn down the strength to 0 (zero). To still get reflection/refraction, disable 'Multiply Specular with Opacity' so you'd still get reflection/highlights/refraction. The proper way would be to have the glass shader refer to the environment map, but I've not managed to work out a solution yet that allows for visibility in the viewport and manipulation within IPR.
Aah, didn't know about the Maya thing.
Tks wowie, looks like we've come to the same conclusions
Just to be clear, technically it's not a Maya thing, but a world coordinate system thing. If you ever imported OBJs from other apps, you'll see the same thing. It's not just scales, but also axis. Though I'm not entirely sure, I think this is also explains alignment differences between the HDRI used by UE2 environment sphere and UE2 itself.
Here's a look at the problem. Left is the environment map lookup and the right is raytracing the actual environment map placed on the environment sphere.
As you can see, 3delight's native coordinate system uses the equivalent DS z axis as y axis (or the up direction for the world). Thankfully, the x axis isn't switched so left is still left, right is still right. At least with an environment sphere that has proper normals (pointing inside the sphere). I think UE2 still has the sphere normals pointing outward which is why HDRI images looks revesed when viewed from the inside.
Kettu's coordinate system hack (inside the render script) simply corrects this and made the corrections as a new coordinate system that shaders can use.
The correction itself is pretty simple. Tell 3delight to use DS native y axis as up and the z axis as forward/backward. If you look closely at the reflection, you can see the tower is on the wrong side as well.
Edit:
I think I've figured out a half way decent solution without resorting to coordsys in the render script.
So this is why it's not a straight forward thing to create a shadowcatcher that works properly with HDRI light? Also, are you saying that the DS coordinate system using the y-axis for up direction is not very common praxis in other 3D apps?
It's actually quite common, but the problem is that there's no easy way to manage it. At least, not one exposed to users.
I did look through some of DAZ Shader Builder macros and they're using pretty much what I was thinking about doing. Instead of building a corrected world space like what Kettu did, they simply realign vectors when doing environment map lookups. So, it only works in that specific case with that specific shadeop.
I think there was a change in 4.9 or 4.10 mentioned by Parris, but I forgot what that actually was.
The major downside with this approach (even with kettu's correction) is that there will be a mismatch/misalignment between the the actual environment sphere and the environment map lookup if you use the horizontal offsets on the environment sphere. While you can make changes to the environment sphere, you can't do it for the environment map lookup.
Another alternative is to use a much simpler environment sphere and apply the AWE AreaPT so you end up with one really big dome light. But as you noticed, some specular/reflection highlights are lost, so the environment path traced sphere should only emit diffuse illumination. Render time can easily increast to double or triple depending on the complexity of the sphere, so the sphere will end up not looking like a sphere at all. Reflections and highlights tends to get warped due to the use of simpler geometry, but that shouldn't be as noticeable with just diffuse light.
Or,just make necessary changes so you can use both techniques, caveats intact.
Tks wowie for this piece of information! Yes I've thought of testing the environment sphere with the areaPT shader but assumed it would cause very long rendertimes. For now using an areaPT lightdisc + the environmental shader on the dome seems to give nice results:)
Made this quick experiment...a plane a sphere and the aweEnvironment with an HDRI. The primitives using the default aweSurface settings, the exposure for the aweEnvironment is set to 5 and the surface settings for the environment sphere are default:
Then I applied the PTarea shader to the environment sphere and set light intensity scale to -1 to prevent overexposing. Now the primitive sphere actually casts a proper shadow onto the plane. Mighty interesting:) Render time was about 6 times longer, which was quite expected. Also interesting to see the difference in specular highlights and raytraced reflections compared to the diffuse light output.
The env. sphere also got a very light red tint with the PTarea shader.
That's inline to what I'm talking about before. Though my render times don't go up that much.
If you apply the path traced shadow catcher to the plane, you'll see it will catch shadows. You can cut render times by making a simpler environment sphere. I think a sphere with 12 The AWE Environment Sphere prop is basically a 20 meters sphere primitive with 19 segments and 20 sides, with a -1000 unit offset in y axis. Using a simpler sphere with less segments should allow for less render times, but still avoid excessive distortion.
You still need to use a HDRI with a strong sunlight to cast shadows. If the sun isn't strong enough, then you'll just have flat ambient lighting.