After searching around for a better replacement for my kit-lens, I decided to give Canon EF-S 18-55 IS a go. The S designation after EF means the lens only compatible with APS-C sensor cameras. It’s cheap, very cheap lens, with image-stabilization to effectively maintain sharpness a couple of stops slower than without one.  It’s lightweight and plasticky, but slightly better than non-IS, especially the grip pattern on the zoom ring. Some reviews suggest that  it’s a much improved optical quality than non-IS version, supported by Imatest’s MTF findings in the lab.

Build quality aside, the obvious downside of this lens is its aperture range. With 1/3.5 on wide-end and 1/5.6 on the zoom, it’s definitely not a fast lens, hence the shot creativity is significantly limited. The stabilizer has nothing to do with this. For portrait shots within the range of 40-55mm where maximum aperture are around 1/5-5.6, the focus range isn’t shallow enough to isolate the focused object, and certain backgrounds would be distracting. That’s why I still have to rely on my 50mm fast prime-lens for portrait works. I would love to have a fast zoom in this range with a constant high aperture of, say, 1/2.8. But given the appealing price-tag on this so-called kit-lens you can’t lose anything. It’s a big bang for the buck only second to famous 50/1.8 prime. The stabilizer alone makes it a daylight robbery, considering lens prices these days. On paper, it’s a pretty versatile walk around lens, and time will tell how long I’ll keep this lens before I get a faster zoom.

The 0.34x magnification ratio is surprisingly handy to deliver pretty close up shots. I was lucky enough to get a fly to test how close this lens can capture. These pics are not “macro” shots, taken with by no means a macro-lens, with close-up sharpness that never hits anywhere near dedicated macro glasses.

It’s kind of stop-and-go modeling of Amelie, one of the most fabuleux characters in modern movies. I had to do it bit by bit between my projects, and had to post it—or I’d let the dust settle on it in my basement. Still I’m not satisfied with the textures, lips bump, and some other details. I’ll post some updates whenever I have a spare time to do it.

The mesh was sculpted with Zbrush

and rendered in XSI using mental ray.

Painted in ZBrush, grabbed as alpha, made it 3d. That’s that. Provided the shot was ready within minutes, I don’t think I’d get any cheaper wine than one (of course, it’s not animated).

A depth-of-field render on roughly 5 million triangles with 60 point-lights and as many shadow maps.

(Lighting: Binary Alchemy Light Rig on Paul Debevec’s HDR probe)

Jolie’s lips were used as modeling reference, though it’s not quite there just yet. Not that I’m a fan of her, but as far as the lips-thing goes, you can’t look any further than Mrs. Pitt’s stuff.

I did some tests with XSI hair. I’ll make natural hairs for this character later, for now I like it this way. The MuhHair shader is used for hair shading, and an extreme size ratio between root and tip makes it suitable for illustration hair look.

Human body is made of substance that scatters when the lights pass through it. Since the skin is not perfect opaque, this scattering effect should be visible and behaves depends on the skin position against the lights relative to viewing angle. For this scene I use EZ_Fast_Skin shader created by Emanuelle–a talented XSIBase member–as an alternative to the native XSI Fast_Skin Shader. It’s faster and, as suggested in the name, easier. The shader effectively calculates sub-skin scattering effects with an array of slots and parameters to simulate human skin shading.

The skin was unwrapped with RoadKill, a tiny handy free UV tool, to map three images for this object. (1)Diffuse, which was manually painted in ZBrush, (2) Normal Map, created with Z-Mapper, and (3) Specularity/Reflection map, which actually is modified normal map.

I used a single map for both reflection and specularity since those two are similar in concept. For physically-based material, specular means reflection. Reflection makes specularity of a surface, and then decays based on microscopic roughness of surface. A very sophisticated implementation of this effect is made by deploying Fresnel Law, with Index of Refraction (IOR) and viewing angle dependency. However in older materials, both specular and reflection co-exist. With this kind of material, it’s possible to have shinny material without reflection and dull material but highly reflective, which are against the physical law.

The good news is, it’s the first renderman interface to XSI to be available for public, as now being shared as open source. Being an interface, it requires a renderman compliant to render a scene, and thankfully, it couldn’t be more delighting as 3Delight is right there at the other end of the pipeline (again, it’s free for a single license). It’s been long time anticipated, as Renderman implementations have always been the first choice in big VFX studios since the very beginning of CGI for very obvious reasons.

Despite its great level of customization, Affogato looks no more than a black box for most regular end-users. It’s certainly not a single click rendering workflow. I once had a discussion with Raffaelle Fragapane —the very guy who was responsible for a large part of shots in Charlotte’s Web—in XSIBase, and he suggested that it’s a production proven interface (well, he’s been using it in real productions). It’s a basic bridge but nonetheless as powerful as any RenderMan-interface in philosophy.

One of main advantages using Affogato is its capability to produce RIB (RenderMan Interface Bytestream). In other words, it’ll need RIB-hacking to bring the best out of it. All I can get is gray phong for now, until I get into tinkering. Raffaelle was right about the importance of this RIB-hacking habit. It may be hard for a script-blind like me, but as an all-time coffee man, having affogato in the pipeline won’t be too far off.

Hello world!