Weapon Prop Modeling: Continuous VS. Pipeshoving??

Amank679

Hi all,

I'm in the process of modeling my first weapon, a custom gun that exists within the Fallout universe. I also noticed that guns can be continuously modeled, meaning most if not all of the body of the gun can be one mesh. I'm confused by this because that adds several extra polygons in places where parts could easily be pipeshoved.

While this is not going to be part of a larger environment and just a marmoset showcase, does it make a huge difference if key parts of the gun are pipeshoved instead of continuously modeled? 
I've also read this may cause baking issues where the assets are pipeshoved, but that also confuses me.

Below are images of the WIP gun for reference.

Thanks in advance!

poopipe

Doesn't make much difference in terms of what happens at the gpu. 

Shoving bits of geo into others can waste texture space and add a slight cost to rendering if you don't trim them off neatly  but equally if you  insist on making everything watertight you can end up using crap loads more geometry than you need to.
You have to balance the cost and arseache against the quality. I definitely wouldn't merge a bolt head into a mesh but if something would be part of the same surface in real life then I would. 

Use your judgement and if called up on it during an interview, explain your reasons - what I usually look for is a thought process rather than a correct answer. 

Amank679

Doesn't make much difference in terms of what happens at the gpu. 

Shoving bits of geo into others can waste texture space and add a slight cost to rendering if you don't trim them off neatly  but equally if you  insist on making everything watertight you can end up using crap loads more geometry than you need to.
You have to balance the cost and arseache against the quality. I definitely wouldn't merge a bolt head into a mesh but if something would be part of the same surface in real life then I would. 

Use your judgement and if called up on it during an interview, explain your reasons - what I usually look for is a thought process rather than a correct answer. 

Thanks for the fast reply, that makes alot more sense!

EarthQuake

Another concern is aliasing. If your low poly is built up from 50 intersecting meshes, you're going to have gross hard edges all over the place. If you merge the main forms into watertight meshes, you can get a more seamless, natural looking result. These days, GPUs can push a lot of polys/verts, so the method where you jam a bunch of meshes together to save a couple hundred triangles isn't very relevant anymore. This was more of a concern 10-20 years ago. The way the mesh above is constructed screams "vintage game art". These days, upwards of 20,000 triangles for a dedicated first person view weapon model would not be out of line.

As mentioned, too many intersecting meshes can waste a lot of UV space as well. Typically this means that for a comparable asset, the one which has too many intersections will have worse texture resolution. In your screenshot you've got some areas where you probably have 40% wasted UV space between the intersections. Generally speaking, video memory is much more precious than geometry drawing speed, and textures take up more VRAM than polygons, so wasted UV space should be avoided.

Personally, I try to merge elements unless:

1. The mesh chunk needs to animate or otherwise be removable, ie: customizable weapon add-ons
2. They create complex overlaps that will cause problems when baking, here it's best to keep the mesh chunks separate
3. It would add an excessive amount of extra vertexes to merge them, something like a series of bolt heads is a good example

EarthQuake

This is a fairly old asset, but you can see the general principal here. The main body of the gun is one wateright mesh, with only the accessories and animating areas being separate elements.

The main thing I would do differently these days is that more of the details would be modeled out rather than represented in the normal map. Like the safety selector switch for instance.

Another general tip: note how that areas closest to the camera, like the iron sights, have extra detail, and the areas behind the camera, like the buttstock, have minimal detail.

Amank679

EarthQuake said:

This is a fairly old asset, but you can see the general principal here. The main body of the gun is one wateright mesh, with only the accessories and animating areas being separate elements.

The main thing I would do differently these days is that more of the details would be modeled out rather than represented in the normal map. Like the safety selector switch for instance.

Another general tip: note how that areas closest to the camera, like the iron sights, have extra detail, and the areas behind the camera, like the buttstock, have minimal detail.

Thank you so much for the detailed response! Really clarifies a lot of burning questions I had! I'll keep these in mind!