Understanding subdivision modeling primer paper

danjohncox

Hey guys, I'm a college instructor teaching folks about environmental modeling. One of the typically confusing subjects ends up being subdiv modeling. I'm putting together a subdiv paper, to help students and any newbie understand maybe, some of the basic rules they should pay attention to. This is my initial stab at it. Please give me your thoughts, I'd like to make sure this is robust and helpful for many people. Feel free to suggest example images too!

Unfortunately I can only show it properly as a google doc. Lemme know if that works for you guys.

https://docs.google.com/document/pub?id=155E77C5nJzkppXyGOQ3PoZAkLQf43QCyUJZa4_kCANM

Xoliul

Seems decent, but only touching the very basics. There's more difficult things to SubD, like understanding the flow of a surface and translating that to a wireframe. I wrote a document on that, I'll see if i can share it when i get home. It's mostly aimed at cars since these are the most difficult situations but it'll help you with everything if you start to understand it.

cryrid

There might be some useful info found at http://wiki.polycount.com/SubdivisionSurfaceModeling?highlight=%28%5CbCategoryEnvironmentModeling%5Cb%29 as well

EarthQuake

One thing I would suggest, is to use a more generic term like "sub-division" instead of "turbosmooth" as that is just a max specific modifier(and even then, only one of multiple methods in max to do sub-d).

EarthQuake

The simplest way to look at turbosmooth modeling is to understand that; the fewer edges you have, the more curved your model will be; the more edges you have, the more hard lines you'll get in your model.

This could be explained a lot better, often times to keep a round shape with details, you need more edges. So less edges = smoother isn't really something I would say is a generic rule to follow.

I would also suggest skimming through this thread: http://www.polycount.com/forum/showthread.php?t=56014 and picking out a few of the more common examples, and do a little write up on those.

passerby

less edges just mean your current edge loop positions are averaged out more, which doesn't always translate into smoother, example would be subdividing a cube yes it averages out to a sphere but far from smooth due to lack of Geometry. IF you got a surface you want to carry a slow smooth curve on your best of distributing edge loops evenly across it to carry the curve, and give some geometry for it to divided.

Michael Knubben

As another example of few edges not necesarily producing a smooth mesh:

In this example one edge on a cube was bevelled, and then the top resulting edge was moved down and forward, resulting in this sharp edge.

vargatom

The simplest way to look at turbosmooth modeling is to understand that; the fewer edges you have, the more curved your model will be; the more edges you have, the more hard lines you'll get in your model.

That is not true. We have a hard surface armor model that's full of curves and it's like 350,000 polygons.

If you want precise control over curvature, soft and hard edges and transitions, UVs, etc. then you absolutely require a detailed model. You can however build this detailed model by starting with a low resolution sketch to lay out proportions, then subdivide it once or twice and add details only after that.

Just look at the "FAQ thread on how U model them shapes" to see the single most common answer to most modeling problems regarding details on curved surfaces:
MORE GEO.

CheeseOnToast

You should probably mention poles and their effect on edgeflow as well.

danjohncox

Wow guys, thanks so much for the feedback and ah ha, yeah you guys are totally right about the "less polys does not equal better curves comment" That rule was poorly phrased.

Xoliul: yeah this is definately simplistic. This part is meant for people who have NEVER properly used subdiv modeling and dont know what it does. My main goal was to help explain how subdiv modeling works. I find there isn't a ton of documentation to help people just starting to learn about it. I'm likely wrong I'm sure, but I simply couldn't find much that explained it the way I'd like. But please toss up your document I'd love to read it over.

cryrid: That page used to have TONS of great stuff it in. But many of those really great links dont work anymore

EarthQuake: Thanks for commenting man. Your normal map, waviness thread actually totally changed the way I teach normal mapping now. Much simpler and less complex with crazy rules. And yeah, I can switch it to Sub-div. You're right about that rule though, its terrible phrasing and misses the mark on what I'm trying to say. Basically that, bringing edges closer together, or rather, having uneven distribution of edges can cause a shape to look more hard edged. However I'm clearly having trouble explaining that in a way that is easy to understand.

Vargatom, MightyPea, passerby: you're very correct, problem is I'm not actually meaning to suggest otherwise. Clearly the wording I used went over my head and the others I had read this over before hand and what I actually wrote is too broad poorely phrased.

CheeseOnToast: I mention poles near the bottom, but i used the term stars. I've fixed that now by including both terms.

Clearly the biggest confusing point is the less edges, more curves, more edges more hardness statement. Thanks much guys, I didn't even notice how totally inaccurate that actually is. I'm simply trying to point out how a person can get an edge that looks more creased compared to most curved surfaces. And generally speaking, adding more edges close together will cause creasing if they're unevenly distributed. My preference is to focus on keeping geo simple. Then getting into the many, MANY exceptions. Would that work?