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
Yes!
Hmm, you said:Turn off the environment sphere. You want to isolate your indoor light first. Alternatively, apply AWE Area PT on the windows or place emitters outside the windows as a proxy for incoming outdoor light.
As I said, after adjusting the ceiling emitters with diffuse and reflection depth off, only direct light, no GI, I tried to make the windows emissive, following your instructions, started a testrender with only the emitting windows, took a long time, wondered why, opened geometry editor and saw they had a poly count of 1500, so I changed my mind since that obviously would not speed up rendering. Instead I used a acube with 5 poygons, the bottom one removed and scaled the cube so it was roughly a meter outside the windows. In my head that is the same as using 5 1-poly emitters placed a meter outside the windows, no?
Yes as I also already mentioned I rescaled the ceiling emitters to fit the light bulbs since they created enormous specular highlights in the window reflections.
I'm not at my DS rig anymore, it's rendering away. I'll attach some screenshots tomorrow.
By the way, and off topic, I really really hate the new forums
Wowie tks for all the help and for being patient with me:) I feel I have made some progress with this thing and learned a couple of new tricks. Eventually I'll get it the way I want it
No problem. After testing a bit more, I think I understand the problem and have come up with a solution. The current AWE AreaPT (and 3delight path traced area light example code) simply replace all incoming light, regardless if there's an additional light/emitter behind it. Mostly likely because that's what the 3delight developers did. Did some changes and now AWE AreaPT actually will 'stack' incoming light much like portal lights in other renderers.
I need to check on the shadowing too. It should also be fixed now.
Since AWE AreaPT is effectively a portal light, I'll be adding transmission with and without refraction, plus a multiply specular with opacity so it can serve double duty as actual glass window shader.
Anyway, the problem with your scene is basically all the surface inside your room can't find enough light. Especially since you want most of the light coming from outside. Once I updated the AWE AreaPT, this should no longer be a problem. It might still be a problem if you just rely on an environment sphere for outdoor lighting though.
I've also looked through the specular/reflection code once again. Mostly to see if I can improve the noise situation a bit. Although it will render slightly longer, noise should less noticeable in the coming update.
Wow, all I have to say about that
Superb!
And I will definitely remember this discussion the next time I set up an indoor scenario, will certainly do things a bit differently from the start. Still have that diner scene in the back of my head;)
Looking forward to hearing more about your findings:)
Ok here is the latest version where I used a cube for outdoor lighting (+ the dome at pretty much default settings). The cube with physical fall off at 3EV, ceilinglights resized at 12EV with adjusted poly shape, no blur and radius 40. I also adjusted some surfaces like the pool water, metals and desaturated the walls/ceiling, got tired of that green shade;) Rendered at 1024 Irradiance samples, non progressive 10x10 ps, rendertime 7 hours 39 minutes 46.88 seconds. Looks like you reallly helped me solve this, it's just a stupid scene but I hate having to leave unsolved things behind. I could still adjust light levels, not totally happy, but enough of Mr Sharky and Ms Rhino for now
!!
...and a screenshot of the emissive cube...
WIP
The Diner revisited. I've been spending most of the day testing various lighting scenarios. @wowie, you suggested adding low intensity emitters inside to illuminate the dark areas to fix the specular noise. However these emitters always cause unwanted reflections, especially in metals and windows. even at very low intensity. Let me first say that I followed your steps to the letter, turned off GI and reflections, the dome and external light, set the ceiling emitters intensity (and adjusted poly shape settings), testrendered (progressive 8x8ps, 1024 samples, 40 min), turned off ceiling emitters and set the external emitter plane intensity, new testrender, all emitters on-testrender, turn on GI at diffuse depth 2 and reflection depth 1, testrender. Zeroed the dome settings and turned it on, testrender. My conclusion is that if trying to use the external PT area plane at a rather high intensity it causes blown out reflections and specular noise, even if in theory it should help. I ended up keeping it at 3EV with specular contribution at 50%, and instead raise the environment exposure to 4EV. Btw, I used two environment spheres, the one visible to the camera is slightly over exposed the way I want it, the other one I used for indirect light and reflections. And as usual the dome image is a png. Ceiling emitters were at 12EV. The environment light is at default setting except for using a temp of 7000K. Override settings: 1024 samples, diffuse depth 2, reflection depth 1, 8x8 ps, non progressive render: 1h 13 min at HD pixelsize. I just can't get under 1h on my machine it seems, so it must suck:) The good thing: it looks clean, the bad thing: I would have wanted more shadowing from the window lettering like in my original setup. Using the environment light as a primary source for external light won't work for that, still happy to be able to produce a clean indoor render;)
Tweaking the skin and environment settings some more. Added the coat layer but either it's too strong or the coat fresnel needs more roughness or the backlighting is to high. The hair is one of those with a lot of semitransparent areas so renders slow. Tried to enable the opacity optimization at 100% and both filters at 50% and that helped some (cut off 15 min), but, not exactly remembering what the filters actually do, I could use some pointers:) And the hands could use some roughness maps, the face diffuse texture needs some gimping...
The updated AWE AreaPT as portal light should allow that.
