Why You Should Never Model with Subdivisions Active - 3D Modeling Fundamentals

In order to make good and clean 3D models, it's really important that you don't model with subdivisions active in the viewport. This will create trouble later on. We explain exactly why in this video.

Support us by checking out our merch!

Subscribe now!

Our episodes are also available on all major podcast services!

#3d #tutorial

[FlippedNormals]

It's great to see that this video has sparked some discussion, but we just wanted to clarify a few things.
We're not trying to say you should't use SubD, you absolutely should. Just don't model with it active. You can preview the SubD while you model by toggling it on and off. This is a great practice!
In the end, you should subdivide your model to get to the final resolution.


So absolutely, use SubD in the end, and use it throughout as a preview for how your final model will look, just don't have it active at all times.

[joeedh]

So as others have said, you can model a really low poly cage and then apply one level of a subdivision modifier. Me and Geoffry Bantle had this workflow in mind when we redesigned the modelling tools in Blender.

[cg_cookie]

The point about getting the low poly model as close to the subdiv version as possible in important areas of contact is a great one!

I find the example a bit odd since in Blender it is easy (and default) to see both the subdiv and the low poly mesh at the same time by turning it on in edit mode but disabling it effecting the vertices. This is hugely helpful at certain points so that you get immediate feedback without having to constantly jump in and out of edit mode. Other times it's distracting and unnecessary. Turning it on and off as you go depending on the situation is a great way to work.

But yeah, turning it on for the cage leads to gross results. I can't recall ever seeing someone seriously working that way, but I'm sure I have shown demonstrations with it on because it looks cool. Will knock that off 😜 -JL

[Draknark]

Thanks guys, I don't have access to a 3D school n my country. This is truly helpful :) love your videos thanks for all the tips!

[magnusrexus]

I agree but the flaw in you guys argument is that once you get to a certain point most people convert the model into a mid poly model to continue modeling or sculpting. Then the model is retopologized.

[jorionedwards]

So this is less "never use" and more "don't use it as a crutch."

[MMXVIII_]

“If this here was a dog, it would be put down” HAHAHAHA that’s an amazing response, thanks for the video

[craftygamerones]

I’m confused, you can still unwrap the model and animate it with the modifier apply

[keyay]

You can also use a subdivided mesh as a base for a shrinkwrap modifier for your low poly. Not sure if this will work for all the cases, but shrinkwrapping low poly onto subdivided mesh + smoothing when necessary (and to be honest, sometimes manually adjusting some verts in tight spaces) really helps me getting nice and clean low poly.

[dpsa]

While I agree that modeling with Sub-D on should not be done if you don't know what you are doing, saying it is "never" ok, is pretty misleading. And as you mention, I agree it's best to switch between the Sub-D preview and the base mesh when working, instead of blindly working on either. But Sometimes you just want to quickly create geometry which is easy to manipulate, sometimes referred to as cage modeling, and you are never supposed to use the lowest Sub-D level in production, but instead you are using the cage model to get the general shape and features in, and then subdivide it once or twice (and apply them) to get to the actual usable model, which you can then edit further. It is way easier to get the big shapes and features modeled in with a model that is only a couple hundred or couple thousand points, instead of trying to tweak a production resolution model. This of course is relies on whether the modeler knows how the cage model behaves when subdivided, so an inexperienced modeler can easily make mistakes and create very bad topology if they don't know what they are doing, but that applies to basically any modeling technique. Cage modeling is pretty much the equivalent of sculpting at the lowest possible resolution to get the big shapes in, before subdividing and moving onto detailing. And the same idea applies to different deformer modifiers, like FFD in 3ds Max, or Lattice and Meshdeform in Blender - it is easier to modify big forms by editing a couple of points, instead of moving hundreds or thousands of individual vertices by hand.

[Grom]

It doesn't matter how lowpoly mesh looks when you are modeling highpoly, because there is no use for that lowpoly anyway. sometimes to achieve desired look in some areas low poly cage can be twisted in a nasty way, but again why bother if most of the time you will be redoing lowpoly later using retopo over the highpoly mesh.

[jamiehemming]

I'm not sure.. if your aim is to get a "quick and dirty" high detail model for normal bakes, etc then I see no problem with it. If you need to use the base mesh for an in game asset for rigging etc, on the other hand, then it would be a problem.

[KingKamakazzi]

As an amateur modeller... yeah, I probably needed to see this video. Love you guys, always giving great information and advice.

[andrewmajewski]

While I generally agree that you want your production model to be very close to the subdivided model, couldn't you just subdivide the low poly model on one iteration to help with achieving form, freeze the mesh, then keep manually adding details? This essentially would still be your 'low poly mesh' that you would later sub-D later for render time. The only thing I can really think of for not modeling this way is if you need very fine control of your topology, but I would think that would be handled in the retopo phase in general. Did I miss something?

[cold_static]

In the Render Properties tab there's an entry called "Simplify" which let's you reduce the level of subdivions for your entire scene if you turn it on.
What I do is I set the Max Subdivisions to 0 per default (saved in my startup file) and assign a keyboard shortcut to toggle it on/off (Shift + Q in my case).

That let's me easily see what my mesh looks like with/without subdivisons and I use it all the time when I'm modeling.

[Gadgetmawombo]

I've found that the "smooth" modifier can fix this without adding polygons. Probably not ideal still but the smooth modifier does a decent job of making the model look like the sub d model without adding polys. But of course it can only work with the polys that exist so you might need to add more manually.

[thirun]

How I model usually (Blender and Maya) is that I subdivide the mesh, go into cage mode and start adding polygons to follow the subdivided mesh, essentially using the subdivision as a 'guide' to create a cleaner looking low poly. I love doing that!
Note that I'll have the subdivision active almost all the time.

[josephvanwyk]

Why not make a tutorial of starting something from scratch, take us through the pipeline process, instead of saying this is the best way to do it - difficult to learn this way.

[TheFlydesign]

I'm using ZBrush to do my modeling. Low cage modeling is so helpful to knock in your forms. I might hit dynamic to see the flow of the model, but I trust my low poly model as my base.

[MsNathanv]

There ought to be some kind of Betteridge law about 3D tutorials that say "never". I can count the number of true 3D nevers on the fingers of one hand. The whole reason your applied subdiv model is hard to work with is because you applied 2 levels of subd instead of 1. And of course afterwards, you go ahead and apply a level of subd on Suzanne to get the bounds more accurate (leaving it hard to work with btw, but that's perfectly reasonable if you no longer plan to work with it; or, if you plan to work with it via a mesh deform or something.) Your plan shouldn't always be to apply a level of subd, but when it is, you might as well be working with it displayed. Now, if the point is, don't make a model with unmodified topo that's awful, concave faces, zero-area faces, degenerate quads, highly non-planar faces? Yes, I agree, don't do that. That's one of the first technical things modellers should learn. But if your anticipated, *applied* subd fixes those problems? Feel free. Indeed, are you starting a character with cubes for legs? Yes, model with a single level of subd, get the boxes where you want, apply that single level, and then move on to the next level of detail (perhaps with a new level of subd, *if* you intend for your legs to have 16+ verts per loop and so plan to apply that single level as well.)