 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
After a couple of IRTC submissions that used radiosity, and then getting
comments suggesting I should try using radiosity, I mean to get it right
this time. I have been using two-pass radiosity. I will post the image on
P.B.I.
Here is my problem. If I use a lower-quality first pass (count 500
error_bound .4 at 100*60) My lighting is just as flat on the second pass as
it always was. So now I am trying to use a 500*300 with the following
settings:
radiosity {
#if (pass1)
pretrace_start .04
pretrace_end .01
error_bound .1
minimum_reuse .0025
nearest_count 20
count 1200
recursion_limit 5
brightness 1
max_sample 2
low_error_factor 1
adc_bailout 0.01/20
save_file "radios"
#else
pretrace_start 1
pretrace_end 1
count 1200
recursion_limit 5
low_error_factor .5
gray_threshold 0.0
brightness 1.25
error_bound .8
minimum_reuse .0025
nearest_count 20
low_error_factor 1
adc_bailout 0.01/20
load_file "radios"
always_sample off
max_sample 1
#end
}
but this, even with the simplified first pass scene, will take over 11 days
to render! By the time I get the second pass done, I'll have run out of
time. Have I forgotten something, or does a good render just take this
long? It seems like I have seen high quality images that did not take quite
so long.
Also, I am having trouble with the smoothed second pass. It seems to lose a
lot of detail in small areas (see the ceiling in the posted image)
I have read the docs several times though, as well as Tim Nikias' two-pass
radiosity experiments, but I'm still not quite getting that clean, crisp,
complex shadowed look.
Thank you for any advice you can give.
-S
5TF!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
I'm not an expert but I would lower your count by a lot. I saw no
visible difference in my POVCOMP entry when I lowered it from 1200 to
about 50, but it cut my time exponentially.
~Mike
stm31415 wrote:
> After a couple of IRTC submissions that used radiosity, and then getting
> comments suggesting I should try using radiosity, I mean to get it right
> this time. I have been using two-pass radiosity. I will post the image on
> P.B.I.
> Here is my problem. If I use a lower-quality first pass (count 500
> error_bound .4 at 100*60) My lighting is just as flat on the second pass as
> it always was. So now I am trying to use a 500*300 with the following
> settings:
>
>
> radiosity {
> #if (pass1)
> pretrace_start .04
> pretrace_end .01
> error_bound .1
> minimum_reuse .0025
> nearest_count 20
> count 1200
> recursion_limit 5
> brightness 1
> max_sample 2
> low_error_factor 1
> adc_bailout 0.01/20
> save_file "radios"
> #else
>
> pretrace_start 1
> pretrace_end 1
> count 1200
> recursion_limit 5
> low_error_factor .5
> gray_threshold 0.0
> brightness 1.25
> error_bound .8
> minimum_reuse .0025
> nearest_count 20
> low_error_factor 1
> adc_bailout 0.01/20
> load_file "radios"
> always_sample off
> max_sample 1
> #end
> }
>
> but this, even with the simplified first pass scene, will take over 11 days
> to render! By the time I get the second pass done, I'll have run out of
> time. Have I forgotten something, or does a good render just take this
> long? It seems like I have seen high quality images that did not take quite
> so long.
>
> Also, I am having trouble with the smoothed second pass. It seems to lose a
> lot of detail in small areas (see the ceiling in the posted image)
>
> I have read the docs several times though, as well as Tim Nikias' two-pass
> radiosity experiments, but I'm still not quite getting that clean, crisp,
> complex shadowed look.
>
> Thank you for any advice you can give.
>
> -S
> 5TF!
>
>
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
There is DEFINITELY a difference in my image. At 50, 100, all the way up to
about 600 it has flat shadows, artefacting in the corners, and all over
boring lighting. I need my count to be at least 750 to look OK.
-S
5TF!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
news:web.41bb4790420c6eda8f722dab0@news.povray.org...
> There is DEFINITELY a difference in my image. At 50, 100, all the way up
to
> about 600 it has flat shadows, artefacting in the corners, and all over
> boring lighting. I need my count to be at least 750 to look OK.
>
> -S
> 5TF!
>
how are set your finish{ambient}s?
if ambients are more than 0, you get luminous and flat shadows
set ambients to 0 unless you want texture to emit light (sky or light object
for instance)
if no ambient value is specified, default is 0.1
you can put #default {finish{ambient 0}} in global_settings block to set
your own default value
Marc
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Yeah, I had trouble with that previously, but the default ambient IS zero in
this image. Only the lights upstairs, and the sky are higher ambient. BTW,
it has speeded up a little, so it will now only take 8 days for a 500*300
render :(
-S
5TF!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
stm31415 wrote:
> After a couple of IRTC submissions that used radiosity, and then getting
> comments suggesting I should try using radiosity, I mean to get it right
> this time. I have been using two-pass radiosity. I will post the image on
> P.B.I.
> Here is my problem. If I use a lower-quality first pass (count 500
> error_bound .4 at 100*60) My lighting is just as flat on the second pass as
> it always was. So now I am trying to use a 500*300 with the following
> settings:
>
>
> [...]