As for the outdoor emitter, the emitter size will determine how sharp/soft the shadows you want.
I think what you want is to use two different emitters. One to simulate sunlight - so small, high EV and a larger one with low EV to simulate sky light. The 'sun' emitter should be higher up and pointed at a sharper angle, so you the light from it doesn't reach all the way into your indoor scene. However, the large one should be right in front of the window (or the actual window itself with the updated area light).
As for render times, it could just be a difference in machine specs. Judging by your render times, my machine seems to render at twice the speed. It explains the differences between your render time and mine on various test scenes.
Good points, but not going for sunny, more like partly cloudy, so might get away with using only one emitter...
Yes, using a late 2015 IMac with a 2.8GHz 2.8GHz quad‑core Intel Core i5 and 8GB RAM...seriously consider upgrading to a new Mac Pro with a 3,5 GHz 8‑core Intel Xeon W/48GB RAM, supports 28 cores. We'll see where this is leading me, not sure if DS4.9 will work under OS Catalina, will probably have to ask around in the forums...
Having two emitters allows you the flexibility to tweak the lighting. And you can always turn off the second emitter at any time.
i5? That explains it. I didn't expect i5 to be that slow compared to i7 though (with roughly the same generation). If you're not 'tied' to Apple hardware, I'd recommend going the Hackintosh route. Hardware wise, it'll be cheaper though you'll lose the option to use Afterburner and maybe Thunderbolt peripherals. I haven't used Hackintosh in a long time, so I don't know about wide compatibility with pre-existing apps.
Haha, yeah I've finally run out of patience, gotta do something about it. Will consider every option available. But thunderbolt is crucial, all my audio gear uses it.
I've been doing some further testing with various external lighting scenarios. First made sure that the opacity mask is working properly, so I used one small emitter, and yeah it works as expected. However I realized I might not want to use 100% translucency shadows on the windows.
Translucency shadows 0%
Translucency shadows 100%
Tried using two external emitters, scaled upt the small one a bit since I don't want sharp shadows, and turned on the larger one, the dome and internal lights, turned on GI and reflections and the noise is back. It's kind of weird, I mean the scene is literally flooded with light. Although I had to scale down the larger one a bit and move it to the side trying to avoid the external emitters overlapping.
And testing one large external plane + the dome without the internal lights, looks totally cool.
Same as above with internal lights on, better but not quite there...
Added a character and some stuff to check light levels, tweaked the lights some more...she seems to cope with the large external emitter gamma radiation quite well.
Made a full HD overnight render (7h) of the Diner scene with 10x10ps 2048 Irradiance samples, diffuse depth 3, reflection depth 2. The window opacity mask turned out to be too low resolution, (obviously mclean didn't expect end users to render it at close range, or just a design decision maybe? Well that's fixable). Basically it looks ok but there is still a bit of noise on some metal surfaces. The frame of the painting for example. Metals are more prone to specular noise than dielectrics? From my experiments it looks like reducing specular strength and/or metal to roughness on metal surfaces helps a bit? Not sure if I need to re-consider the lighting scenario once more or just try to edit the problematic surfaces...
Btw, I used 40% specular contribution on the external emitter to avoid overly strong highlights, maybe I should try that on the internal ones as well?
Yes. That's what I was talking about noise. Should be fixed in the next update.
Tks! Looking forward
Testrender of a 4K window opacity map, remade...
...meanwhile, in another part of the Universe...
Here's a quick overview of the opacity filter and what they do. Basically any input value below the first filter will be treated as 0 (fully transparent), while any input value above the second filter will be treated as 1 (fully opaque). The filter themselevs has a gradient though so whatever value you choose for the first filter will always be less than the second filter. The optimization will try to cutoff half of your input range.
Like this
You will likely be tweaking the second filter which will lead to a more solid output (to avoid excessive render times rendering less than 100% opaque and discarding areas with 0% opacity).
It'll be different with every map. Here's an example with hair, where you'll see more noticeable differences as you play with the filter values.
0% optimiation, 100 % optimization at default filter values, 100% optimizatin with the second filter value set to 100%. Opacity optimization blends between the original value and filtered values.
Ah, great, tks! I used 100% strength and both filters at 50% and that really helped, but maybe 25 and 75 would be a better startingpoint?
Short answer: it depends on the input values and target use.
For instance, faraway/out of focus figures probably can have 100% optimization with a very strong filter. But even then, it will depend on how screen space your model is on the screen.
@wowie
Please bear with me for asking for some more advice. I revisited my "desertstorm" scene to try and fix it up a bit while waiting for the next update. Long story short: Despite having set the environment and lights temperature to 8000-9000 K and arealights color to a blue tint I get this reddish overall "mist" that reduces the contrast and that I would like to get rid of. I have tried various tonemapping settings to no avail. On this untouched render I used Aces tonemapping and raised the lower luminance to 0.2. Had to render at diffuse- and reflection depth 1 to keep rendertimes down. The environment light settings are default, only the tonemapping and temperature altered. Saturation was 0.2. The environment sphere is a jpg with a dark blue ambient color. Environment exposure 1. I'm attaching a postworked version to show what result I'm looking for. I understand that a bit of postwork is required in most cases but I would like to come a little bit closer already inside DS. Any tips for setting up tonemapping etc? Or you think it's my lighting or surfaces? I've been thinking maybe overexpose the arealights and reduce scene exposure to see where it leads me...
