Nicolas Alvarez wrote:
> A generic way to see if a light affects a scene:
> Render a very small image for each light; each image having only one light
> turned on. Take the images that aren't completely black. Those are the
> relevant lights :)

I know which lights affect a scene. The problem is I have hundreds of 
lights, and hundreds of frames.

http://darren.s3.amazonaws.com/Castle/SavedOUT/index.html

Take a look at the "floor plan" link near the bottom. That's one floor. The 
big room in the middle has (about) 30 light sources and needs probably 200+ 
images, depending on which way you're facing and where you're standing. That 
floorplan image with all the lights for just one of the six floors turning 
on rendered in the pixels-per-minute timeframe. Turning off those lights 
while rendering (for example) the bedrooms (or the other floors, for that 
matter) speeds things up immensely, which is why you see so many curtains 
hanging there.

Right now, I run the macro expansion once for each room, then render all the 
images for that room. It would certainly be more convenient to run the 
expansion once and render the whole building, but that's not going to work 
unless lightgroups can very efficiently interact with bounding boxes to 
ensure the blocked lights don't take up a lot of time. It took about half as 
long to run one expansion as it does to render one frame of the smaller 
rooms, so it would have to be very efficient to make it worthwhile.

> I thought about it for re-renders of long animations. Render the whole
> animation at a tiny size, compare the tiny images with the ones I did
> before the last change, take the differing ones, re-render only those at
> full size.

Not a bad idea.

-- 
   Darren New, San Diego CA, USA (PST)
   The NFL should go international. I'd pay to
   see the Detroit Lions vs the Roman Catholics.

Post a reply to this message