With soft edged RT shadows, is GI calculated after the light and shadow are calculated, or is GI calculated for each shadow recursion?

The render time seems to indicate the later.  I’m trying to optimize a drag and drop project set up.

Thanks,
Yon

Hi, Yon

Black Box - 20 August 2008 06:41 PM

With soft edged RT shadows, is GI calculated after the light and shadow are calculated, or is GI calculated for each shadow recursion?

The render time seems to indicate the later.  I’m trying to optimize a drag and drop project set up.

For GI reversed illumination shading (when GI ray hits something) soft shadow lights are calculated as not soft (one ray per light, no antialias). It’s not always “enough correct” even with 1 GI bounce. With 2 or more bounces it’s “always incorrect”. Better and faster solutions are coming in EI8

Thank you Igors,
as always, very direct and helpful.  I very much appreciate the inside view. It is much better to rely on facts, than anecdotal evidence.

Perhaps I could trouble you to clear up one more question.  Which is faster?  Rendering reflections in a mirrored ball (and then reversing the image), or rendering reflections through a virtual glass lens .  I want to recreate elliptical perspective, such as the eye or for an extreme case, a fish eyed lens sees.  I realize photoshop may do this fastest, but not in an animation.  I imagine reflection is faster, as it’s one bounce per ray.  It’s just hard to line up the shot in a mirror:)

I look forward to see what you guys are cooking up for EI 8.

Take care,
Yon

Hi, Yon

Black Box - 21 August 2008 08:32 AM

Perhaps I could trouble you to clear up one more question.  Which is faster?  Rendering reflections in a mirrored ball (and then reversing the image), or rendering reflections through a virtual glass lens .

At least for EI it’s same fast/slow because in both cases “secondary points” shading (reflected or refracted) is performed.

Black Box - 21 August 2008 08:32 AM

I want to recreate elliptical perspective, such as the eye or for an extreme case, a fish eyed lens sees.  I realize photoshop may do this fastest, but not in an animation.  I imagine reflection is faster, as it’s one bounce per ray.  It’s just hard to line up the shot in a mirror:)

We don’t understand too much in “elliptical perspective”, “fish eyes” etc - but intuitively we guess it’s what RT should do automatically - just because it’s RT

A common rule is: for drectly visible (primary) points all is optimized ("samples" in GI windows shows what’s calculated). But for secondary points no any optimization is performed . So, for example, if you see a scene thru RT transparent glass, so GI works with “Ambient occlusion” shader speed - all points are fully calclulated (slow).

Hello Igors,
That is very compelling point about GI, and pretty much keeps this form of perspective from being implemented with GI.  There’s always some little kicker.  Maybe this technique could work with V8 and the Photon mapping.

I’ve attached an image which shows what I’m after.  Elliptical means 4 point perspective.  It is closer to what the eye sees.  If you are 50 stories up and look at a building which is twice as tall as you, it vanishes in 4 directions (well really infinite) not three.  The four being left, up and down.  The only way I can see to do this in EIAS is to point the camera at a mirrored sphere. 

Maybe 4 point perspective can be a feature of EI 9

Yon