Polygon Modeling Practical Basics

By Ali Ismail

Sections

- Topology

I - Support Loops

II - Hard Edges

III - Tris & NGons

IV - Boolean Operations

V - Crease To Blended

Introduction

It's quite common for someone who starts out with modeling in polygons, to find some challenges in completing a high quality model, which is a faithful representation of the reference and have a topology that is suitable for the task and on top of all; finish in a timely and efficient manner!

There are a lot of tutorials on how to model specific objects, and ample material as well on theory, using terms which might seem too intricate for someone who is not familiar with them. In this article, I hoped to give some of the practical basics to keep in mind when modeling, and avoided being too technical.

Actually, modeling is far easier than most believe. But, like anything it needs some practice and time.

Perception

In the tutorial, 3D Modeling Fundamentals. I emphasized the importance of perception and proportions for a good model, here again I would have to state that most modeling difficulties are usually found in how you see something and translate your vision into an "understanding" of the model. Tools are only secondary.

Traditional artists went into a lot of effort and different exercises, to improve their perception, doing things as far as fetched as trying to feel touching the object while drawing it. Anyway, I will not go any further into perception in this tutorial. You can check my 3D Modeling Fundamentals Tutorial, watch this primer video, learn more from any traditional art source or simply practice, practice after all makes perfect as they say!

Topology

Topology can simply mean how things are connected, in a way, think of a mesh wireframe. to be quite frank with you, it's not a science in it-self. At most, there are few rules of thumb that you can use as guidelines, if anything, it's an art.

Suppose, I give you a scanned 3D model, and let you do all the retopology, will you find it a difficult task? if not, but you somehow find it difficult to 3D model in general, then you better work on perception; Apologies for repeating this one more time in a tricky way, but I had to make sure to state the difference.

What would be the worst thing that could happen if you did a wrong topology? not much really, you can always correct them and if you do a wrong topology for the most important part that needs a topology which is the human face, and it somehow passed all the directors and animators and went on screen, you are only going to have a few unconvincing facial features or glitches here and there, it's not a matter of life and death :) .This is not in anyway to depreciate our craftsmanship, but It's good to remember, so that you work with enough relaxation for optimal performance.

Study Wireframes

The beautiful thing about working in 3D, is that everything is out in the open, and you can learn anything you want quite easily.

are you ever curious how Pixar wireframes look like, well. Look them up! how about high quality scenes, vehicles and characters. It's all out in the open. There is no shortage of wireframes and 3D models you can download.

You should always study all the material you can find. Look up VFX breakdowns, making of and anything you can get your hands on.

Topology That Works

Any topology or wireframe that achieves your goals, is a good topology. Always consider what you are trying to achieve and time and resources you have.

Drone Camera Wireframe

For example, the topology above is what can be considered a perfect clean topology, but in that case, it was the wrong topology! because I didn't need to go into all of that effort to model it, for that particular project, I could have managed with a high-res mesh and used boolean quickly. The small glitches and artifacts would not have been noticed by the viewer. And even if it did, it might have had the effect of adding variations since it was a dirty reflective model.

Looney Toon Car Wireframe

Is the topology for the car above wrong? Just because there are tris, it doesn't mean it is a bad thing. Perhaps for this project, I only needed this level of quality, and using tris and ngons represented a faster and more efficient way of working.

I am not advocating the use of tris everywhere and every time, but, I just suggest that you know your goal in modeling before starting, and then deciding what method will achieve the best results.

I attended a talk once, by Steven Stahlberg, and when the talk finished, all I wanted to ask was: how dare you sir, use tris and ngons on a character? all of my indoctrination told me otherwise. Well, I held my mouth shut to avoid making a scene, Now, I can finally see his point. Use what ever topology that works and you decide for yourself. I would still avoid ngons on a character, and use tris and ngons carefully on meshes, and generally speaking avoid having a tri or an ngon on a curved surface, but you should test for yourself and decide if it works or not.

