Categories:


Archives:

Hi John,

by Yorik on 19.01.2011

Hi John,

I think it pretty complicated. This is, as you say, a subject that comes and goes, there are several techniques out there but nothing really significant, architect are interested but not much, etc. I myself tried to do that several times, with different techniques, with more or less success, but always with a lot of work.

Basically I think there is one, big problem: The way to work in 3D (model and texture) to obtain a good result on the web is similar to working with games, and is totally different from the way architects work in 3D. Architects use to build 3D with specific software such as Revit, Archicad (or in the best cases Sketchup), which are high-level software, made to produce accurate 3D models, at the costs of efficiency. Those software usually use other types of geometry (CSG, BRep, etc). When you need to put your models on the web, there is only one geometry type: Mesh (all 3D web formats such as papervision, o3d, x3d or blender) use meshes. When you try to "downgrade" your high-level model to a mesh with few faces enough to be fast on the web, you get a hell of a time converting, cleaning and reducing complexity by hand. Also architects almost never use UV mapping, because most "real-life" materials use some kind of procedural repetition, so that is another task that must be done manually when converting.

And don't forget that architects usually want badly some features that are still hard to find in web 3D, such as exact sun position and shadows. You need to estimate those by using things like shadow baking in textures, but that's another extra work.

So to resume, the big difficulty would be to find an efficient and not-too-manual way to convert solid-based architecture models into clean, low-poly, uv-mapped mesh models. If you find a way to do that, then you might change the whole picture...

by Yorik on 19.01.2011