Topology standards for hard surface modeling?

Elarionus · Feb 2019

I learned quite a bit about topology of models during school, but I wanted to try and push it a bit further, to make sure that I'm working both as efficiently as possible, and to a standard that the industry will accept when I graduate. I've found lots of reference on the polycount wiki for organic models, and I learned quite a bit about that, but I'm primarily a hard surface modeler. I do have an advantage with that, because the shape of objects generally lends itself to the topology. But for rules about triangles, n-gons, and quads across surfaces, I want to dive deeper. I thought I knew the rules, until I watched Senechal and Frick model, and noticed that they use N-Gons all the time, making things that just look like "improper" meshes to me, but it turns out amazing, and most of all, it works.

I just want to learn more about why these things work at certain times, why they don't, and how I can make sure my model turns out great with no shading artifacts, a good polycount, and a delicious topology. I've browsed through youtube, and found some okay things in there, but a lot of contradictions as well. Are there any resources that some people have found to help with learning the ins and outs of the more technical side of modeling/topology?

sacboi · Feb 2019

"Any topology or wireframe that achieves your goals, is a good topology. Always consider what you are trying to achieve and time and resources you have." - Ali Ismail

An ethos I've followed for quite some time and sure considering the amount of material available online nowadays there may at first seem to be a broad difference of approach but really the art of topology can be best summed up as primarily testing for yourself whether it works or not for what you have in mind.

Theory and Fundamentals

hansolocambo · Feb 2019

An edge you won't see is a useless one. Optimize your wireframe to be sure that cuts are there only when they're necessary to the final look of your object. Prefer quads to triangles when modeling (edge loop friendly, etc). Coincidence but I just did today the following image about quad topology. Keep it handy, it might help :

Image: https://us.v-cdn.net/5021068/uploads/editor/b5/1skg1i9tdbsx.png

A good way to have a clean wireframe is to work your hard surface with ZBrush's ZModeler Brush (BZM). This way you'll have a very clean quad topology to work fast and a nice beveled / "chamfered" result with Dynamic Subdiv / Smooth Subdivision set at 4 and CreaseLvl set at 3 (That's my usual setup for hard surface). Final retopo with 3D-Coat (I still find this soft better than anything else for unwrapping and retopology).

Image: https://us.v-cdn.net/5021068/uploads/editor/to/8a4hw49x8s8e.png

Alex_J · Feb 2019

i'd start by making a gun or something like that. Just on your own with whatever you know right now. Just do your best to make it as efficiently as possible. You will discover a lot of the common practices all on your own, as much of it is simply what makes the most sense.

Then go and look at professional artist guns on artstation. Marmoset viewers or pictures with wireframes and texture breakdowns. See if any obvious differences jump out at you.

Then dig into the info on the web. Just start with the wiki and go from there. You can probably get most the info you need from this site alone. Now you will understand it more readily and be able to implement it properly. Of course doing some studying before you get your hands dirty may save some frustration, but, for me at least, a lot of times it's more trouble to try and understand it first if I haven't got my feet wet with some work.

Eric Chadwick · Feb 2019

How The F*#% Do I Model This? - Reply for help with specific shapes - (Post attempt before asking)

http://wiki.polycount.com/wiki/Subdivision_Surface_Modeling

musashidan · Feb 2019

N-gons are an integral part of my hard-surface workflow. The obsession with quads is an outdated and perpetuated myth. For hard-surface your main concerns are that the in-game mesh is optimised and has a clean silhouette. I only triangulate it non-destructively, before baking, using a modifier in Max. This same n-gon heavy mesh will be my high poly. Because I know where to place strategic edges to resolve the sub-D my only concern is the surface shading. This is the sole purpose of the high poly. To bake down the smoothing from. Topology doesn't matter to me, all n-gons become quads upon subdivision.

Everyone has their own workflow. This is the one I've developed that suits me. It's very fast(low/high are created simultaneously and high is practically free) It's non-destructive(low/high are using modifiers) I don't waste time obsessing over quads. It's liberating as you just concentrate on creating the forms and shapes and don't restrict yourself because you might think the forms are 'too hard' to sub-D model due to pinching you can't resolve. This is me though. The workflow has come after years of modeling. Refined with each new asset created. I've dabbled in all sorts of hard-surface methods and found for game asset production assets this is the best way(for me)That's the only way to do the same for yourself, just a case of bringing assets all the way through the pipeline and seeing what's most efficient.

Here's an example of a recently finished model created this way. Note that the on-card vert count is pretty decent for a large weapon with attachments. You will see similar FPS AAA assets come in at 80-100K.

CreativeSheep · Feb 2019

N-Gons I'd stay far away from, unless giant plane on your mesh even then if you are efficient you can do it with quads.

musashidan · Feb 2019

CreativeSheep said:

N-Gons I'd stay far away from, unless giant plane on your mesh even then if you are efficient you can do it with quads.

Bad advice. Why would you stay away from them? Sub-D mesh becomes quads. Final in-game asset before bake becomes non-destructively triangulated. It's all down to understanding topology, first and foremost. Blindly following the 'all-quad' mantra is folly.

sacboi · Feb 2019

Correct.

