Tim Cook <z99### [at] gmailcom> wrote:
> render time 1:18:08.  three noncoincident boxes containing df3s (red,
> green, blue) for screen with emitting 10*each colour, 10 brightness
> factor for radiosity, 1600 samples, 0.1 error bound, 4 recursion limit,
> 6 nearest count, 0.01 minimum reuse, 0 grey threshold.

Tim, it's a crazy thing you tried. It did work out, though! The render time
isn't as bad as I would expect with such high settings. The emitting media is
meant to simulate a plasma screen, yes?

I'm not sure if you are aware of this, but there is a way to reduce radiosity
artifacts even when you've reached your 1600 samples limit. You can do this by
adding a tiny surface normal to your objects. I usually use normal{bumps /*.25*/
scale .001} and antialiasing with a low threshold (+a0.01). You should still
keep a low error_bound, but you can get away with lowering the nearest_count
down to 1 and maybe even use a lower samples count.

The biggest problem is reducing the appearance of noise, but with adequate AA
settings you can make things look pretty smooth. IMO, it's preferable to the
standard radiosity artifacts. I think even Thomas would agree :)

Sam

Post a reply to this message