Those settings don't appear very useful - start with the default or with
one of the settings from rad_def.inc and only modify something when you
know it has the intended effect. Simply randomly trying out numbers
won't lead you anywhere except to long render times.
Christoph
--
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 23 Sep. 2004 _____./\/^>_*_<^\/\.______
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
>
> Those settings don't appear very useful - start with the default or with
> one of the settings from rad_def.inc and only modify something when you
> know it has the intended effect. Simply randomly trying out numbers
> won't lead you anywhere except to long render times.
>
> Christoph
>
> --
Well, let me explain why I used them, and hopefully someone can tell me
where I went wrong.
radiosity {
#if (pass1)
pretrace_start .04 <--- The pretraces, I reduced in size to get
a little more variation in the colors and shadows of that big, blank far
wall, by avoinding large sample areas.
pretrace_end .01 <--- See above
error_bound .1 <-- I reduced the error bound until it looked good.
Larger is too low quality, and lower changes very little.
minimum_reuse .0025 <-- The default.
nearest_count 20 <--I was using this because that's what Tim Nikias
reccomended for 2 pass radiosity. Ensures more even samples, as I
understand it.
count 1200 <-- I used a high count because it looked better. Low
count gives the bland results I posted in p.b.i.
recursion_limit 5 <--- I have several panes of glass in the scene.
Without a higher recursion limit, the scene gets very dark.
brightness 1 <-- The Default
max_sample 2 <-- Increaced for the lights upstairs, which have a
high ambient.
low_error_factor 1 <-- Tim Nikias' value, again.
adc_bailout 0.01/20 <-- default, I believe.
save_file "radios" <--'Nuff said
#else
pretrace_start 1 <-- Pretraces set to one for second pass
pretrace_end 1
count 1200 <-- count set the same as above.
recursion_limit 5 <--same as above
low_error_factor .5 <-- Looking at this again, I realize I never
deleted this. Correct number is below.
gray_threshold 0.0 <-- Default
brightness 1.25 <--Looked better this way. The room is a little dark
w/o
error_bound .8 <-- Value multiplied to smooth out samples
minimum_reuse .0025 <-- Default
nearest_count 20 <-- Tim, again
low_error_factor 1 <-- Same as above
adc_bailout 0.01/20 <--Default
load_file "radios" <--'Nuff said
always_sample off <-- Two pass radiosity
max_sample 1 <--So I can see the partitions in the lights, now that
the light from them is calculated
#end
}
Thanks for your help!
-s
5TF!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
>
> minimum_reuse .0025 <-- Default
>
'scuse me. Turns out this value is NOT the default. I reduced it to gain
more
data, and smooth out the effects, so that all that interest in the shadows
I want looks OK.
-S
5TF!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Can anyone tell me what values they use? If mine are *that* far off, maybe I
am texturing improperly.
Thanks, though, for all the help so far. From the reactions, I obviously
don't have *something* right.
-S
5TF!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
stm31415 wrote:
>
> Well, let me explain why I used them, and hopefully someone can tell me
> where I went wrong.
>
> radiosity {
> #if (pass1)
> pretrace_start .04 <--- The pretraces, I reduced in size to get
> a little more variation in the colors and shadows of that big, blank far
> wall, by avoinding large sample areas.
> pretrace_end .01 <--- See above
Pretrace settings only influence the way samples are taken, you will
hardy achieve any general change except in artefacts. The pretrace_end
should be choosen with respect to the error_bound so the majority of the
samples are taken during pretrace. Your value probably is too high.
> error_bound .1 <-- I reduced the error bound until it looked good.
> Larger is too low quality, and lower changes very little.
Lower error_bound != higher quality!
> nearest_count 20 <--I was using this because that's what Tim Nikias
> reccomended for 2 pass radiosity. Ensures more even samples, as I
> understand it.
This first of all immensly increases render time and memory use. And
your argument does not even mention it increases quality in this case.
> count 1200 <-- I used a high count because it looked better. Low
> count gives the bland results I posted in p.b.i.
You mostly need a high count because your other settings require it.
> recursion_limit 5 <--- I have several panes of glass in the scene.
> Without a higher recursion limit, the scene gets very dark.
Transparent surfaces are influenced by max_trace_level but have no
relation to recursion_limit - this is only relevant for the diffuse
reflection iteration. The idea of two pass radiosity is to avoid
refractive/reflective surfaces in the first pass.
> max_sample 2 <-- Increaced for the lights upstairs, which have a
> high ambient.
max_sample *limits* the brightness of samples and does not increase it.
> nearest_count 20 <-- Tim, again
complete nonsense in combination with 'always_sample off'.
> max_sample 1 <--So I can see the partitions in the lights, now that
> the light from them is calculated
Changing this in the second pass is a possible artefact creator.
Again my recommendation: start with the default and only change settings
when you see they have the intended effect.
Christoph
--
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 23 Sep. 2004 _____./\/^>_*_<^\/\.______
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|
