POV-Ray : Newsgroups : povray.binaries.images : Reservations : Re: Reservations Server Time
14 Apr 2024 12:47:44 EDT (-0400)
  Re: Reservations  
From: Chris R
Date: 11 Mar 2024 10:30:00
Message: <web.65ef14787bcd374e24fcb13d5cc1b6e@news.povray.org>
Alain Martel <kua### [at] videotronca> wrote:

> > This is an update to the "Interior Brick Scene".  I moved the point of view back
> > a bit to see more of the room.  A bunch of the models were updated, including
> > the spilled champagne on the table.
> >
> > This was about a 2 1/2 day render on my laptop, and does not include any
> > radiosity.  I tried doing adding in radiosity earlier and the render rate
> > dropped to the under 100 PPM, which means it would take weeks to finish.  There
> > are a lot of light sources in the room, so maybe radiosity wouldn't add much to
> > the scene anyway.  Also, with the rough surfaces on the brick walls, I find that
> > I have to do a lot of tuning to get around radiosity artifacts.
> >
> > Anyway, I may do some renders from different points of view, and the letter
> > sitting on the napkin ring needs some tweaking so you can tell better what it
> > is, but I'm pretty happy over all with the results.
> >
> > -- Chris R
> Some times, removing most or all of the lights and using radiosity can
> improve the rendering performance. When you have a lot of lights, the
> shadow tests can really bog you down.
> Replace the lights with some high emission surfaces for the
> illumination. Bonus : Free soft shadows.
> Use the importance mechanism. The light need radiosity{importance 1}
> The count can be over 10000.
> Then, have
> #default{radiosity{importance Desired_Count/High_count}}
> Desired_count can be relatively low, like 80 to 120.
> Using the two values format for count usually seems to give better
> results. count 14000, 15557
> Also use the two values format for nearest_count. This improve the
> performance during the pretrace phase. nearest_count 3, 20
> Use a small value for pretrace_end. The default is usually to big.
> pretrace_end 0.0025 or 0.00125
> Use a smaller value for low_error_factor. In many cases, the default of
> 0.5 is to big.
> If some surfaces are causing to much problems, you can use two objects :
> The object that we see have no_radiosity.
> The fake object have a much simpler geometry and texture. A simple plane
> with a brick pattern with no_image no_reflection no_shadow
These are good suggestions.  I'm going to start doing some experiments replacing
the light sources with radiosity emission, and then playing around with the idea
of simple objects with no_image etc.  That would probably work really well with
the rough brick walls and should be fairly easy to implement.

I'll keep you posted...

-- Chris R

Post a reply to this message

Copyright 2003-2023 Persistence of Vision Raytracer Pty. Ltd.