Some generic mapping Qs...

[SMASH]

Ok, so a few questions:

1. If I model a level in MAX and import it as a model how do you do custom collision? I've tried before and couldn't get it to work.

2. Do I put a skybox in the model or is that later?

3. Does anyone know a good custom map tutorial to follow? Or mb someone who modeled a level in MAX can make one...

[Zeph]

I haven't gone too far into doing things in Unreal for gameplay, but if I wanted something to have custom collision I'd make the custom geometry along with my model but UV map it off to a small unused part. I'd use an alpha map to give that full transparency. It's crude, but it'll work.

[l SINERGY l]

This might be what you're looking for:

http://hourences.com/book/tutorialsue3modeling1.htm

[FireScythe]

The UDN has info on making your own collision models here : http://udn.epicgames.com/Three/CollisionReference.html

Although importing a whole map as one object kind of goes against the workflow of the engine. You'd be better off using it as a reference.

The skybox is actual full size geometry in UT3, take a look at the official maps for some ideas :D.

[Phopojijo]

Making the full level in Max isn't a problem.

The thing is you need to split it up into chunks and export them separately and place them separately. You do not want to just export the whole map as one visible section.

The whole idea behind deferred rendering is that you can only render what you can see. If your entire model is a single mesh -- you can see your entire model all at once. This is not good... especially when you consider that you're ALSO getting penalized for dynamic lights touching your mesh.

Say someone in red base is firing a flak cannon. If you are in bluebase and your ENTIRE map is ONE SOLID MESH -- you will get a performance drop whenever the game tries to calculate lighting all the way across the map.

So:

Advantages for separating your map into small meshes, placed in unrealed:

1) Multiple instances of the same mesh will be cloned in memory. (Say you have 40 lightposts... only 1 will be in memory)

2) If the entire mesh is invisible to the camera -- it doesn't exist performance wise (pretty much)

3) If a dynamic light doesn't touch the mesh you see... it pretty-much doesn't waste any performance.

Disadvantages to leaving as one contiguous mesh:

1) More prone to import errors.

2) There's no instancing or occlusion.

3) Things that happen completely out of sight can hurt performance elsewhere.

As for collision?

You can I believe either import custom geometry, use BSP, switch to per-poly collision, or simply turn collision off (if it's something irrelevant)