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
This is not making the spikey thing go away. If I turn visibility "OFF" it stops emitting light in my Iray preview. Help?
By "spikey thing" I thought you mean the rays from the spotlight. I don't know how to turn the visible representation of point lights off without also turning off the light.
Oh. Bummer. :( I frequently will hang a tiny, very subtly point light near a figure's face to accentuate their features. Having the "spikey thing" there makes it really hard to gauge how much light is reaching her face (relative to the rest of the scene) because it blocks my view of the face. Grr. Well, Matte-object, transparent mesh light it is, then!
Couldn't you simply scale the point light down to a smaller size? I just tried a quick test and a point light in front of the models face appears to give off about the same amount of light whether the point light is at 100% or if its at 1% scale. You could then adjust the lumens if there is a noticeable difference in the amount of light Perhaps this won't give you the results you're looking for, but it seems to work properly.
Yes, I just tried that too, and it worked. No apparent difference in light output even with 0.01% -- doing IES profile tests at the moment, so I have the light shining down on a 50% gray panel. So, similar in functionality to setting a mesh to .001 or something really small to hide it, but still allow it to generate light.
Yes, that works for a point light, thank you! However, I've just stumbled onto something else. If I give the light geometry to soften the shadows (say by selecting "sphere" from the light geometry pulldown) there's no way to make the sphere invisible. I realize that unbiased engines don't want you to "cheat" with invisible glowing geometry, which is why it isn't built into the engine. But cheating is quick and convenient, and I can do that with geometry by making it a matte object and reducing the surface opacity to a very tiny number. Since lights don't have surface properties, I can't reduce the opacity of a light, and therefore have a big, glowy ball in front of her face. (Unless I'm missing something else?) Argh.
By the way, reducing the size of the light when geometry is selected doesn't impact the light emitted in any way, as you noted. It does, however, alter reflections, and it isn't possible to make the geometry disappear entirely.
The spotlights can use point, disc, rectangle, sphere, or cylinder geometry. But not a torus. So if you want to do ring lighting, you'll have to set up a torus primitive as a mesh light, or use an HDRI with a ring light. That's the only thing I'm missing with the Iray spotlights.
Sounds like I did the same thing you're planning to do -- I got a light set to learn from as well as to use. :-) The one I got is called "I-Radiance Studio Lights" by Dimension Theory.
I imagine all the other light sets recommended are excellent, also, but I'm really happy with DT's I-Radiance. I did learn the basics of Environments and Intensity and Rotation -- and once I understood the basics, I was able to branch out and start making my own lights, as well as to better judge which commercial light sets would be good additions.
Good luck! IRay lighting is a lot of fun. And I don't like fiddling with lights. :-) A note though...the I-Radiance Studio lights don't appear to use conventional scenic HDRIs. They're grayscale images or black and white, if that makes a difference. But there are free HDRIs out there. :-) I wanted a light set that'd illustrate concepts.
Under Light properties, Render Emitter: Off.
I did a few experiments with spotlights over the weekend. I found that the renders took appreciably longer with the spotlight geometry set to anything other than point. Four to six times longer. No other changes in the scene. Can anyone else confirm this?
Sorry, not my experience. Here's what I just got. Scene is a studio style figure. Clothed, paperroll backdrop. rendered at 1125 x 750 for test.
Dome and Scene, 3 spots (main, fill, rim)
As points: 47 secs. 205 iteration to 95% convergence.
As rectangles (100cm x 100 cm) 49 seconds, 207 iterations to 95% convergence.
This is typcal even for larger scenes. Not enough difference to matter, especially given the much nicer lighting with the "softboxes".
My test set up for this was quite different. Scene only lighting. 6 regular planes as an enclosure (4 walls, floor, ceiling). The "subject" is a sphere primitive. One single spotlight in the scene, aimed in the direction of the sphere. I don't have my tone mapping settings handy, I'll edit this when I can get at them, but I did adjust fstop, shutter speed and ISO to fit an indoor environment with a single light source. The light was 5500K, 900 lumens. I realize render time should reduce when adding lights, but I wanted to minimize the number of variables in the test.
Important note, I'm rendering this on 5 year old hardware, the video card is a GTX 460. So - super slow.
It went roughly as follows (I'll edit this later when I can access the actual test results), in terms of the spotlight geometry:
point: 1 m 30 seconds complete
disk: 3 minutes to reach 85% complete (cancelled render at that point)
rectangle, sphere: 6+ minutes to reach 75% complete (cancelled render at that point)
For comparison's sake, to test out mesh lighting, I turned the spotlight off, added another sphere primitive, set its emissive surface color to white, also 5500K, 900 lumens. I placed it pretty close to the "subject" sphere (I think -100 x, 75 y, 5 z). It took about 4 min 30 seconds to complete but the result quality was much poorer than the spotlight (very grainy).
I'll update with accruate metrics and perhaps some images later.
You need to run renders of actual scenes with definition and variation in the subject. Because of the way Iray reports pixel convergence, it takes appreciably longer to "render" just one all-white shape (tall rectangle) than it does of the exact same scene with the same rectangle plus a full textured and clothed human added. The type of lighting doesn't matter. I know, because I regularly do this. The actual convergence and Iray's calculation of it, as a percentage of completion, do not even come close to matching.
The spotlight shapes are built-in Iray area lights, and are defined in its programming. They are as efficient as the engine can possibly make them when used properly because they are native to the engine. Ttwo other main ight types native to Iray are point and infinite. These correlate in D|S as Pointlight and Distant Light.
