Thanks Alain,

It does help tweaking various parameters and I got rid of the blotchiness I was
getting before although at the expense of rendering time. Wondering if there's a
way to render it quickly?

Cheers,
Ankur

Alain <aze### [at] qwertyorg> wrote:

> > Dear all,
> >
> > I created a small video out of Jaime's office scene (obtained from:
> > http://www.ignorancia.org/en/index.php?page=The_office ). The video is available
> > here at www.doc.ic.ac.uk/~ahanda/output120fps.avi . As you may notice, the
> > radiosity is flickering a lot at the back of the room. To let you know how I
> > rendered this scene, I used the following arguments to the povray commandline
> > tool
> >
> > (a) povray +I<input_file>  +O<output_file>  +W640 +H480 -d +RI<radiosity_file>
> > +RFO (for first pass)
> > (b) povray +I<input_file>  +O<output_file>  +W640 +H480 -d +RI<radiosity_file>
> > +RFI (for second pass)
> >
> > I'm a bit clueless as to why this flickering is happening. Could you please
> > suggest me some ideas to help remove this flickering in the scene? Thanks a lot
> > for your patience.
> >
> > Regards,
> > Ankur
> >
> >
>
> When using radiosity, the sampling tend to change for each render as
> they are taken, at least partly, at random.
>
> You can increase the number of samples taken, even going over 2000. Use
> importance to get more samples for bright objects or patterned ones.
> You controll importance as follow:
> In a #default block: #default{radiosity{importance Default_Importance}}
> In the objects: radiosity{importance Object_Importance}
> If you use a large number of samples, controling the importance becomes
> crucial to get reasonable rendering times.
> Bright objects and those used as light sources should have importance 1.
> A broad and uniform surface, like a wall or ceiling, can have importance
> 0.01.
>
> More samples will soften out the random noise.
>
> You can use +wt1 to limit yourself to only one working thread, but the
> render time will increase as you can no longer benefit from multiple cores.
>
> You can try a lower pretrace_end value. I often use pretrace_end 0.01,
> 0.005 or even 0.0025.
> Using a smaller value for low_error_factor may help. Default to 0.5, you
> can use something like 0.3 to 0.2.
> You can try increasing nearest_count up to 20.

Post a reply to this message