Download

3D Modelling aproach: Subdivision vs manual modelling

Hello everyone,

I’ve been modelling for a year, and my usual approach for modelling game meshes would be to make a high poly mesh by making a low poly base mesh, with all details that I want, then subdivide it couple of times, and add support edges to retain shape and volume. From the high poly make a low poly mesh manually and uv it and bake textures.

But for a while a friend of mine has been telling me that this approach is for noobs. The industry professionals go and manually create all the shapes and make it high poly details without subidivision/turbosmooth (3ds max) and get very awesome results and usually a lot better than subdivision.

I’ve watched several tutorial series by Digital-Tutors and all have been using the subdivision method.

I want to hear your opinions which way is really better and if the second method actually works at all since it seems literally unreal to make a mesh like that. I mean, if you had a 8 edge cylinder, you would have to go and manually make say 60 edges, and then connect all those edges together with the rest of the mesh manually, and in such a way to keep the topology.

Tell me what you think please, I want to know which method is better and if the non-subdivision method exists, since I’ve never seen any tutorial or blog post or anything about it.

Best regards,
YoungWolf

I don’t often use subdivision modeling, there’s smooth type shapes which benefit from that greatly, but I find it easier to put in details without having to worry about how those details will subdivide. Also, with subdivision you end up with a lot of extra polygons that aren’t necessary so if you model things specifically you can get a more efficient mesh
However, with a subdivision mesh you can smooth it as much as you need to, so if you’re far away you don’t need to turn it up as much, and if you need to be closer then you can increase it. You can’t with regular polygonal modeling, so you have to make it as high detail as you expect to use it.
Sometimes what I do is use subdivision modeling to make a smooth surface and then in 3ds Max I use the freeform modeling tools to model a mesh on top where I put my details so that I have a generally low-poly mesh with high amount of detail.

Alternatives are pretty much:

  • Digital Sculpting
  • Various spline modelling techniques, including NURBS(Most widespread uses are to a create a surface by rotating a spline, or by sweeping a spline along another spline.)
  • Polygonal modelling
  • Using various modifiers to achieve the shape you need.
  • CSG
  • Photogrammetry
  • Mathematical
  • Certain content creation software-specific approaches.
  • Outsourcing (this one is my personal favorite :smiley: )

Your friend is probably jealous of your results and is clearly trolling you.

I always had an impression that “industry professionals” use a technique or combination of techniques, that will yield them the path of lowest resistance for specific case.
But I can say without a doubt that Sub-D modelling is still a primary way of high poly creation of most hard surface models.

Well Unreal 4 has changed a lot of the “rules” so to speak by not hard coding limitations which usually resulted in the development of lower fidelity assets that even if one wanted to go hi res so to speak due to the demands of the game engine. The result as to this form of making assets to fit into the 3d space is what would be called a “fit to finish” asset the left little or no room for what I would call “Next Gen” development of a working asset if one did want to scale up their project due to the fact that you can not make not enough do more yet on the other hand Unreal 4 needs a rethink as to the process of asset development that is not typical of most if not all “game” engines

UE4 being whats called a closed edit environment then it should be the job of the environment to add features in app that will allow the art to scale as to the needed output by replacing the process usually done as a “fit to finish” process.

For example.

One can add Tessellation if there is need for more detail for things like displacement maps.
In 4.14 Auto LOD for static mesh was added so it can be assumed such a feature for skeletal mesh is in the future.
4K images should be used and materials constructed that outputs to the desired resolution with out texture resizing.

Overall though the fact the UE4 can parse the art means that decisions can be made with asset in hand and evaluated based on fact and not an assumption that building optimization into the creative process is still a requirement.

There is nothing NextGen about butchering the art based on ten year old ideology and I tell my guys make it the best you can do and once in game lets talk about it then. :wink:

I suppose this is the one my friend was talking about. What I cannot understand is, how do you work with so much high polycount? retaining all the topology and connecting all those hundreds/thousands of edges verts and polys. It’s a disaster for a very small object, I can’t imagine that for a huge model like a gun or car or something like that? Also he somewhat mentioned that this technique requires ~ a month to get a good high poly car / gun.

With that said, are there any tutorials or read on the topic? I really want to understand that.

Take some time to research and find out how professionals work. In a production environment, realistic gun with a lot of good reference material can take 2 days or less to model the high poly. Tor Frick has some videos on this youtube channel doing an environment or gun in 1 hour and a detailed sci fi bike in under 4 hours. But he’s also crazy fast and has a really streamlined workflow.

Of course if you have a vague concept and need a detailed model, that may take a lot more time.

There’s a lot of different workflows out there, like trying to use booleans and zBrush to speed up what would normally be subdivision modeling http://polycount.com/discussion/168610/proboolean-dynamesh-hardsurface-workflow-tutorial/p1

Also small things like using floaters or painting in details in Substance Painter or NDO so you don’t have to model them, really saves a lot of time in the end.

It doesn’t take longer to do, the point of using one technique over the other is that it’s easier for whatever you’re doing and therefore takes less time.

For example with this:

30321f39af3d9fc1178ebd93060eb653aa5232fa.jpeg

The body was done with subdivision, but I added things like the panel lines after subdividing because it’s easier to just put those details in than to try and get them in the subdivision mesh. And, it ends up with fewer polygons.
Part of modeling is also figuring out how to make the tools work for you, so you have to problem solve what will be the easiest way to get the shape that you want. It’s not the case that you’re handling crazy amounts of geometry, the tools are designed to help you do that.

I think I’m starting to get the hang of it. Instead of putting everything into single mesh you split it up in pieces. Well I normally do exactly that. For example if I’m making a pistol, I put the receiver on a different part, the grip things that you wrap your hand around different was well and so on. But still I would have to add subdivision to make them smoothed. As I can see here your mesh is the same. Also aren’t the separated panel lines also subdivided?

Thanks for the halp all! Hehe

The panel lines are modeled after applying Turbosmooth. If I did it before doing Turbosmooth it would pinch and there would be a lot of unnecessary loops.
You can see stuff like the side fin shapes that are not using subdivision, just chamfering to get rounded edges. And all of the engine parts are done without subdivision.

An alternative might be what Modo offers (aside from their awesome implementation of Pixar Subdiv): the round egde shader. It makes the mesh look like it’s subdivided (with the nice edge highlight we are looking for) it also creates this bevel effect between 2 intersecting meshes AND can be baked into a Normal Map. Combined with Boolean, a lot of polygon and bevel it is among the most inspiring and fast workflows. With other Packages, you might want to look up Face Weighted Normals, which can create a similiar effect.
Tor Frick is the person to look up for this.

In 3ds Max you can do rounded edges as a render effect in the Mental Ray Arch&Design material–though you can’t bake the effect. For Vray there’s the Edges map which can give a similar effect.
I prefer though to have the rouneded edges modeled in, plus you have more control over where it ends up and how round it is.