>> Yeah, each light source is a 16x16 area light. Given that the tree has 
>> hundreds of thousands of tiny needles - wait, just read the render 
>> stats! Look at the number of shadow rays, for crying out loud! o__O
> Over 4 BILLIONS!

Hell yeah.

>> I did also try this with radiosity - but it makes absolutely no 
>> visible difference. (Well, given how puny the light sources are, the 
>> extra bounces don't contribute much.)
> As Christmass lights are geting smaller and smaller, actual point_lights 
> should be "good enough", especialy if you have 100's of them.

With point-lights, the walls are covered by giant needle shapes. It 
looks very bizzare and unatural. With area lights, the walls are motled 
with shades of light and dark - go look at a real tree, this effect is 
actually pretty much right on!

Now, if I actually had *hundreds* of lights on the tree, maybe it 
wouldn't matter... but most sets of lights contain 20 or 40 lights, not 
several hundred. ;-)

> Don't forget to set fade_distance to a suitably small value.

Already done. 50 cm. (Fading = quadratic.) This makes a big difference...

> If manual placement is to tedious, why not try some procedural 
> placement, maybe integrated in the building of the branches.

I don't know what algorithm TomTree uses to built the branches.

> In real 
> life, the lights are often placed between branches.

Yeah... maybe I should just wrap a "string" round the tree and use a 
physics simulation to let "gravity" position the string? (I can use the 
trace() function to make it stop when it hits a branch.)

OTOH... it's not going to be fast!

> If the parceing gets to long, why not use a switch to turn off the 
> needles while you place and test the lights?

Already doing that. Takes 30 seconds to parse even without the needles. 
(Can't turn the detail down any further because I need to see where I'm 
placing things!)

Post a reply to this message