General question for the 3D modeling crowd

hello guys, to make it short:

Now I've seen people do both things when connecting 2 meshes. Are there any downsides to the "just stick it in and pray"-method when it comes to rendering in a game engine, or is that an accepted way of doing things?

 

just depends on what the asset is.

Is it a hero prop that will be very visible on screen? What is the quality level of the rest of the games graphics? That sort of thing.

Neither way is right or wrong, just depends on the situation. Obviously one way is faster and easier... while the other way will look more realistic.

If you do connect them though there is no reason to leave those edgeloops running all the way around. Stich them off at the connecetion, they aren't doing anything.

Polycount is the big forum for 3d artist. Most questions like this you can find already answered or discussed by searching or reading the wiki there.

 

Have you tried both methods and examined the difference in your game? If it looks the same, it really doesn't matter what anyone thinks.

I used to never just stick objects into one another, but these days I tend to do it unless it's something that requires a bevel at the join. Generally speaking, I think the main issues with joining them are:

• Takes longer and the work isn't fun.
• Sometimes baking artifacts if the join has some stretched triangles or something like that.
• When you have a tiny thing joined to a big flat surface, you often find yourself having to spread out the transition from small -> large faces to prevent weird normal bake artifacts.

And the disadvantages of sticking stuff through eachother:

• Sometimes the part that's embedded inside another mesh causes baking artifacts, especially if it comes close to the inside surface of another face.
  
• If you separate the parts to bake e.g. normal map, you have to then put them together to bake AO which is a bother - it's often better to be able to bake iteratively without doing a lot of messing around.

It's really up to you, and it depends on what you find most annoying and what suits your workflow better.

One thing that sticking stuff through eachother works well with is a workflow that is based on a lot of modelling and less reliance on normal maps. If you look at the following image, I'm pretty sure the modeller is not going to bother joining the tiny details to the body, that would be hell right there:

 

That is some great advice, BGTIMEMASTER. Sorry for posting in the wrong forum

 

Great post too, Billy. This really clears up things for me. Thanks for taking the time

 

this isn't the wrong forum by any means, just tons more 3d artist at polycount as that is the purpose of that forum. Here is more programmers, seems like. Only a few of the regulars here are doing much 3d I think.

 

to add a little more to billy's example :

here is a model you can get for free from the substance share website. I think you would probably classify this as a hero asset that could belong in a AAA game. But it provides a good example hard surface topology and shows what kind of stuff might be built with tpology versus normal map baking, and also what topology will just be floating versus connected. As you can see, many of the grab bars are just stuck into the main bulk.

It comes with textures so you can see which details are baked into the normal map and which are built with geometry.

Just one example but the more you dissect other peoples work and experiment on your own you'll get a feel for it.

 

Nothing to do with rendering, but you have to be careful when doing this sort of thing when you make a skinned mesh, especially if you want Blender (or whatever software) to skin the mesh with automatic vertex weights. In that case, these sort of unconnected edges may be seen to come apart when animated.

 

Like mentioned here, what is visible on the model is the only thing that matters. You don't need to connect things if model will look normal regardless. Make your life easier when modeling anywhere you can.

 

Connecting things also adds to the vertex count of the model can add hundreds of verts in the case of that featured model

 

The one at the bottom will be terrible for lightmap baking, otherwise go for it.

 

I have done some optimization and render profiling work. I never had any issues at all with the meshes built like the first or second example. In my experience, doing it the lazy way did not cause any slowdown in rendering.

 

@Speckqualle
Me, I would have made a hole, at the bottom of the mallet head
and connect the vertices, of the mallet handle, to the vertices
that make up the hole, of the mallet head.
But to make that work properly, both the mallet head hole and mallet handle hole,
must have the same number of vertices. Of course, there are tricks to achieve that.