PDA

View Full Version : Big Question


sphere314
02-22-2005, 11:06 PM
hey everyone, long time reader, first time thread-starter.

ok, my experience in modeling isn't vast, but i have a good understanding of how things work. however, although this may be a tad advanced for where i am or what i need to do for classes and im sure this may require some long answers, i am still curious. So, how exactly would you start modeling a person? every time ive done it, which is about 5 times hehe, ive started with a poly cube and just extruded down and attached the arms in if i couldnt extrude them out. also, for the head, i assume you would build it separately and attach it, but how would you start that? im sure there are a million ways to do this, but im curious as to what is the best way. the "body topology" thread is very interesting, and it is a great deal of help, but i am unsure as to where to start from scratch, square one. the best head ive done, i never even finished, but i had started it with a nurbs sphere, adding isoparms and pulling CVs and such. i understand the concept of edge-looping also, yet i am not exactly sure how to implement it. i see everyone here modeling the most insane things and im having trouble trying to figure out the best way to make water from scratch for a sunrise.

i know this is a lot to cover, and i apologize for that, but i guess i have a few questions. dont worry, i wont ask anything about rendering... at least not now. Thanks.

Vertizor
02-23-2005, 08:17 PM
There is no one single "best" way to go about it. Everyone has their own preferences. But I'll do the best I can to summarize a few methods and the pro/cons.

Box modeling
This is the one where you start with a cube, slice it up, move verts around, slap a sub division on it to smooth it and viola!

Pros:
- You see and work with the whole mass at once which is great for maintaining overall proportions.

- If you find yourself focusing on, say the mouth, then you go and make the chin, and realize they're out of proportion of each other. By laying out the foundation and then adding details later, it's easier to keep things in proportion.

- If you measure "progress" by the number of polygons you have in your model, well this gives a false impression of progress because poly count can increase rapidly.

- Being able to see the whole mass of the object at once has a great advantage of predicting where new edge loops are needed, how they will flow, and help you better control that flow.

Cons:
- It gets messy quickly. You'll be looking at a lot of vertices and polys that will just get in the way when you want to focus on one specific area.

- Everytime you add a slice or add an edge loop, you have to follow that whole loop and correct the vertices or else they'll quickly bunch up on you and you'll just have to go back and fix them, it's tedious.

- You basically have to manage the whole object, a small change to one area might upset another, so you'd have to jump around a lot and make a lot of little fixes before really moving forward.


Polygon-by-polygon
With this method you basically start with just one polygon. Slice it, and/or add more as you go along.

Pros:
- In the early stages you'll be working with a smaller number of polys/verts. So it's easier to manage.

- Lets you focus on specific areas. You can model parts in "islands" like the eye socket, the nose, the mouth, those can be seperate "islands" of polys and later connected.

- Without all the extra polys getting in the way you are free to rotate around and really see the part you're focused on. You only focus on the polys/verts you add as you go along.

- Easier to work on parts that have to be dense in polygons, like the 3 major facial parts (eye, nose, mouth). The smooth "flat" surfaces like the forehead and cheeks can be quickly assembled once the harder parts are completed. Starting off with a really dense mesh will be harder to keep smooth later on.

Cons:
- It might be easier and more manageable in the early stages, it gets to be more tedious as you progress. Because you are adding a little at a time.

- Harder to predict potential "pitfalls" and eventually you'd have to go back and fix things anyways.

- Takes a long time before you start seeing results, that's just the nature of the beast.


When I start a figure head, I like the box method to quickly shape the head mass. Keeping it low poly (with sub division for smoothing) helps to maintain smoothness on the continous surfaces like the cheeks and forehead. But for dense areas like the eyes/nose/mouth I find poly-by-poly gives me more control and I can go crazy adding edges without worrying about messing up other areas. So my method is basically a hybrid of box and poly-by-poly.

If you have a 3D package that allows for n-gons (polygons with any number of edges) it easy to break the uniformity of "every poly has 4 edges." That way you can have many polys where needed, and low for smooth continuous surfaces. This is NOT recommended though unless you know you can get away with it. It would make animating a lot harder later on. You might get away with a few n-gons but if you're good you won't need them.

sphere314
02-24-2005, 02:52 AM
Thanks for your reply. but im afraid its only raised more questions hehe.

when you say "one polygon" for poly-by-poly, do u mean like just one face of say a cube and just extrude edges or just model a part, then combine it and merge vertices with another part? if thats the case, how would you know if all ur edges and vertices would line up right on the other pieces without having any n-gons? and is converting to sub-d the best method for smoothing? or do you just mean to smooth the polygon? wouldnt it be better to just soften the normals? sry, im still a little confused about the little things... o i use maya btw, if that helps understand where im coming from...

KolbyJukes
02-24-2005, 02:58 AM
- Takes a long time before you start seeing results, that's just the nature of the beast.

I agree with everything but this. I only model with poly by poly and have found it the quickest of all the methods. I build at full detail and see results immediately, as opposed to box modeling where i'm slowly 'chiseling' in the detail.

essencedesign
02-24-2005, 03:27 AM
I Agree with Kolby, it takes good visualization, but poly by poly, is the way I model to, Oh and I also use lot's of splines...they are great for getting a quick form similar to "box modeling" but the divisions are set in a much more fluid and intuitive way....It may seem daunting but just keep at it you 'll eventually be giving advice too.. :thumbsup:

Vertizor
02-24-2005, 02:08 PM
You guys (KWAK and essencedesign) are obviously more skilled/experienced modelers. For a beginner (which admittedly I'm still in that stage) it's harder to visualize the whole mass to do a poly-by-poly model. Just looking at the wip section it's evident that people who haven't mentally grasped the shape they're trying to make will end up with either: flatness when is should be more round, or lumpiness where it should be smooth.

when you say "one polygon" for poly-by-poly, do u mean like just one face of say a cube and just extrude edges or just model a part, then combine it and merge vertices with another part?
That's pretty much it. If you start with a cube, delete all but one face. If you program allows (haven't really used Maya myself) then you can just start by drawing that face point-by-point.

if thats the case, how would you know if all ur edges and vertices would line up right on the other pieces without having any n-gons?
The thing is, you won't know. Unless you were good and planned ahead by knowing where to add edges. Usually you'd just find ways of fixing it when it comes time to merge the pieces together. It's not the most elegant way to do it, but if you have this great looking nose that you want to add to an existing head... you'll find a way to make it work.

and is converting to sub-d the best method for smoothing? or do you just mean to smooth the polygon? wouldnt it be better to just soften the normals?
The ratio of normals to polygons is the same so you really can't just play with normals to make a "square look like a sphere" so to speak. Unless you're doing normal mapping. What little time I did spend with Maya, its sub-d tools is just like every other app. You have your base mesh that's however many polys as you've modeled, then the sub-d mesh which has many more polys and appears smoother and rounder. As you're modeling your low poly "cage" the sub-d mesh is updated because it's being controled by that cage.

sphere314
02-24-2005, 03:32 PM
i think im starting to get it now. so would you really build up all the detail as you go, or should you get most of it, but go back and push and pull vertices or split some faces later on to refine it all?

kwak:
i saw on your batman that when you were building the legs, you were building just the front of them at first, unless im mistaken. is that a smarter way? my experience, which of course is not nearly as much as yours, tells me that it would be better to select all the bottom-most edges and extrude them all down while "keeping faces together" or whatever. that way you avoid having to merge more vertices while moving around the leg, from front to back.

CGTalk Moderation
02-24-2006, 04:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.