View Full Version : Improving LW's radioisty, how-to

09-21-2003, 04:54 PM
Hi all,
if we'll not get a render-engine plugin class or a totally new, powerful engine, here are some little improvements which should improve GI usability within LW:

- Object based and Surface based sampling; this means we should have Ray x Evaluation, tolerance and Evaluation spacing values for every object and surface. In Lightscape, this was a very useful thing... if something complex is distant, we should be able to reduce these parameters on a surf/object basis.
Again, in architectural viz there are often very little things which does not require too much GI detail, but are very poly-expensive.
If these are envelopable, we could get a LOD by distance or by other parameters.

- Obj/surf based intensity: often a surface does not bounce enough light to lighten an ambient. By now we have to raise bounces (time expensive) or to raise light intensity (not controllable because it burns lighten areas).
A nice add would be a "radiosity intensity" parameter within surface edotor, so a wall can be normally lit, but it will bounce 500% light than we see.

- Ambient Lattice. Since ambient is kept in account by radiosity, think about controlling locally Ambient intensity with an object (invisible) with a Vmap. We could use also negative values to darken some places in a room, and strong values near doors to simulate and and easily control skylight. GI would look much more rich without too hassles and without more bounces.

- GI distance falloff; bounced photons should not hit other surfaces if they falls beyond a date distance. This would cut times killing a lot of photons which would not give a contribute to lighting. This could also related to a number of bounces; i.e, i can decide that ALL photons will bounce 2 times, and only then redundant photons will be killed.

There are some others things but these would be imho the most interesting.

Paolo Zambrini

09-22-2003, 11:31 AM
Could be very usefull a new pre-pass sampling. Best i can found today is used in Cinema 4D / Cebas Final Render. It can be saved to improve speed and make more consistent GI solution in animation process.

Other usefull thing could be the ability to set sampling value for first bouce and sampling vaue for secondary bounces.

Other good idea is that GI could be computed after image complete rendering (as option, of course). This avoid compute radioisty in not visible polygons, and his antialiasing could be independent of antialiasing of global scene. And radiosity pass could be saved in other image/redering pass.

Juan J. Gonzalez

09-22-2003, 12:03 PM
Hola Juan

A post rendering is tricky, due to reflective & refractive surfaces, cause they need to reflect/refract radiosity affected environment.

I agree about setting the rad-settings per object.
Much harder on a per surface setting :(

Sampling values for first/second bounce is good idea.

The most important aspect I encounter, is the "standardization" of results to avoid flickering, in case of interpolation.

Additionally I will add a better shadow noise reduction that will be calculated during the rendering, and not as a post effect that digests details.

Finally, additing inverse falloff or inverse square falloff lights, creates terrible noise, a NO-NO case for radiosity renderings.

09-23-2003, 04:01 AM
Great! Many ideas I had... but never posted! :p

There's a couple of thing NT can quite easily implement in forthcoming LW[8] (if they don't want to please us with a renderer revamp):
Pavlov's "Obj/surf based intensity"
juanjgon's "sampling value for first bouce and sampling value for secondary bounces".

Hoping we'll see them... at least! :D


09-26-2003, 09:04 PM
Yeah, I think a photon map pass or a multipass irradiance pre calc step in the renderer that then gets avaraged in the actual render would make a huge difference in quality and speed. This is what VRay does and it's brilliant. www.vrayrender.com It's even faster then FinalRender.

09-30-2003, 02:46 PM
V-Ray is amazing, maybe is not the most accurate of them all but it renders beautyfull images in no time...Amazing Gallery by the way.


10-01-2003, 03:54 AM
Ok here I go, COULD WE ASK ..... at LEAST... for multitrhread surface baker....

PLEASE, enough time lost...

Does one uses the bake illumination tecknik... ?

10-01-2003, 06:03 AM
Hervé could you explain or provide a link to the bake illumination tecknik?

10-01-2003, 07:47 AM
VRay is good for stills, since it bakes the irradiance for stationary objects, while animated objects are ignored or are not considered 100% into the GI-calculation.

LW is mostly an animation software.:D

10-01-2003, 11:20 PM
Hello Wizlon, here you go...

Be careful, there is a mistake in that tut..... When finished, put the luminosity channel to 200%, not 100%...... it's strange but I've seen this tut somewhere else, where it says 200 %, but can't recall where ... ah well, just rememnber that.... also I hope you're on a PC, coz you need to use HDRShop from Debevec....

About HDRShop, each time I see a tut that uses it, they always say, "remember, no commercial use...." let's be honest, no hobbyist would mess with that or so rare...." I think NT (the tutorial is on their Web site... sic.) should somehow clarify that situation...


BTW, anybody else using that teknik ?