View Full Version : modeling for games

01-11-2006, 03:31 PM
I have heard that if I am modeling for games I should always use 3-sided polygons and not 4-sided. Is this true, and if so why?
I generate alot more polygons if i have to use 3-sided so it seams it should go slower..

01-11-2006, 04:03 PM
Game engines require 3 sided (triangles) polys, to enable them to be used in the engine. You can either generate the model in 'quads' and then triple them or generate them in triangles by setting your preferences...

Which methiod to use??? not to sure myself, as I'm just in the process of making a game myself....

As for number of polys (triangles).... see what the game engine can handle, but also be wary that you may have a limit on points....


01-11-2006, 05:07 PM
Every model will be made of triangles in the game engine, so when you count your polygons i LW, count triangles. Using quads won't speed things up or result in a more poly efficiant model. Generally, you can keep quads in your model if you like, at least if they are rather flat ones. Most people like to triangle the quads themself in LW, since that way they can make sure that each quad is triangulated the right way, so it will shade correctly. If you leave quads in LW, the engine will split them into triangles for you, and chances are the engine don't split them the way you would have.

01-17-2006, 07:34 AM
I personally first model using quads and then gradually optimize polygon count and flow by making triangles and welding points. Especially the polygons which are in joints that deform (ankles, knees, elbows, shoulders and such) should be triangulated manually so that the edges flow naturally...

01-26-2006, 01:01 PM
I try to keep nice edgeloops and as such quads are handy, but i am not afraid of "freezing" some quads into tris (as that gives you control over the rotation of the triangles and not a random triangulation by a game exporter), and also split a quad into 3 triangles to add detail curvature smoothness, such as around wheelarches.

04-27-2007, 07:46 AM
Depending on the engine or converters (as LW is not a standard for this stuff) it might be essential to have triple polys or the programm might fail in doing whats needed.
If in doubt asked the programmer of the engine...

04-27-2007, 07:57 AM
yea right.
it depend on engine.
if the engine can load 4verticies-poly,of course can use quad poly.

but ,at lease Direct3D(actually mesh paradime on D3DX) need triangle only.

04-27-2007, 08:13 AM
As people have said, the games engine will typically split any quads into tris before outputting them and sometimes it will split them the way you don't want. Best case examples are joints on characters, the crotch area in particular - the amount of time I've spent just bending legs backwards and forwards to check the joints don't break when bent makes me cry if I think too hard about it.

05-05-2007, 08:50 AM
I've been using Lightwave to make dts models for the Torque game engine recently. Torque requires tris, but I prefer working in quads.

I've found it useful to model in quads (then optimise to tris where necessary), subpatch the model with a desirable weight and set the subpatch level to 0 in Layout. This has a similar outcome to tripling quads in Modeler except it keeps your quads intact in your model. I'm used to working on subpatches so I build my models as cages from the start. Setting the subpatch weight to less than 100% smoothes the model within the cage. Setting the sub-patch weight to 100% keeps your edges as they are.

This technique has helped me better manage my saved model iterations (I hate saving tripled versions when I'm working and I don't like working in tris).

This technique only triples quads and doesn't multiply tris (i.e. you can optimise by including tris within a model). It doesn't allow you to control the direction the quad is split, but for most of the models I've tested in Torque this hasn't been an issue. I can triple individual quads in Modeler where I want to control the split direction of a particular quad.

One added benefit of using a subpatched model is that it's upwardly scalable: I can increase the subpatch level in Layout to get a smooth silhouette for high resolution rendering without changing my mesh, my rig animations or my UVs. I've even used this as the basis to generate quick LoDs.