Modeling a complex building

Ville Pakarinen

I'm about to start modeling this building for Crysis:
http://www.mybesthelsinki.com/HelsinkiArchitecture/Stockmann_Argos_building_Helsinki.jpg

It's a corner of a department store located in Helsinki Finland. I have never low-poly-modeled anything that detailed. That's why I'm not completely sure how to aproach this. I'd like to hear what you guys think

Talbot

I'm not an expert but I'll try and help. :P

In many buildings, including this one, there is a lot of repeating geometry. So what you need to do is take advantage of it. So model every style of windows and duplicate them a according to your reference image.

I don't think I explained that too well so maybe someone else can join add to what I said. :P

Here is an example of a similar model.
http://boards.polycount.net/showthread.php?t=52743

I hope this helps at least a little.

Ville Pakarinen

What an awesome thread! Nice wireframe pics.

But isn't duplicating bad for processor? I mean, the more seperate objects the more rendercalls processor has to do. I have understood that it is good to model thing in large chunks as much as possible. And duplicating increases polygons.

http://www.hourences.com/book/tutorialsmeshrendering.htm

renderhjs

create a box,- and thats how you can start

seriously:
- try to find more reference pictures, maybe blueprints resources are everything.
- check out cgtextures.com if you dont know it already,- looks through some of the textures you could use
- find similar buildings with more details like single objects,- single parts that are better visible

then you could:
- create a smal sketch that catches what you want to archive,- what you want to include

starting in 3d:
- I like box, plane modeling alot- you just start with a plane of bos (6 faces) and extruce,- cut extend ec.t from it.
Its always good to seperate into multiple objects instead of 1 giant mesh with ugly all over the place devided loops.
Keep the models at the beginnig very plain to better fix proportions,- adding details later is no problem.

Jeremy Lindstrom

http://boards.polycount.net/showthread.php?t=53002

Ville Pakarinen

Here's an issue I've been pondering:



I cut window holes in two different kinds of walls using boolean. The wall in example A has been divided so that every window has its own polygon. This gives nice looking topology. In example B six windows are cut off of one continuous wall. This gives messy topology but fever faces.

People seem to use the version A more often. But why? Isn't fever polygons better than good topology? After all buildings are totally static, they don't need to deform at all.

Eric Chadwick

Two reasons:
1. Long thin triangles create rendering problems, they should be avoided at all costs.
2. Building textures are usually packed into a texture sheet, so the sections in A are needed to section off the UVs. See Chris' tut here.
http://chrisholden.net/tutor/max_tiling04.jpg

More tuts and info here, including Chris' full tutorial...
http://wiki.polycount.net/CategoryEnvironment
(or see Dekard's link above, same content basically)

Vailias

Duplicating isn't bad for the processor nor does it always increase the number of draw calls if the rendering pre-process is well done. You can draw a very large number of instanced pieces of geometry with a single draw call. Even still what is expensive GPU wise is state changes, loading and unloading of things, so if the majority of the scene you have can used cached vertex buffers and textures it can draw very quickly. You run into problems with having many unique objects and unique textures that can overflow the video ram and require loads from disk.

Also fewer polygons aren't always better, as mentioned by eric but also this thread.
http://boards.polycount.net/showthread.php?t=50588
batch batch batch.