These links are quite interesting:

Ngons and Tris: Sloppy modelling, or intelligent strategy? | Polycount wiki Topology | OpenSubdiv Modeling Tips | Multisided Polygon by Andrew Silke.

Animatable Topology

There is no shortage of theories and talk on topology for the human face and this and that. (CGS Thread | Polycount | Head Tutorial). for the beginner this might sound overly complicated, and it is! there is only one thing you need to know about topology for animation (Deformation) and that is: will you be able to deform it correctly?

Deforming Box

You see how it is quite easy to bend an object when the loops are placed in the direction perpendicular to the deformation, when they are not, it becomes quite difficult, and if you have no loops at all, you can't even make a deformation.

This is really, the whole deal about topology for animation. Any topology that will enable your mesh to animate and deform correctly is good enough, of course it is quite tricky sometimes with heads and joints, especially if something deforms in multiple directions. But, hey, you get the gist of it now and are very much capable of studying any wireframe and making your own decisions. Don't let anyone convince you otherwise!

Rules Of Thumb

There is nothing new under the sun, all the complex polygon models are made up of smaller features that usually will go under one of these categories:

I-Support Loops

This one, is used a lot in hard-surface modeling, Some begginners forget to use it sometimes. If you want a sharper edge, you know of course how to add one or a double edge loop

Support Loops Test

But when you do add support loops, you risk deforming the mesh, please remember that any extra loop you will add, will change the curvature of the mesh, If you want to add a loop with minimal damage to the original mesh; it's always better to add it at higher subdivision iteration.

Support Loops Deformation
Curvature in Subdivision

II-Use Hard Edges Loops

A technique beyond support loops, using hard edges or opensubdiv, can save you the effort of support loops and give you fast and efficient way of modeling, especially at the beginning phases.

Not only is it time consuming to add support edge loops but it would be much harder to modify the base mesh. the alternative? try to use hard edges/smoothing groups with conjuction with mesh smooth and work on phases or stages as much as possible; getting first the basic details then moving to higher resolution and adding more detail as you go along.

You see below, how the same mesh, but with different smoothing groups can give very different results.

Hard Edges

In comparison you can control the mesh with support loops, but it will take more time and will be harder to change. Try incorporating hard edges in your workflow as much as possible and you will see a jump in the productivity and efficiency of your modeling.

No Support Loops

You can have a look here at this youtube video "Double Smooth" by Ben Mathis, for something similar to the concepts above.

Tool tip:

3ds Max: Set smoothing groups to define your surface then add a MeshSmooth/TurboSmooth with Smoothing Groups ticked Or if using OpenSubdiv adjust the edges Crease values manually in the Editable Poly Edit Edges Tab. I have noticed some bugs in 3ds Max where sometimes you see a strange subdivision artifact if using smoothing groups, always double check your mesh and look at the resulting wire-frame and if you notice anything wrong you can fix by deleting the surface and rebuilding the polygons again or as a safe measure by exporting into OBJ and importing again.

Maya: Either use the Crease tool from the Mesh Tools menu and apply OpenSubdiv Subdivision type or select your edges and from the Normals menu select Harden Edge then Select Smooth Options and use Maya Catmull-Clark as Subdivision type and make sure the Hard edges check box is ticked under preserve tab.

III-Tris & NGons

As we spoke earlier, you can feel free to use any topology that works, even if it means your wireframe render is not going to win a beauty contest. General rules to follow would be to use tris or ngons in places where the surface curvature angle is minimum, or if it's a hidden area, or if you are aware of the distortion and find it to be OK.

This video does a fine job of explaining it, Subdivision Surfaces: Artifacts by Greg Petchkovski.

How to Turn Into Quad

There are automatic tools to convert tris and ngons into quads, but in here, I am referring to the manual way of doing it when you are modeling something. To be quite honest, I am not a big fan of theorizing these aspects too much, it's quite simple to start with, and will come naturally as you model. But, it's good to know few things in general and get an idea.