Edit: I've been doing some more testing. It looks like increasing the arealights exposure and reducing scene exposure actually takes me closer to the result I'm looking for.
Seems like you've got the gist of it. You want to increase the actual lights not the camera exposure. Rather than adjust tonemapping luminance, maybe dial down your actual environment exposure so there's less on directly lit areas.
Ok, but so far I haven't really used scene- or camera exposure to add light to a scene, I've pretty much kept them at default values. I'm experimenting with a low light scenario now. The main light is a candle and I use 5 fairly large arealight planes at very low levels as fill lights. I set up the light intensity using default scene exposure and testrendered, looked nice. Then I used the intensity scale offset on the lights and added 2 EV to each light and decreased the camerabased exposure to match the original light level. I'm getting some odd results. Looks like the specular rays are not attenuted by the same rate as diffuse rays, so ended up with a lot of specular light. Is this expected behaviour or again something I'm missing? Do I need to adjust also the scene specular exposure?
The problem is your environment sphere. Sorry, but you need to STOP relying on the environment sphere. Especially at exposure 1. There's no point in fiddling with the lower luminance to darken less lit areas because you're already have too much uniform (ambient) light.
Either use a proper HDRI with high enough range or dial down the diffuse exposure on the sphere (as much as possible). Rely on area lights to light your scene and have enough surfaces for the light to bounce off
Let me see if I can explain this with esamples.
OK. Here's a render with and without diffuse on the environmet sphere enabled. There's a ground plane though and it's enabled on both renders. As you can see, most of the lighting is direct light even with a white environment sphere and ground plane.
If you want to have a flatter light, scale up your emitter and dial down the intensity scale. Hence you'll end up with something like this.
Place additional emitters or bounce planes/reflectors to your liking. Here's a very good example of a tailored lighting setup for a specific shot: https://www.artstation.com/artwork/QrNrGx
Scene specular exposure will not work when you enable camera based exposure. If you feel you have too much specular light, you can adjust specular levels through the light or through the scene specular exposure, or do it per surface. If you have large enough emitters, you'll get very large almost diffuse specular and those will appear dimmer than highlights from small emitters.
Tks, point taken. That's exactly what I'm trying to do here.
All I'm saying is:
I added 4 EV to every arealight in the scene. I turned off camerabased esposure and set scene exposure to -4. This gives a totally different result than keeping the environmentlight settings at default. And using camerabased exposure to reduce scene exposure by 4EV gives yet another result. This is not necessary a bad thing, I'm just trying to figure out what works best for me in any given situation.
The original render:
4EV added to lights, 4 EV subtracted from scene exposure, the environment sphere set to visible to camera only:
If I'm not mistaken there is much more bouncelight, differences in reflections and the SS on the candle is gone.
Didn't complete a render with camerabased exposure but a quick test showed that all reflections are more or less gone, especially on metals.
And I know 4 EV is a lot, just had to test how it works..
Edit: I was wrong about the SS. There's a glitch with the candle, if I close the scene and reload it the SS and reflections are gone, everytime. Have to reload the candle and reapply the mats and it starts working. Sorry about that! Really annoying;)
A sidenote...normal maps are still not transferred over from IRay Uber.
So this was done in Maya and postprocessed in Nuke? Do you know what rendering engine he used? I studied the setup and read the comments, interesting!
Wrapped this up with a larger size final render. Environment sphere off, 1024 Irradiance- and 256 hair samples, no other overrides, non progressive 16x16 pixelsamples. Same environment light settings as the original. What's interesting is that the SS on the candle lose its gradient at anything higher than 10x10 pixelsamples. With 12x12 you start seeing a sharp line. Any idea why this is happening? Solved by spotrendering at 10x10 ps.
Here's the original, the fixed one and a slightly postworked version to get some sheen on the skin. The skin looks dull and flat, so the light setup probably sucks or the skin settings are off, oh well...
@wowie
Decided to build a reflector like the one that artstation dude used. The two main problems to use it with your shaders as I see it:
1. Apparently there is no way of overexposing diffuse color anymore, not even with tonemapping off. Not even by turning diffuse strength limits off. Must be something under the hood? Makes it hard if not impossible to get a sufficiant brightness on the surface to show up in eye reflections.
2. Speaking of eye reflections, highlights from direct lightsources or surfaces using the environmental shader is no problem but reflections is a constant source of frustration. In theory it should be easy...transmission to 100% reflection strength to 100% adjust roughness. If I use an arealight plane with very high intensity to lighten the reflector to the max it barely shows up in the cornea. So the only way is dialing in 50% metalness. Or skip the reflector and use a direct light source.
It produces some nice indirect bouncelight though:) Here is a test with no environment light, only bouncelight. Cornea with 50% metalness:
This is about the max intensity I can get with default environment lightsettings, unless I scale up the reflector.
Thoughts?