View Full Version : Trying to improve Maxwellrender withLW

03-05-2005, 02:11 AM
I found this thread where try of comparing a same scene with different render engines:


I try to fake (and improve) Maxwell GI with Lightwave. At this moment I diminish the render times reducing Ray Recursion Limit from 16 to 4 (about 8 min AA_Low, and 13 min aprox with AA_Medium)



03-05-2005, 05:23 AM

Nice render. Looks neat.


03-05-2005, 04:26 PM
Thanks Anesthan,
I've made a small update. The curious thing of this scene is that the whole effect of bounce lights is not made with lights, but with gradients :)


03-05-2005, 05:23 PM
Thanks Anesthan,
I've made a small update. The curious thing of this scene is that the whole effect of bounce lights is not made with lights, but with gradients :)


Looks extremly well but i'm not sure tha ti understand this render 100% :).

Is it FAKE radiosity (LW only) or is it Maxwell Radiosity with LW ?


BTW Is there a demo of maxwell for LW to try ?

03-05-2005, 07:46 PM
Thank Lewis,
All is made with LW. What I do was to take as reference this image made with Maxwell by Andronikos:

(As far as I know there is not an available demo version)

As you know, Maxwell render engine is based on the physics of real light, the problem is that it takes too much time for render (several hours); however with LW and gradients we can do it in quite less time :)


03-06-2005, 12:30 AM
Nice image, but the lighting is totally odd. The white frame on the wall for example gets a lot of light on the top edge in your rendering, although it gets only light at the bottom edge from the window, where its dark. The left side of the frame is totally white, although there is no light. Than i guess you use a dirt shader, but there are no shadows in "reality" in the edges of the room or at the bottomedges of the furniture. And where should the weird yellow caustics come from ?

I would suggest you do a "normal" grainy raydiosity in LW first to find out where the light travels before you start to experiment with your own lighting, that will help a lot. Otherwise interesting experiment.

03-06-2005, 11:38 AM
I like it. Even though it might not be accurate, it looks nice and most the time in viz work thats all that matters. Plus it works for animation which is great. :)

03-06-2005, 09:13 PM
Ingo: thank you to take the time of to observe and to comment. I agree with you about the frame, although I think that the bottom of the frame should be almost completely white (as Maxwell render). I don't understand well when you talk about the left side of the frame. If you refer to the frame (wooden), the left side is almost white because the light that enters through the window hits directly with the white wall and bounces in the left side of the frame (as Maxwell render); if you refer to the brightness of the side left (bottom) of the frame (wall to the speaker's side) this shine is justified because the speaker's sides are metallic and they work as a mirror or diffuse reflector that projects the light that enters directly through the window. The caustics is because the yellow sphere is translucent half, transparent half; as I said in the thread that I posted above, I know that this render is not accurate, but as Otacon says (and believe me, he knows what he says), most of the times in the viz work, advertising work and in several other CG areas, the important thing is that the image is agreeable and believable, not mathematically correct. But that is the advantage of this technique, you can achieve the look that you want and in little time. :)
Btw, if you want to try this technique and to make your own experiments, you can take a look at this thread:

If you can improve it (in some aspect) please, don't doubt to share it with us :)

Something that really bothers me and I am surprised that you have not mentioned, is the soft shades direction; they should be lengthened in direction contrary to the window's position; something that we could do more easily if the ref_vector/cam_direction option of ambient occlusion shader, worked with multiple cameras; however (with more work) it can be made with spinning lights, shadow designer, an arealight or baking the ambient occlusion shader (refvector/camdirection option) in textures.

Otacon: Your work is great man, I appreciate your comments a lot :)


03-07-2005, 01:02 AM
Hi Gerardo,

i was talking of the left side of the TV frame which is completly white, but nearly unvisible in the MW rendering. Also the bright spot on the white shelf next to it looks odd. And the caustics are unbelievable to me because there is no light reaching the ball directly, me thinks for the reflected light there is to much caustic. Also i'd like to see the rendering without the ambient occusion shader, its mainly useful for exteriors.

I agree that a picture doesn't have to be mathematicly correct, but at least it should be overall believable, thats what i miss in this picture. My guess is that you do to much or use to many tools to achieve a good result. Just look at otacons pictures, there are only one or two additional lights, thats all. That will work fine with standard raytracing too.

Of course i tested your nice "luminosity gradient raydiosity", especially nice if you have FPrime to tweak it.

03-08-2005, 02:01 PM
Now I understand to what you mean. In the original image those areas are not white:
That is because I added glow later with a very low treshold, I also think you are right about the caustics, I've diminished the intensity now, I think the image has lost some visual force but it looks a little more "overall believable", thanks for that :)
In fact I haven't made a lot to get this image and this technique doesn't manage many tools (at less not for me); is really simple, only LW gradients and an occlusion shader, no bounce lights, neither "luminosity gradient radiosity"; are you sure that you have tried it? The advantage of this is the simplicity and the control on any aspect of illumination, and you shouldn't worry about flickerings of lighting in an animation, what you see in a frame is what you get in the rest; also if for example, an object that affects a surface moves away from this in animation, its irradiation influence will diminish and vice versa. Besides you don't need FPrime, this technique works well with standard Viper :)