I should also point out that by using an area light, the emitter is enlarged, and so is the coverage area of the light falling on the scene -- the same physical property that softens shadows when the area of the source light is increased, also spreads out the rays. The luminosity value of the area lights is for the entire emitter. To get the same light onto the scene as you increase the emitter size, you either have to increase the luminance, or reshape it to a smaller cone. A pointlight with an IES profile can do this in a controlled and preditcable way.
The 3D light shapes are more "lossy," because they emit much of their irradiance away from the scene. When you set the luminance to 5000 lumens, for example, that's 5000 lumens going in all directions, not just the part facing toward thre scene. The off-axis light influences the scene only indirectly, such as reflections off walls.
It's well known that renders take longer as the scene brightness goes down, so it's important to be watchful of the total scene lighting, not just the individual values of the fixtures. There are some light meter helpers for some of the other nVidia renders (Mental Ray for one), but Iray may not support it, and I know it's not in this version of D|S.
I'm trying to get insight into what the different properties of a spotlight do, what effect they have etc. So the tests I'm doing are serving my purpose in this particular case. I just found it interesting that changing the geometry had such a big impact on the render time (at least for me, I imagine the impact would be reduced for those who have modern hardware).
The hardware makes it relative. Even if you are only rendering with a CPU, or a card with few cores, the relative render times are more or less consistent. On slower hardware the difference are more pronounced, however.
You might find the following well-hidden performance tips article from nVidia helpful. Though it's counter-intuitive, bright white is hard to render, even when thoroughly lit. Note some of the setting suggestions, and the comment near the end about avoiding "glowing objects" (mesh lights) as they can be sources of further noise, which can impact convergence and render times. They recommend shape lights -- they mean the built-in area lights such as disc and rectangle. (Area lights are what they're called in the Iray programmer's documentation.)
http://irayrender.com/fileadmin/filemount/editor/PDF/iray_Performance_Tips_100511.pdf
nVidia themselves don't have many other end-user articles on performance improvement, but reading through the various forums (Iray has been around for several years, and Mental Ray/Vray for even longer) turns up some interesting gems.
Thanks for this link, the pdf is short but it does have useful info.
I have a hunch about the render time disparity related to the area light shape, which I'll try to test out tonight.
So my hunch was wrong. I moved the camera way back to take in the whole room, thinking that maybe the renderer was truncating calculations for light rays emitted by the point geometry that would leave the render port and not bounce back, whereas all the light rays are directed into the render port with the other geometries. Didn't matter.
Rendering with the area light set to point geometry took less than a minute to complete. Rendering with the cylinder geometry took nearly 6 minutes to get to 89% (I cancelled at that point). The image quality with the cylinder was much better than the point light though.
Now, to mess around with the beam component...
Try other shapes, too. I have a vague memory of reading someone saying that cylinder lightsource works slower than disk or sphere.
(I might remember wrong or the person was wrong, but worth checking)
Yes, as I mentioned a few posts back, my tests showed that point geometry was fastest, disk took roughly twice as long as point, rectangle/cylinder/sphere took roughly 4 to 5 times longer than point.
It's worth knowing for someone like me who's got old hardware if I'm trying to reduce render time (for whatever reason).
My experiments with the beam component property indicate that there's a somewhat linear correlation between increases in the beam component value and increases in render time. A low value results in a less focused beam with more area lighted and gradual transitions from light to dark and shorter render time, a high value results in a more focused beam with less area lighted and sharp transitions from light to dark and longer render time.
That's good information. I didn't realize cylinder/sphere/rectangle were mostly the same... which encourages me to use disks more (which I like, anyway).
Also interesting about beam component. Funny, because I actually prefer the more light/gradual/softer light.
As I noted above (I think it was this thread), if you want to control the profile of the light it's better to use a Pointlight with an IES profile. It's more controllable, and if it matters to you, more physically representative of a real light. With a profile, a Pointlight takes the shape of the profile -- it's no longer a 360 degree globe. Andy Grimm does a good job explaining their use here, and includes some examples:
http://www.daz3d.com/forums/discussion/59100/using-other-light-sources-in-iray/p2
A disadvantage of point lights is that they have no "camera" view, so you can't visualize by aiming through them as you can with a spot. You can, however, parent a camera to the light.
Rendering times with shadows and light falloff will depend on the geometry of the objects in the scene, and such things as displacement maps versus bump/normals, and SubD's. Given the "wrong" type of geometry, shadows and falloff calculations can seriously degrade performance.
As a side note, I'm amused at how often I go 'what the hell am I doing?? I don't need that light to be accurately modeled around a bulb -- it's out of frame, I can't see it!'
Not to mention how often I realize that a precisely modeled light bulb in frame is next to invisible within a shade and with all the light coming off of it, so I could just use a point-sphere.
Yes, use a point light and just adjust the Kelvin to whatever the mood you are going for.
This is a good thread! Lots of useful info! Thanks everybody!
I'm going to run my tests again tonight and report back with more accurate metrics. I'll also throw using an IES profile into the mix just to see what kind of effect that will have. Anyone have a particular favorite free IES profile they'd like me to use? Otherwise I'll just search around for a random freebie.
In my sig is a thread to an IES resource list...the Artist Friendly IES profiles would probably be the best for your purposes.
Gracias. Unfortunately I probably won't get around to messing with this until Sunday night. Real life getting in the way and all that.