Modeling discussion?

Chernikov

Spaceman
I was curious if some of the more experienced modelers were open to discuss technique and method... less of a specific-program or a specific-problem line and more best practices and a professional approach to the art.

My leading question(s): when modeling complex objects, and with an eye to implementation in game engines, is it considered acceptable to use separate intersecting meshes (such as an engine mounted in a wing) or is it best to keep everything in one mesh? If a single-mesh approach is better, do you find it more practical to model separate parts and cut/weld vertices as necessary to join them, or to start with your core object and grow/extrude as needed? And with modern effects such as bump mapping is it more effective to create a very simple mesh and effect (affect) as necessary, or work from a more complex model and refine it down as the mesh develops?

Or are meshes rather outdated and I need to stop living in the 90s?
 
Its not meshes that are outdated (kind of like saying wheels are outdated on a car) but worrying about intersecting objects etc. etc. is redundant in many modern game engines. The amount of extra processing power needed to calculate those intersections is rather small in scale compared to all the advanced shaders and post-processing effects in play now. Back in PS2 days (where I cut my teeth) it was more important.

That being said, there is always good practice in keeping things clean. One nice tidy mesh is always going to be easier to troubleshoot than a whole mess of things. When you are working with thousands (or millions) of tris you don't want to be looking for that one backward-facing polygon thats making your engine crash by hand...

Also, keep in mind that if you are using open-source engines, they are often built on a whole stack of legacy engine code that may complain about the odd thing.

In regards to normal mapping, many of the latest modelling programs (zbrush, sculptris) have in-built tools to change the detail level of your model as you go. I'm going to fire a shot in the dark and guess that since you are posting on a Wing Commander forum that you are interested in creating some kind of Wing Commander mod, involving spacecraft, in which case these programs will be of little help. I would suggest building your craft as detailed as you like, generate a normal map from that, then cut down to fit the game engine. If you are planning on using any next-gen game engine, this may be less cutting-down than you think!

Remember craft in-game only fly by as close as the camera gets. You don't need to have mass amounts of detail on the inside of an engine intake or individual screws modelled around a cockpit gauge! A little trick I use is to turn on wireframe mode, highlight all the verts of your model and zoom out a little. If you have many areas where collections of verts are densely concentrated, you may want to trim the detail down a little there.

Thoughts, people?

PS - My Three rules - Don't intersect curved surfaces, build in quads where possible, and never EVER use any tool like 'Smooth Surfaces' or 'Meshsmooth' (unless you REALLY REALLY want to).
 
Chernikov, you're on the right track. The important thing to remember is that modeling is a toolbox, and whatever trick from the toolbox works for your particular situation tends to be the right one.

It usually depends on the game engine, but lots of static meshes use intersecting entities, it's only when you're dealing with skinned objects (humans, animals, etc) that work on a skeleton rig that it's important to have everything on the same mesh. Even then you can cheat a little.

Bump mapping is an effective technique in many circumstances, but not all. On my models, for instance, I will tend to model deeply cut areas like intakes and gun ports, but I save time and energy by including rivets and panel lines into the bump map.

Getting into program specifics, Modo lets you make cuts on top of a textured model without screwing up the UV mapping, so if you have a textured detail that you think would look better as a modeled detail, you can make it very easily by cutting the model using the texture as a guide, it's very handy.

There are a ton of great programs out there more affordable than 3d max, and let you learn essentially the same techniques.

I reccommend Modo, which has a very good free trial period, and tons of free online tutorials in movie form, which are especially helpful.

Also Houdini, which has a free, nearly fully functional version especially for learning, or a $99 version for what they call 'apprenticeship' which gives you a completely functional version as long as you only use it for non-commercial applications (i.e. you can't make money off your work with that version)

I haven't played wit Blender much, but it is also free, but quite a bit different from the "standard" format that a lot of 3D programs use. But it is quite capable and has a wide support base in the form of a very active online community.

Hope this helps, *I* don't think you're living in the 90's!
 
It's less that I have a specific project in mind, and more that I'm dipping my fingers into modeling again and trying to do so seriously this time through. My intent is to avoid bad habits and create with an eye to a usable finished product on all projects, which requires knowledge and experience that isn't found in user manuals or in my past. Thanks to both of you.
 
Meshes are still standard fare. Even if they end up only being the base for sub-division objects, they're not going away. It's the plaster of virtual sculpting.

For me it depends on what my desired level of detail and poly budget is as to how I model extra detail ('extruded from' vs 'new mesh').

I'll paraphrase a conversation I had with another modeler. You have a space ship and you want to actually model the armored plates so the crease lines aren't just a texture effect. You can 'draw' the plate lines with new edges on the existing mesh, inset each new polygon, and extrude the plates out. Or you can create a single new, separate face for each plate and extrude that face.

The advantage of the former method comes when you want to subdivide the model and need interaction between the meshes (the plate lines being a rounded gutter instead of sharp angles). The advantage of the latter is that you have a much smaller final poly/vert count.

Having individual meshes also isn't a problem. In 3DS Max for example you can attach meshes together so it becomes a single mesh, even though it's made up of many unconnected elements. It can then use a single UV map and texture.

So... it's a question of what you need as to which is going to give you more.
 
There are a ton of great programs out there more affordable than 3d max, and let you learn essentially the same techniques.

I reccommend Modo, which has a very good free trial period, and tons of free online tutorials in movie form, which are especially helpful.

One of my long-term goals is actually to save up for Modo. I have/use Max and Maya, and the stuff I see Modo capable of blows my mind.
 
Back
Top