PDA

View Full Version : How do you learn "Correct" Modelling?



Dahfish
10-12-2004, 10:56 AM
I have been a regular reader of these forums since starting to teach myself Lightwave a few months ago, although I am very much a novice stumbling along in the dark. I find inspiration in your work and it keeps my determination to improve alive.

I have gone through many tutorials from various sites, however I have been unable to find a resource teaching the do's and don'ts of correct modelling.

There are many methods for stitching points into polygons (see example), but how do you know when something is right or wrong?

Are there such things as bad habits? And when does it matter?
Do different companies / industries (e.g. games) have specific guidelines?
Or is it a case of if it looks good, anything goes?

How do you REALLY learn?

I have attached some files to show as an example of what I mean.

The zip file contains the model and scene files of an electrical household socket (UK) that I am using in a model of my house that I've been working on.

The area of concern is the front square face area (X,Y no Z depth). The top row has this section unattached, resting in place. The bottom row I have MERGED the points so it is attached. As you can see there are texture problems (procedural) only on the merged versions.

The Left and Right sockets show 2 different methods used to create the front face: Boolean Cuts and point by point polygon creation 'p'.

Linking back to my overall problem: How do you "correctly" model such an item?

If you need any information just ask and thank you for your time.

Dah.

Thomas M.
10-12-2004, 11:06 AM
I don't think there isn't anything like right or wrong when modeling something. It simply works or it doesn't. As we are working with a highly flexible medium, there is always more than one way to reach your goal.

The problem which you discoverd is something LW doesn't like to much. The easiest way to avoid these smoothing errors is to bevel to opening for the plugs twice on the same layer as the front panel before beveling inwards. LW is really about learning by doing. If you need to animate your object than there are for sure some things you should avoid, but otherwise it's hard to come up with a simple formular for modeling. It is like good design: It's hard to describe how it should look like, but you recognize it instantly if you see it.

Cheers
Thomas

UnCommonGrafx
10-12-2004, 11:20 AM
It would seem you've learned how to already. ;) Now, you just have to learn how to accept what works and what doesn't.

I would use the top one, learning that the bottom one has 'issues' when points are merged. Lesson learned: sometimes you don't have to or want to merge points as that will give you rendering errors.

Sounds trivial but it really isn't. Having been down the same road of "but my other app..." I can only suggest that you learn to "Go with what works!"

jin choung
10-12-2004, 02:49 PM
this is something that i probably asked a lot early on:

GUIDELINES:

1. most modeling for film/tv come from a hollywood aesthetic... it is all and ONLY about FACADES.... the amount of verissimilitude need only be good enough to sell the shot... no less of course... and NO MORE. cuz more is time and time is money.

2. lots of derivative guidelines from 1. so if you are never gonna see the other half, don't build it. if you only see one room in the mansion, for heaven's sake, don't build the 99 other rooms that 'really should be there'.

3. you ignore rule 1 completely if you are modeling for pleasure, design, or games.... lots of times, here the rules are not at all the same as tv/film but you get other rules instead.

many times, if you are modeling for manufacture, you must build 'watertight' shells... every mesh must be a 'solid' that encloses a volume without 'holes'.

in games, you are usually prescribed poly count limits. 'make it look frickin' awesome in 200 polys or less'.

4. it is absolutely OK to just push unconnected shells into each other to build a model... there's no need to build 'water tight solids' in every situation, no need to merge all the verts of one part into the verts of the part it's sitting on top of.

do whatever is fastest and easiest and don't do what is not unless you're gonna see the results.

5. do not create and work with geometry you did not intend to - be aware of and remove JUNK POLYS:
a. frequently, 1 point polys are useless and unintentional. delete them.
b. 2 point polys are almost always junk too. if it's not, you'll know it.
c. two polys that share the same verts (and they're right on top of each other with welded verts) is junk too.

learn how to diagnose your mesh using 'w' statistics and do proper cleanup. junk polys will create lots of problems from conversion, to visual artifacts, to errors if used in a realtime game engine.

come on guys, there's lots more common sense rules like this. let's list off a bunch more for the lad....

jin

jin choung
10-12-2004, 03:00 PM
aw hell,

here's another one....

MICROBEVELS: nothing in the real world has razor sharp edges. put in 'microbevels' on edges that will be prominent in your model to make it look that much more real. you can safely ignore this on smaller or insignificant details but it's a good idea for most edges to be nice and beveled.

the beveling ends up catching specular highlights in renders and it REALLY makes a difference in how your model 'registers' to a viewer.

and true, if you 'microbevel' a 90 edge, you are just replacing it with two 45 edges that are razor sharp too.... but this is indeed a fudge and a single microbevel will suffice in most situations.

jin

bobakabob
10-12-2004, 05:10 PM
Hey, Jin, excellent advice there and nicely summarised - so many modelling tut books fail to mention these fundamentals.:cool:

Dahfish, if you want to learn organic modelling there are some great threads not only on this site, but on cgtalk exploring topology and how to aim for economy in your style. Check this one out...

http://www.cgtalk.com/forumdisplay.php?f=25

Cheers,

Bob

mattclary
10-13-2004, 05:29 AM
I think it was Larry Shultz (Splinegod), who once said something like this: Modeling is just a problem that you use your tools available to solve.

