I used to be a Linux-beta-tester of Realsoft3D long time ago. Even wrote a .r3i import-plugin for gimp once. I also got into RenderMan (RM), wanting to write a RM-binding (using RIB) for Realsoft3D. The RM-display-driver I got done and had RM (actually the free RM-implementation BMRT) render into Realsoft3D’s viewports. That was quite something! But as time went on, I watched blender accelerate in development and surpass numerous commercial 3D-suites in features… not to mention the user-community which formed behind it, after blender became OpenSource. It was obvious that Realsoft3D was a dead end for me, although I regard my CG/3D endeavours as a modest hobby at best.
Concepts for CG are the same, independent of the tools you use. Still, using fast and sophisticated tools makes your work/hobby more enjoyable. So I need to transfer my knowledge from Realsoft3D to blender. There are many disciplines in CG which you need to tackle, if you want to create good results. There’s the idea for an image or animation (which you need to cut into scenes to tell a captivating story), you need to build your characters/environment (modeling), then you need to “dress them up” (shading/texturing), the scene has to be properly lit (lighting) and you have to animate all of that (rigging/animation, sometimes involves physical simulation if you want to go the full stretch). All that is a huge pile of work, where I usually only enjoy the modeling-, shading- and lighting-parts.
Let’s get used to a new workflow then.
Thanks to Shawn Kirst’s normal-map plugin for Gimp 2.6.x, I created from this source color-map a derived bump-map and a derived specular-map. These put to use for a floor-material, together with a nice lighting setup in my scene, I get…
… as a result (it’s not about the monkey, but the lighting and floor-material).
Not photorealistic, but good enough to look decent. Such scene setups are commonly used to show off models you built in near life-like lighting-settings, without the need for hour long renderings. The above scene rendered in a couple of minutes on an intel Core 2 Duo (2.0 GHz) using the internal renderer of blender.
I also looked into luxrender, which is an unbiased renderer under the GPL, that comes with a blender-plugin. An unbiased renderer models its lighting calculations after the physical properties of light in the real world: wavelengths. It does not use a RGB-colormodel for its internal computations like many other renderers do. Furthermore the term “unbiased” means it does not make any assumptions or takes any shortcuts during the calculation. This has a huge advantage: true photorealism. The advantage comes at the cost of huge rendering times. An unbiased renderer is never done with an image-rendering. Instead it continues to improve it over time. After the first few seconds the image is very noisy and blocky. As it continues the noise gets finer and finer. You can let it go at an image until the noise resolution falls below the threshold of the images pixel-resolution. But that can take ages. What blender became for the 3D-suites in the OpenSource domain, luxrender will for the field of unbiased renderers I believe. I would not be surprised, to see it prosper as well as blender does. But luxrender has some tough commercial competition to beat Maxwell and indigo among a few others. Unbiased renderers are still new territory for me and I don’t wield this tool well yet, thus I don’t feel confident enough to show something in public right now.
