Should I merge meshes for claws and the like into the main mesh?

[Pocho_Azul]

I built an overcomplicated, high maintenance creature, but I am fond of him anyway. In the lessons, the orc has separate meshes for the horns, teeth and shackle, but I realized that I made a lot more than that between the teeth, foot claws, finger claws, armlets, bracelets, anklets, tongue, eyes and horns, he has as many around 60 bits of crap floating around him. Joining the upper and lower sets of teeth into two sets cuts it down a good bit, since they will move rigidly with the rest of the mouth, and I guess I could also join all the toe claws into a single mesh, since I am sure we won’t be rigging toes. But what about the rest?

Would there be major performance drawbacks to having all of this junk floating around as separate meshes? This is a hypothetical question, since this model is not headed towards a particular project. What would be the consequences (besides making or saving extra work) of:

1. Leaving the fingers, and so on, as independent meshes and simply parenting them to the appropriate bones
2. Joining them to the lo-poly body mesh, but not fusing the meshes
3. Joining them to the lo-poly body mesh, and fusing them into a single, completely manifold mesh

This is my lo-poly guy (with some temporary shaders applied)

He is loosely inspired by this painting:

[FedPete]

Most of the questions can be answered by defining the usage of your model.

How are you using this model and what does it need.

It is your first model and you want to do it exactly right the first time … but it is a learning curve.
You are asking good questions, but the answer is difficult and depends on many things.

In the commercial world, there are many models of the same figure for different purposes. It’s not one model to work with. Okay, there will be a master model or concept in the 2D art of real clay. And even only in 3D. But how it’s made and used, depends totally on the end goal.

That knowledge can only be learned by experience and a lot of projects.

So have fun, be creative. Don’t think too much about efficiency and poly counts.

[NP5]

FedPete covers it really. It depends! lol

Blender can make things for so many later uses that there can not be one right answer, for almost any process.

For example, this idea of low poly is ‘good’. No it is not! It is a compromise to work in certain uses. Baking is fudging the real detail. Details using normal maps are cheats on real modeled details. If your end use is a 3d printed model it can only print REAL geometry and if it is at all low poly it will print the facets out even if the visualisation tricks on the computer like smooth shading lead you to believe it is smoother than it really is.

Oh and the worried looking Orc Blue Smurf thing, looks great!

[Pocho_Azul]

@NP5

Oh and the worried looking Orc Blue Smurf thing, looks great!

Thanks! He does look nervous. He kind of reminds me of Ernest Borgnine, for some reason.

@FedPete

So have fun, be creative. Don’t think too much about efficiency and poly counts.

For this model, sure, that’s why I was comfortable with going overboard with extra details like fingers and and so on. But, I would like to know for future projects where it might matter more. I am almost certainly going with option 2, since it will require less mucking about with the main high-poly mesh than three. So that’s why I described this as a hypothetical question.

I am interested, not so much for this character as for future ones. This character is just practice, but I am still interested in the answers. I know the requirements vary depending on the playback hardware, game engine, and even on different rendering pipelines within the same game engine. Am I mistaken to think that, in most environments, every separate object (option 1) places a multiplicative burden on game engine rendering systems? My understanding is that, in some rendering pipelines, number of individual objects x number or textures x number of lights x number of rendering passes = total draw calls. So adding one additional doodad to a model may result in many dozens ( or more) of additional tasks per frame. What I have no info about is things like: whether multiple mesh islands in the same object matter, or if having hidden holes in a mesh matter at all.

[Mark_Jackson]

Things like clothes, horns, teeth, claws and trinkets can all stay separate meshes.

What I tend to do when rigging is to join all the meshes together temporarily and then use automatic weights to get the base weights on the mesh. This way the algorithm blender uses to apply automatic weights takes into account surrounding vertices of the same meshes and I can get nicer weights from the first moment so that each part moves equally with the bones.

Later I separate each part again. Once they are all rigged to the same Armature (skeleton), this will import fine as one “object” with separate components into an engine like unreal/unity.

There’s no performance hit on doing this and allows for greater control in a game engine.

If the fingers are separate, I would personally make sure that these are modelled as part of the palm as it helps to avoid hard shading artifacts on parts that should share the texture space. Claws and nails can stay separate in my opinion.

[Pocho_Azul]

Thanks for your response!

@Mark_Jackson

If the fingers are separate

All of the main body parts are in a single, manifold mesh, it is just claws on the fingers and toes, teeth, tongue and the like that are separate objects, at the moment.
I did discover one semi-major drawback to having composite, non-manifold meshes. The rounded pegs on his jewelry were separate meshes from the main band, originally. This interfered rather badly with getting a clean normal bake, so they are all fused into single meshes now.

@Mark_Jackson

Later I separate each part again. Once they are all rigged to the same Armature (skeleton), this will import fine as one “object” with separate components into an engine like unreal/unity.

There’s no performance hit on doing this and allows for greater control in a game engine.

Are you really saying that there is no performance hit for adding additional game objects to a scene, or am I misunderstanding?

My prior research and experience is that, all other things being equal, that two 500 triangle mesh/gameobjects has a significantly greater impact than a single 1k triangle mesh, just as four 512x512 textures makes a bigger hit than a single 1024x 1024 texture, even though the number of triangles/pixels are the same. In a simple project, a few additional objects may make little detectable difference, but multiplied many different prefabs, shaders, lights, etc. my impression is that this can add up. Am I mistaken? I am not worried about this specific project, I just mean in general.

[Mark_Jackson]

Well in terms of polygons, if they’re the same, this won’t make much of a difference. They’re loaded in and then they’re rendered.

For textures, two or three or four mesh objects can share the same texture map if the UV’s are made to fit. This is optimized and won’t make much of a difference. So you could have the arms and legs mapped onto the same UV space but then separated into different mesh objects but using the same material and texture maps.

[Jack_Mak_Morn]

That’s a really cool idea to use a japanese Oni or Ogre for reference! The result speaks for itself