That REALLY hit home with me, as I hadn't even realized that is EXACTLY how my mind works when I model. When I want to make something, I try what I think will work, and if it doesn't I try something new. It almost becomes a form of troubleshooting for me.

You are on the right track, you can't start truly thinking for yourself until you have some practice under your belt, so tutorials are the right way to go. As you learn the toolset, you will have ideas on how you can better use them.

No matter how long you do it, you will ALWAYS learn something from the community that you had no clue about. The uber-coolness of LightWave is reflected in it's user-base, make use of the community.

Follow what jin said. Number 5 should have been number 1. Someone else mentioned smooth shift with no displacement; I'm not a big fan of that because when you merge points, it no longer works. If merging points breaks your geometry, your geometry needs to change (IMO).

re your model: I think what you have will work, you just need to adjust your smoothing angle. If not, one way to improve what you have done (presumably a drill or boolean), is to give it more geometry to work with. Instead of using one flat poly, subdivide it. Having the extra geometry allows the cut to better "define itself". Once again, that's how I prefer to work. There is no ONE right way to do anything, pretty much.

My illustration is off a bit. You want as many lines going through the area to be cut as possible.

Dahfish
10-13-2004, 05:30 AM
Thank you all,

Some pearls of wisdom there and certainly the advice I was looking for. I feel a lot more confident that I haven't taught myself any particularly bad habits and that I am probably not the only one with these questions :)

I shall post my house modelling and progress that I've been using to learn and experiment with under the W.I.P. forum soon for some critique / advice.

Take Care

Dah.

Silkrooster
10-13-2004, 06:18 PM
I have to agree with Mattclary on what he said about your object. I have read in a book that when you model an object a polygon can only have so many points, that is the limit in Lightwave. When you use the boolean tools, you are more than likely going to exceed that limit. Therefore it is a good idea to first divide that poly once or more so you don't exceed the limit. That I think is one of the main reasons why booleans don't work for newbies.
Silk

John Fornasar
10-13-2004, 10:46 PM
A few things here... as jin said, think Hollywood set - if it isn't going to be seen, don't model it.

However, when you are learning, it feels good to be able to do complete objects, so here are a few thoughts...

in the top left switch housing, you have 559 polys, 200 of them just for the screws, which are barely visible. that is a hell of a lot of polys (unless you are doing a product rollout with extreme closeups.

I'm going to disagree with everyone on the booleans though.. they do work, you just have to follow some rules. (or better yet, I won't disagree, just say that they work for me).
Every object must be solid, both the object layer and the cutters.
When doing more than one cut in a solid, do all the cuts at once. (this goes for lightswitches or windows in a building.) LW will produce "clean" lines for you (hopefully the .jpg uploaded). I started with a solid box, booleaned out the inside, then cut all the openings in one shot.
If I needed to "remove" the front for a product shot, I would take the resulting box, copy it to another layer, and cut the front from one layer and the back from another.
Here, I "assumed" the insides would not be seen, so I deleted the "excess" polys.

colkai
10-14-2004, 03:35 AM
As an alternative to boolean, try using stencil and then bevelling / smooth shifting the result.
This way, I can provide the microbevels and at the same time, remove the smoothing errors. The first microbevel inset will separate the stencil from the main surface, so it doesn't try to smooth the object. The last pic was done without microbevels.
Thus...

Matt
10-14-2004, 07:04 AM
The other way to avoid smoothing errors without having to microbevel is to select the polygons aound the holes, cut and paste, by doing so you are 'disconnecting' the vertices. Not the best way but quick and dirty if you're in a hurry!

Also a word on modelling. Plan, plan, plan! (Ok, three words!) You can make a modelling task go so much smoother if you sit down and plan (use a pen and paper) how you are going to model it. You might change that plan once you get into the details, but it's still a good idea!

SplineGod
10-16-2004, 03:30 AM
An alternative to beveling or creating microbevels, get this plugin :)
TBs Edge Bevel Shader - http://home.att.ne.jp/omega/tabo/3dlabo/p_lwp.html

John Fornasar
10-16-2004, 10:59 PM
TBs Edge Bevel Shader

Nice one... useful and fast. Thanks for the tip, Larry

MooseDog
10-17-2004, 09:31 AM
there's no need to build 'water tight solids' in every situation

excellent excellent advice and insight. one us learners have a tough time getting over conceptually. the final image is what you model to as jin said, not the "perfect" model. in fact the perfect model is the one that produces the image you need, and working towards solving that is the right way up the learning curve. that's my expereince at least.