 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Along some B5, Farscape and BSG maratons of the recent years, I've become
interested in developing all kinds of alien life forms of my own as well. =)
The character I'm showing here is modelled in TopMod and has 7 separate objects
that I've then combined in POV. Rendering is done in the latest RC of POV beta.
This is also one of my first experiments with SSLT. That in addition to
radiosity, area light and focal blur camera it surely took some time with my
Phenom II x4 925. Total rendering time was 07h 06min 08s.
I'm still trying to find ways to make the render go a bit faster yet not to lose
much in quality. But if anyone notices any complete oddities in my
global_settings, light_source or camera (for me it's been a lot of trial and
error), I'd appreciate all kinds of advices and tips & tricks. =)
global_settings {
ambient_light color rgb <1, 1, 1>
assumed_gamma 1.0
max_trace_level 5
radiosity {
brightness 0.85
recursion_limit 1
pretrace_start 0.08
pretrace_end 0.01
count 150
error_bound 0.05
nearest_count 10
normal on
media on
always_sample off
}
subsurface { radiosity on }
mm_per_unit 20
}
light_source {
<0,0,0>
color rgb 1.0
area_light
<-20, 0, -20> <20, 0, 20>
3, 3
adaptive 1
jitter
circular
orient
translate <0, 20, -20>
}
camera {
location <0, -1, -11>
look_at <0, -1, 0>
right x*image_width/image_height
focal_point <0, 0, -1.475>
aperture 0.35
blur_samples 5, 25
confidence 0.9
variance 1/100
}
As with any remarkably higher max_trace_level, count or blur_samples or smaller
variance the render time slows down to the ground and can take up to 16-20 hours
or even more depending on settings, so it'd be just excellent to be able to work
around it to have even a bit more sensible times for my scenes to come.
Post a reply to this message
Attachments:
Download 'alien_life_form_sslt_1024x768.png' (484 KB)
Preview of image 'alien_life_form_sslt_1024x768.png'
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 03/31/2013 07:22 AM, Pekka Aho wrote:
> This is also one of my first experiments with SSLT. That in addition to
> radiosity, area light and focal blur camera it surely took some time with my
> Phenom II x4 925. Total rendering time was 07h 06min 08s.
Nice job with the sslt!
> I'm still trying to find ways to make the render go a bit faster yet not to lose
> much in quality. But if anyone notices any complete oddities in my
> global_settings, light_source or camera (for me it's been a lot of trial and
> error), I'd appreciate all kinds of advices and tips & tricks. =)
I don't think there are /that/ many gains to be made ... other that
perhaps radiosity. Have you looked at:
http://wiki.povray.org/content/Reference:Tracing_Options#Radiosity_Options
also check out rad_def.inc (included with the distribution) maybe some
of those settings might give satisfactory results with less overhead
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"Pekka Aho" <pek### [at] gmail com> wrote:
>
> camera {
> location <0, -1, -11>
> look_at <0, -1, 0>
> right x*image_width/image_height
> focal_point <0, 0, -1.475>
> aperture 0.35
> blur_samples 5, 25
> confidence 0.9
> variance 1/100
> }
>
Superb image, love the floor it is standing on as well the contrast with the
alien is excellent. For me slighly less blur or maybe higher quality (by this I
mean blur sampes/variance/confidence etc) is the only suggestion I have but that
will certainly not help your render times and will only be a slight improvement
as the image is great.
Sean
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Many thanks for the feedback, cheers! =)
I looked at the radiosity values again, and tried once more with error_bound
increased from 0.05 to 1.8, I raised count from 150 to 200 though and also had
low_error_factor 0.6 added in there to see what happens. In camera I reduced
aperture from 0.35 to 0.25 just to finetune it a bit more. The result however
didn't differ that much from the first one as the changes were so minor
afterall. The render time was 06h 53m 56s so only a tad less.
I looked at the rad_def.inc as well and tried some of the alternatives, but in
the end I think my own settings made the scene more to my own liking.
