| GI Lighting Experiment (Picture in Message)
• | | | |
 | GI Lighting Experiment (Picture in Message)
by Mark Simpson on Aug 24, 2001 at 2:14:54 am |
Here's a picture of the experiment I'm trying to do in V7 w/radiosity. At this point I'd like to hear suggestions on how to improve it, and reduce render times. The majority of the render hit is when the portion of the image containing the glass renders. Everything else renders pretty fast.
Notice that there are some artifacts where the back wall meets the ceiling, and there are some weird distorions in the horizontal mullions encasing the glass dome. Also, there is a general blotchiness to the shading of the scene, which if subdued some will look nice I think, but it's too much right now.
The glass dome is supposed to be housing a glwoing ball with light rays shooting from it, which I have turned off for the sake of render times. I'm rendering another larger version with smoothing tags added to the structure (hoping to kill the artifacts), and the luminous object in the dome turned on.
In this image the lighting consits of an external omni light, with volumetric, hard shadow, and dust set to 100%. There is also a sky object with a luminance only material applied in which the luminance color is pure white, and set to a brightness of 750%.
The radiosity settings are strength=70%, accuracy=30%, prepass size=1/2, diffuse Depth=2, Stochastic Samples=100, Mimn resolution=5, Max resolution=65, recompute=first time.
Mark Simpson
| | | | |
• | | | |  | Re: GI Lighting Experiment (Picture in Message)
by SunnyC on Aug 24, 2001 at 4:23:35 am |
Hi Mark,
Nice image. Has certain ambience to it. To get rid of the blotchiness you need to increase your accuracy setting. 20-30% is fine for test renders, butI'd got for 50-70% for a final render. Apparently you need to keep your prepass at 1/1 for max effect (ref manual). Keep the depth at 3 at stochastic at 300 to begin with. Try raising stochastic samples for a better result. I tend to start with a min=2 and max=10 for test renders, then work my way up for final renders. What AA setting are you using? If you want more light to spill into the room, raise your radiosity strength to 150% and see how it looks. Be careful of over exposure. Having volumetrics and hard shadows will definately affect your rendering time. Is hard shadows absolutely necessary? Anyway you may have already know most this, but incase you haven't then I hope it helps.
Good Luck and make sure you post an update of the image.
SunnyC
| | | | |
• | | | | | Re: GI Lighting Experiment (Picture in Message)
by Mark Simpson on Aug 24, 2001 at 5:33:40 am |
Nice image. Has certain ambience to it.
Thanks, I'm trying to get a feel of a concrete wharehouse at mid day in the blazing desert sun.
Is hard shadows absolutely necessary
Well to simulate bright sunlight I would think they are.
I'll try your suggestions, although they sound like they'll send render times through the roof. Maybe I'll try a few tests with the glass dome hidden, so it won't be so brutal. You have given me some areas I can turn down though, so maybe that'll help.
Mark Simpson
| | | | |
• | | | |
Okay, here is a test with Sunny's GI settings. Things are a lot smoother now, but a bit bright and now the sunlight color is a little blue after I fussed with the luminance color, so I think I'll bump it back to pure white. Things are a little to well lit now, but I'm getting the idea a bit, so with some further tweaking, I can get it right. A couple more questions, the areas that really slow down the renders are the glass dome, and the light blue chrome railings. How can I adjust the GI in the 'illumination' channels of each so that I get some of the great look that radiosity does to the highly reflective surfaces, without taking such a hard render hit? Also, the light spill still looks a little grainy. Is this a matter of the 'accuracy' setting, or do I need to bump up the stocahastic samples or resolution more? Thanks, Mark Simpson 
| | | | |
• | | | |
Is that a production scene? If not, would you mind posting it so that I can try and get some good settings for it? As a quick fix I would suggest that you turn off recieve/generate GI for all of you windows. They aren't in the shot and don't really contribute too much to the overall lighting.
Cheers,
Donovan Keith
| | | | |
• | | | |
Hi Donovan,
It was good to meet you at Siggraph. That scene is just for learning, although if it gets to a point where it's cool and renders fast enough to use for a quick animation (camera moves) it might make it into a demo reel after fleshing out some more with some more props and the like. In any case I'll get the file to you.
There is glass and such in the actual scene, in fact I started a test render with everything turned on, but the prepass is taking more than an hour because of the glass and volumetric lighting contained in it. I'd sure like to find a way to make that part render faster (I will somehow before this is all over).
Mark Simpson
| | | | |
• | | | |
Donovan,
I sent you the file last night via the link on your name and it came back. Is there a better place to send the file?
Mark Simpson
| | | | |
• | | | |
Hi Mark,
When I registered for CreativeCow I mis-typed my e-mail address. Unfortunatley, I can't seem to change my email information. So, for now just send it to: donovan@tartarusdev.com
Cheers,
Donovan Keith
| | | | |
• | | | | | Re: Test #3 (Image in message)
by Mark Simpson on Aug 24, 2001 at 4:21:57 pm |
I interrupted this one mid render as it was taking too long (all night). But it's looking better. I like the way some of the color from the objects in the glass dome are bleeding on the ceiling and walls. So, I think I need to make adjustments to the radiosity and AA settings, by using render tags, and reducing the impact of the glass/chrome objects in the scene. Then I can go back to modeling and texturing a bit more. Boy, this radiosity stuff really becomes a painstaking process. Mark Simpson 
| | | | |
• | | | | | Re: Test #3 (Image in message)
by Mark Simpson on Aug 24, 2001 at 4:51:26 pm |
What seems to slow everything down the most, is when the glass dome is rendering. So, I put a render tag on it, turned off cast and recieve shadows, deselected 'seen by GI', and set force antialiasing to 1x1. In the glass material's illumination dialog, I turned off cast and recieve GI. I've turned off all refraction in the material. I'm using an SLA shader.
For all the chrome items on the dome, I used render tags to force anti aliasing to 4x4, and I turned off seen by GI, and receive/cast shadows have been turned off.
On the railings, I have turned on 'cast shadows' but everything else is the same as the chrome items on the dome.
For all chrome items, I reducted the GI influence, where still on, in the material illumination dialog to 25-50%.
I've also turned off the illumination of my external omni light (that is acting as a sun light)
After all this, the render still bogs way down as soon as it hits the glass. I would think that those items would render at the same speed as if the scene radiosity was not being calculated, but that isn't the case.
Can anyone tell me why this is, and what can be doen about it?
Thanks,
Mark Simpson
| | | | |
| |
|
