Hey everyone, this is my first time posting here. I can see there are a lot of pros here so I'm excited about the responses I will receive.
I am currently a student going for a BA in Game Art & Animation, focusing primarily on hard surface/environment modeling.
I have a question that has plagued me for a while now. Through my modeling experience I have found that extracting some faces of a mesh as individual objects and/or combining separate meshes tends to save tons of triangles. The best example of this would be a cylinder floating above a square face, combined, but not physically connected.
My question is: is there anything negative that could come out of this technique, specifically for hard surface models? Does it cause any rendering issues or any visual glitches at far distances, or create collision anomalies (despite edges/vertices being perfectly lined up with each other)? Are lighting, normals, or other texture maps affected by a technique like this?
I figure I should ask now while I am still early in development so I don't "grow up" doing things completely wrong.
If this is indeed an acceptable way of modeling, I have one more question: would it be better to have edges of floating geometry flush with the other mesh it rests against or have the edges intersecting slightly?
Thanks everyone!
Replies
When it comes to modeling lowpolys, sure you can save a few tris, but generally there are drawbacks such as:
1. You will waste more uvspace, i've seen some really excessive models created with like 40 different chunks, and you'll often see like 50% of the uvspace is just wasted with black AO on it because it intersects.
2. Every time you split an object, you're creating a hard edged seam, and you will need anti-aliasing to soften it out. If your mesh is mostly solid and seamless, apart for areas that need to be animated or areas that need to be exploded to avoid raycasting problems, you can get a nice seamless result, as the baked normal map can cover up any uv seams.
So you waste texture space, it looks worse, and its generally a pain in the ass to deal with(bake). You can save a few tris, but really, its not worth it. With current hardware, getting more texture resolution vs saving a small amount of geometry is always going to be king.
Here's a screen shot specifically what I'm asking about. I used the edge loops to control how the textures were displayed, but decided to extract the lower portion from the top instead tying off the extra edge loops as triangles. Is this bad form?
Depending on what your planning on doing with the piece I would add the triangles to the top conjoining poly so there are no holes (preserve your uvs and cut then weld.)
If you don't you will probably run into sparkling artifacts when the player, or camera is moving (it will probably look fine in a still shot).
http://www.chrisalbeluhn.com/Vertex_Welding_Tutorial.html
give this a read through, its for max, but the explanation is quite helpful.
(I'm no expert, and if I'm incorrect, by all means someone step in and correct me)
you'd be much better of modelling in some detail with those verts
are you relying on baked vertex lighting or something?
If floating geo is not an issue with the particular engine you're using, under certain circumstances it can be more efficient and cleaner to float the geo rather than weld it into the mesh. Using floating cards for decals is a good example.
I can't see a reason why you would need to float the geo in your example though.
This explains a lot of my original question, thank you!
DOG-GY, r_fletch_r, & Harbringer,
This specific piece is a modular extension to a ledge set. I bent the under side out to show the whole model better. The part where it looks like it has ngons is where the two separated faces meet. I extracted the bottom half to keep the edge loops from wastefully continuing to the bottom, as I only needed the edges on the top to control how the bricks tiled.
My question doesn't pertain to this piece specifically but more for general practice. Is it bad/wrong to have an object constructed with multiple disconnected faces or should I try to stick with a contiguous mesh even if it means a higher poly count?
Thanks everyone.
Check out the 'Occlusion' part of this write up for some other info with gaps and holes. (something else to keep in mind specifically for Unreal)
http://www.hourences.com/tutorials-ue3-modeling-2-advanced/