Changes that I'd still like to try for slight quality tunings would probably
take me back to the good ol' 386/486 days for the rendering times, hehe. So I
think I'll just have to upgrade my computer at some point now. :D
Post a reply to this message
Attachments:
Download 'alien_life_form_sslt_1024x768_2.png' (485 KB)
Preview of image 'alien_life_form_sslt_1024x768_2.png'
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> Many thanks for the feedback, cheers! =)
>
> I looked at the radiosity values again, and tried once more with error_bound
> increased from 0.05 to 1.8, I raised count from 150 to 200 though and also had
> low_error_factor 0.6 added in there to see what happens. In camera I reduced
> aperture from 0.35 to 0.25 just to finetune it a bit more. The result however
> didn't differ that much from the first one as the changes were so minor
> afterall. The render time was 06h 53m 56s so only a tad less.
>
> I looked at the rad_def.inc as well and tried some of the alternatives, but in
> the end I think my own settings made the scene more to my own liking.
>
> Changes that I'd still like to try for slight quality tunings would probably
> take me back to the good ol' 386/486 days for the rendering times, hehe. So I
> think I'll just have to upgrade my computer at some point now. :D
>
In some cases, you can gain some rendering time when using subsurface:
In the global_settings, have subsurface{samples 25 25 radiosity on}
With low translucency, you may possibly go even lower. Using lower
values and high translucency can result in some interesting effects
around highlights, like a nice star shape...
low_error_factor default value is 0.5. In my opinion, you should not set
it higher than that, especialy when you have a multicores CPU.
For your area_light, at 3 by 3, adaptive 1 don't have any effect as it
start sampling at that value. Start with adaptive 0 and use a denser
aray, like 9 by 9 or 17 by 17. Only increase adaptive if you get shadoe
artefacts.
adaptive 0 start with a 2 by 2 aray.
adaptive 1 start with a 3 by 3 aray.
adaptive 2 start with a 5 by 5 aray.
To have some notable gain, you should use values that are at least 2
steps larger for the lightlets aray.
A low density area_light with jitter and antialiasing can be very slow
by forcing antialiasing for every pixel in the penumbra.
In this case, i'm not sure that the area_light is usefull at all. Try
using a simple point light.
Alain
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
nice. plus bnw. cheers.
"Pekka Aho" <pek### [at] gmailcom> wrote:
> Along some B5, Farscape and BSG maratons of the recent years, I've become
> interested in developing all kinds of alien life forms of my own as well. =)
>
> The character I'm showing here is modelled in TopMod and has 7 separate objects
> that I've then combined in POV. Rendering is done in the latest RC of POV beta.
>
> This is also one of my first experiments with SSLT. That in addition to
> radiosity, area light and focal blur camera it surely took some time with my
> Phenom II x4 925. Total rendering time was 07h 06min 08s.
>
> I'm still trying to find ways to make the render go a bit faster yet not to lose
> much in quality. But if anyone notices any complete oddities in my
> global_settings, light_source or camera (for me it's been a lot of trial and
> error), I'd appreciate all kinds of advices and tips & tricks. =)
>
> global_settings {
> ambient_light color rgb <1, 1, 1>
> assumed_gamma 1.0
> max_trace_level 5
> radiosity {
> brightness 0.85
> recursion_limit 1
> pretrace_start 0.08
> pretrace_end 0.01
> count 150
> error_bound 0.05
> nearest_count 10
> normal on
> media on
> always_sample off
> }
> subsurface { radiosity on }
> mm_per_unit 20
> }
>
> light_source {
> <0,0,0>
> color rgb 1.0
> area_light
> <-20, 0, -20> <20, 0, 20>
> 3, 3
> adaptive 1
> jitter
> circular
> orient
> translate <0, 20, -20>
> }
>
> camera {
> location <0, -1, -11>
> look_at <0, -1, 0>
> right x*image_width/image_height
> focal_point <0, 0, -1.475>
> aperture 0.35
> blur_samples 5, 25
> confidence 0.9
> variance 1/100
> }
>
> As with any remarkably higher max_trace_level, count or blur_samples or smaller
> variance the render time slows down to the ground and can take up to 16-20 hours
> or even more depending on settings, so it'd be just excellent to be able to work
> around it to have even a bit more sensible times for my scenes to come.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|
