Hi Mark, Henrik and Lee Hope more people will get back to using GI in Realsoft again soon as it is better in V6. Reflections are improved so not quite so many loops to jump through to get a decent image.
I thought I would attach this test scene I am still working on as it shows the bigger picture of what brought about the artifact discussed yesterday (now gone of course). It's a little dark at the mo, but I kind of like it so far. The car model is not mine, it's from the Dosch collection, but at least I managed to import something via sds plus plugin (Ronnie's plug in):) Obviously had to set all the shaders, but looks ok so far. Remember now in v6, the recursion level in the rendering settings defines the light ray bounces effectively, so better GI can be achieved in your scenes. This scene had a recursion of 6. Rendered at 3000 pix wide, scale 0.5 on a quad in 2hrs 12 mins. The only lights in the scene was a fill distant light at a very low intensity of 0.06 to initiate the shaders, the rest is illuminated by a single analytical cube imitating a light box above the car set to an illumination value of 1.8 1.8 1.7. That's it, the rest of the scene is as you see. No other clever tricks. I do like to set the camera flashlight very low also at '0.04 0.04 0.04'. Gi shader ray count was 6 and GI blur at level 6 and threshold 26. GI blur reflection at default. Have a nice weekend everyone. Regards Jason -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Mark Heuymans Sent: 17 April 2008 20:15 To: user-list@light.realsoft3d.com Subject: Re: GI rendering artifact fix >> As you can see in the attached pic, I am getting the white blur around >> the edges of the tyres on the floor. > > Yes, I remember this one - it is a surface normal issue. Rotate the floor > surface 180 degrees to fix the problem. You can also flip the normal > direction in the material assigned to the floor surface. > > The error is related to flat surfaces, hollow cubes etc. In other words, > cases where the surface normal side is somehow unpredictable. If you model > everything from solid blocks it will not show up. > >>>>>>>>>>>>>>> > > In this example, I changed the floor to an analytical cube and the problem > disappeared. > > Hope this is of use to some of you. Timo has kindly added it to the wiki > also. > > Regards > > Jason > I always assumed this only had to do with Blur, interesting to see there's more to it. When V6 arrived I spent many hours fiddling with the Blur Width control curve, and all sorts of combinations of blur width, strength, raycount, etc. It's certainly much better than it was in V5. But GI is still very slow in real-life scenes that involve more than a few simple analytics... I'll check this out in some test scenes, thanks! -Mark
<<attachment: Pixel Perfect _Audi TT _03_WIP_18-4-08_copy.jpg>>
