"Subdivision modeling was developed as a quick approach to sketching. " - what? What kind of sketching? SubD modeling is a universal modeling approach, and is used for final images as well. It’s just it has its cons, especially when modeling products and adding small details.
"Uniformity: SubD performs well when designing organic shapes and where little variation is required (hence why it was primarily created for animated cartoons). " - it should be elucidated what “variation” means. It means that adding small details is more complex than with NURBS, where you can boolean small shapes easily, without thinking about loops. But good luck animating deforming NURBS. For animation I would convert it to sub-d poly first with auto-retopology.
Sub-d modeling is indeed used primarily for characters. It is also used for many objects, which don’t require the utmost precision. LIke, having many parts, but not complex ones.
But when you start having interpenetrating shapes, the sub-d becomes slow and unwieldy. This usually concerns product modeling and visualisation. So the article is correct, that for automotive concept modeling sub-d is not precise enough and cumbersome.
The article is pushing VR, but VR is not a modeling technique. It’s a way of interaction, like a keyboard or mouse.
I think a decent modeler would encompass several techniques:
- 2d sketching.
- Zbrush clay modeling
- Sub-d modeling and usual poly modeling
- Sub-d boolean interactive modeling, like Mesh Fusion in Modo.
- NURBS modeling for final precision (if it is needed)
It depends on what the final output is for. For characters - SubD., for product viz - NURBS or retopoed Mesh Fusion. Medium complexity objects - Sub-d or just poly modeling without subdivision.
Those are different industries, and different worlds.
Also, animation and texturing require some planning. As texturing NURBS is harder than Sub-D models. So I think people would retopo NURBS with auto-reopology tools prior to unwrapping.