View Full Version : Best to model with fewest "pieces?"

05-12-2004, 02:55 PM
Is it best to model things using as much contiguous geometry as possible?

For example, a building..just a big cube, but a few knife cuts and bevels later you can have doors and windows and cornices, etc..apply different surfaces to each.

Will such an object render more quickly than a building with holes cut into for windows, windowframes made with glass, separate pieces for molding, etc. Is it worth the time saved to strive for this kind of one-piece modeling?

If so, is this universally true for modeling?



jin choung
05-12-2004, 07:43 PM
a very good, essential question.

most of the time, it is irrelevant but generally, i try to keep things together when it makes sense to. i keep things apart if it makes no visual difference and will save me polys.

the number of unconnected pieces make no difference. but POLYS do.

sometimes, you will want you join things because instead of a sharp termination at the join, you want a molded, flowing join to show a weld, solder, glue, what have you.

at other times, you can save polys by NOT trying to bevel something out from an existing surface but just create a separate piece and just 'sticking it in' a FLAT PLANE (a single quad) with no connection whatsoever.


re-articulation of 'the rules':

1. model only what will be visually apparent (carefully considering all possible purposes that you intend to use it for)

2. economy is essential - modeling things or taking pains on things that will never ever be seen is a complete and total waste not only of your time but of the renderer's.

if you boolean a hole in a quad, you automatically create extra geometry. if it's not visually worth it to do that for the sake of 'joining' attachment parts, you are violating rule 2 but not gaining an inch on rule 1-

just keep the quad a simple, untouched quad and slap on the part.

your question borders on the concept of SOLIDS, which is geomtry type and methodology that is somewhat different from what poly modelers and nurbs modelers are (SURFACE modelers).

solids are shapes that must ENCLOSE A VOLUME and be 'watertight'. you work with these primarily by using boolean operations.... and i believe that the primary advantage is that such models accurately reflect how things are in real life and therefore are readily MANUFACTURABLE. (most if not all computer assisted manufacturing REQUIRES that all pieces be watertight solids).

another advantage is that solids have certain cool operations that they tend to do well like 'shelling' (taking a solid and then hollowing it it out so that it is a model that is a shell of a determined thickness.

you will never see free floating 'paper thin' polys or sides floating around a solids modeler.

but to my knowledge, most solid modelers are for manufacturing applications and are not widespread in entertainment/media.

also, sometimes, there are constraints determined by an application (such as a renderer [i'm thinking of game engines]) where consideration must be made to keep models, rooms, etc. as solids and as 'single, volume enclosing pieces'.

you don't use 'solid modelers' to generate such assets but they must be continuous volume enclosing models.

but if you are modeling strictly for a render, the original 2 rules apply.