trueSpace Radiosity Tests
Mostly modeled in Rhino. I am starting to experiment with
modeling in Rhino then doing quick renders in Carrara then more complex ones in
tS.
1st Radiosity tests:
The images in this table are tests using radiosity on a
Rhino model I built just for this purpose. The Solid plane is just large enough
to extend outside of an environment sphere I made using Infinity. The sphere and
plane were not excluded from the radiosity solution. I am not sure how much
effect the sky had on the colors of the building but am confident the brownish
tint visible in the underside and front of the roof are from the brown ground. I
can't think of where else it could come from. The sun light is very slightly
yellow and the sky light is very slightly cyan.
The first two rows are the final product. I
probably should have reduced the sun light more but I got tired of this test and
learned what I wanted to learn. I settled on using quality of 1. After a while I
found that convergence was set at 99 and because of the cryptic name I didn't
realize that is the percentage at which the radiosity solution quits. Much of
the action happens around 99.96% so that was too low. I set it to 100 so I can
just quit the solution when the test renders look good. I found the light spots
on the floor did not render unless I quad-divided it and the more I did the
better the square shapes looked. I found when I turned up the sky light too high
the detail near the bottom front of the roof would disappear like it was
smoothed out of existence. I had to keep in below one sixth on the slider. I
think the sun was set at two thirds on the slider, probably too bright as the
results seem to show. I found a quality setting of only 1 gave me a usable
solution as early as 40 steps or so. But continued to reveal detail in the
ceiling right up to 99.99 percent which was up between 5,000 and 8,000 steps. So
for some uses radiosity can be quick. For new users it is worth noting that I
consistently had to wait almost 8 minutes before the radiosity solution showed 1
step completed, but after that the counter started going up faster than 1 step
per second on my PIII 500. While it was doing that first step doing a
CTRL-ALT-DEL would show trueSpace not responding. This does not always mean a
program is not working (locked up) especially in trueSpace which seems to do
this in several cases. Busy and not answering your calls.
The third row shows why I needed radiosity,
direct light, and sky light to illuminate the important parts of the model. I
could have gone without ray tracing and added fill lights but to get the whole
model just right would have been a real pain. If you are doing animation you can
use this to create a lighting reference of how things should look so you have
something to compare to when you place fill lights for simple ray tracing.
Front view. |
Ceiling hole light. |
Angle view. |
Up at ceiling. |
|
|
|
|
|
|
|
|
|
|
|
Side view. |
Through skylight. |
Column bases. |
Detail view. |
|
|
|
|
|
|
|
|
|
|
|
Sun, No Radiosity. |
Sun. Radiosity. |
Sun. Radiosity. |
Sky. Radiosity. |
Sky. Radiosity. |
|
|
|
|
|
2nd Radiosity tests. Focusing on what radiosity
does to bumps maps:
The images in this table are tests to see how bump mapping
holds up when using radiosity. The texture is the Darktree flagstone texture. A
single distant light in the default, overhead position. No other objects in the
scene. I am not sure how adding objects or lights would affect things but more
even light will fill in shadows so it should make bump map effects less
noticeable they same as you would find without radiosity. This was meant to test
whether bump map effects seem to be present at all using radiosity.
tS default sphere. No Radiosity. |
Radiosity quality=9 full solution. |
Radiosity
quality=1 full solution. |
|
|
|
Cube subdivided 4 times in Thermoclay. No
Radiosity |
Radiosity
quality=9 full solution. |
Radiosity
quality=1 full solution. |
|
|
|
3rd Radiosity tests including my fake radiosity trick:
The images in this table illustrate a new trick I figured
out for faking some of the effects of radiosity.
Distant Light. No Radiosity. |
Distant Light. Radiosity. |
Distant Light. Radiosity. Too dark. |
Same with tone mapping set from 1 to 2. |
Distant light and two big area lights at
intensity+10. No radiosity. Coincidentally very similar exposure to the
one on the left. |
Same as on left but larger view. Note how
roof is lighter for the bottom 1 meter. This is where I placed the
down-facing area light. |
|
|
|
|
|
|
|
|
|
|
|
|
Area light
moved higher. Sun brighter, shadow mappings set to low detail by accident. |
Shadow map trouble with higher settings.
Can't remember which. |
Shadow map set to high settings. |
Trouble with shadow mapping producing wild
and unpredictable results. This is the best it got. |
Back to Radiosity. Increased mesh density on
all objects, especially floor. Sun=1.0 Sky=0.1 & no shadows. |
Same but close-up of pot. |
|
|
|
|
|
|
3rd Radiosity test conclusions:
Radiosity looks good but screws up bump mapping but its
fill effects are hard to achieve using fill lighting because they are nice and
uneven. Ray tracing without radiosity produces hard shadows that really don't
look that good but learning to control mapped shadows may be harder than dealing
with radiosity. My trick with area lights took so long to render it made
radiosity look a lot more practical though the textures looked better with my
trick since it was NOT radiosity.
In the future I will probably choose radiosity where bump
mapping is not important and ray tracing with area lights or light arrays to
produce soft shadows when bump maps DO matter. Later I hope to tame the wild
beast that is shadow mapping and turn it away from the dark side of The Force.
This whole odyssey has taken way more time than I expected,
partly because my desire to use models instead of simple shapes for my testing.
But I think the models served the purpose of showing me where and why I would
want certain characteristics in my light and shadows so it was probably worth
the extra time. These test also left me a bit depressed that CG is not as far
along as I had previously thought. I would much prefer to put up an object I
call a sun and have its light behave like a real sun in every way unless I
desire something different. Now that I know better how much will be wrong with
my images I am less excited about producing them even though I now know enough
to make them much better.