The steps by Pedro Amaro Santos which you can find here: Polygon Table and his small updates: Moving Poles and Shifting Grid Topology Keys, and this Polycount wiki: Subdivision Surface Modeling are quite useful.

To me, I can simply say, These are my guidelines:

1- You can take edge loops and change their flow as you like, or direct them out of sight.

2- If you have 2 tris close to each other, you can easily remove them.

3- If you have one tri only, there is no way you can remove it, unless by removing the whole edge loop or extend it to an open edge somewhere.

An easy scenario.

An easy scenario.

Moved one tri (extended the loop slightly).

Moved one tri (extended the loop slightly).

Woop.

Woop.

Let us try another way.

Let us try another way.

Looks messy, take out 2 edges and..

Looks messy, take out 2 edges and..

Move vertices and you are done!

Move vertices and you are done!

This method is longer.

This method is longer.

Voila! Quads.

Voila! Quads.

This took longer to finish.

This took longer to finish.

One more.

One more.

You can sometimes shift loops.

You can sometimes shift loops.

Same results.

Same results.

IV-Boolean Operations

Don't shy away from using booleans, it can increase the speed of your workflow, and there are some simple tricks, which can help you use the resulting mesh and get around some its limitations, Polycount Boolean Thread. Just know what you are doing and keep testing as you go along.

Tools like Mesh Fusion, which you can use in The Foundry's Modo, will save you the effort of cleaning up and give you extra options to easily control the mesh and change the design.

Another tool you might consider is actually Blender! check this Youtube video.

Of course if boolean is such an integral part of the modeling you do; you might consider moding in NURBS.

If all what you need is a small part somewhere, and you know what you are doing.

If all what you need is a small part somewhere, and you know what you are doing.

Why not make a quick boolean operation.

Why not make a quick boolean operation.

Of course you will have to deal with all of those vertices.

Of course you will have to deal with all of those vertices.

I did a boolean operation here, I took my time to make sure there aren't any major glitches.

I did a boolean operation here, I took my time to make sure there aren't any major glitches.

Sometimes, you can use boolean operations to help you make a cut.

Sometimes, you can use boolean operations to help you make a cut.

with proper cleanup, it will subdivide smoothly.

with proper cleanup, it will subdivide smoothly.

V-Crease to Blended

These transitions, can be quite difficult to master, there are many ways to achieve-the transition.

1. Space Loops Apart

The standard way works best.

The standard way works best.

You will get a gradual transition when subdividing

You will get a gradual transition when subdividing

2. Terminate a Loop

Or this one.

Or this one.

You will get a different form.

You will get a different form.

3. Hard Edges

You can use smoothing groups/hard edges

You can use smoothing groups/hard edges

Subdivide

Subdivide

You will get a mesh, that you can subdivide again without hard edges this time; to get a smooth mesh.

You will get a mesh, that you can subdivide again without hard edges this time; to get a smooth mesh.

This way can be incredibly fast to block out models, start with hard edges, subdivide while preserving the edges then subdivide again to smooth.

This way can be incredibly fast to block out models, start with hard edges, subdivide while preserving the edges then subdivide again to smooth.

4. OpenSubdiv

Use edge Crease.

Use edge Crease.

Normal crease algorithm.

Normal crease algorithm.

Chaikin crease algorithm. Both don't seem to provide an easy way for a smooth transition. Maybe later someone can come up with a better subdivision method than the ones cur.

Chaikin crease algorithm. Both don't seem to provide an easy way for a smooth transition. Maybe later someone can come up with a better subdivision method than the ones cur.

Summary

I tried to keep the information as brief and simple as I can, I didn't go into differentiating between the mesh you are working on and the subdivision you will add later, nor did I use a complicated terms and techniques. I even avoided adding details on how to turn or redirect loops, corners, poles, and any details about facial loops.

I have learned how to model intuitively with no theory at the start and I believe most can, a lot of the things you will learn by trial and error and observation; will stick to your memory much better than if given out freely. I hope no one takes the things posted here too rigorously. Just have a quick look at it. and then get back to modeling and see how it goes.