CreativeSheep · Feb 2019

musashidan said:

CreativeSheep said:

N-Gons I'd stay far away from, unless giant plane on your mesh even then if you are efficient you can do it with quads.

Bad advice. Why would you stay away from them? Sub-D mesh becomes quads. Final in-game asset before bake becomes non-destructively triangulated. It's all down to understanding topology, first and foremost. Blindly following the 'all-quad' mantra is folly.

SubDivision surfaces become triangles in most render engines; if you are strictly talking about Unity or Unreal; I'm unfamiliar with those. Although unless you are creating still, even then I would suggest quad, nGons simply can cause problems.

cryrid · Feb 2019

If someone avoids ngons out of fear that they could potentially cause problems, then they never get a chance to learn how to use them properly so that they don't. Ironically they'll probably create a whole new set of problems by trying to stick with the misguided quad-only dogma when a triangle or ngon would have made their job much easier. From OP's post it sounds like they want to learn how to make proper use of them, so avoiding them outright is counter to their request.

musashidan · Feb 2019

@CreativeSheep mate, I think you may be confusing yourself. Why would the sub-d mesh be going in-engine? As I mentioned above, the sole purpose of the high poly is to bake down from.

It's probably best not to advise people on things you are uncertain of yourself.

oglu · Feb 2019

Take a look at the kitbash from Andrew and study how he modeled stuff.
You will learn a ton im sure.

https://www.artstation.com/artwork/nrqDX
https://www.artstation.com/artwork/ynde5
https://www.artstation.com/artwork/6q03n

musashidan · Feb 2019

Fantastic resource alright, @oglu and a must read for thise new to sub-d modeling, but real-time production assets have different needs than film/vfx.

CreativeSheep · Feb 2019

@musashidan The stuff which Andrew created as posted by @olgu mostly if not completely shows all quad models, therefore far from confused. Most, if not all render engines at this point can accept or convert polygon models to sub-division models, making quad models especially for film/vfx essential, even with games I wouldn't model completely in n-gons as n-gons have become more of a lazy approach to modeling whether you agree or not. Some have modeled successful still images from n-gons but I wouldn't make it a hard and fast rule to follow. Besides that may be your workflow to bake down most things, but what you do is not what everyone does.

Alex_J · Feb 2019

I see tons of n-gons in Andrew Hodgson's work there, @CreativeSheep .The only person who said anything about modeling totally in n-gons is yourself. Read the thread again, look at the referenced images -- the argument for n-gons seems clear to me, but your's is not. What is the point you are trying to make exactly?

oglu · Feb 2019

Andrew does show on Artstation two diffent workflows. One are the background obejcts or orbject without subD.
The others are mostly quads cause of the displacement they use on those SubD meshes.

There is still so much confusion about ngons and "bad" geo.
Ngons and tris arnt bad. But you need to know where you should work with quads.

1. If it does shade fine it does bake fine. So ngons arnt bad for normal baking. If there are shading issues clean them up.

2. If it does deform fine during animation ngons arnt a problem in this either.
In most cases they wont deform nice so quads are better here. Take a look at all the disney chars they have a nice even topo.
Even if quads are better that does not mean your trousers cant have triangles in areas they dont deform that much.

3. If your still render does look good and it does fit into your system ram everyhing is fine even with "bad" geo.

If not you have to take a closer look why.

Image: https://us.v-cdn.net/5021068/uploads/editor/1n/gljuyomsfxd4.jpg

4. The main reason to work with quads in FX is displacement. If you have thousands of objects in your scene with displacement you have to be carefully how much poly your basemeshes does have. If you have only a view objects close to the camera with displacement the renderer will take care and subdivide the objects more. But if you have hundred millions of unique poly in your scene think twice.

Here a good vid about creases and why they are used. Same reason for a clean SubD mesh.
At min 26min.

https://youtu.be/xFZazwvYc5o

musashidan · Feb 2019

CreativeSheep said:

@musashidan The stuff which Andrew created as posted by @olgu mostly if not completely shows all quad models, therefore far from confused. Most, if not all render engines at this point can accept or convert polygon models to sub-division models, making quad models especially for film/vfx essential, even with games I wouldn't model completely in n-gons as n-gons have become more of a lazy approach to modeling whether you agree or not. Some have modeled successful still images from n-gons but I wouldn't make it a hard and fast rule to follow. Besides that may be your workflow to bake down most things, but what you do is not what everyone does.

Again, you're confusing 2 totally different workflows and when you talk 'render engines', you're lumping them all in as one. I'm specifically talking about modeling hard-surface production assets for a real-time engine. You keep talking about still images which has no relevance whatsoever with production assets. For a real-time baked asset the high-poly can be anything: from a polysoup CAD .stl model to a perfectly all-quaded sub-d mesh. It doesn't matter. Only the end result of the surface shading matters. The production asset itself will be triangulated BEFORE being baked and put in-engine, so N-gons during modeling have ZERO relevance either. You say 'N-gons have become more of a lazy approach'. I say you simply don't have the modeling experience or are failing to grasp exactly what I'm trying to explain. Of course what I do isn't what everyone does, but it's a very common workflow in the industry. I'm merely giving the OP some food for thought, based on real-world production experience, which I presume is what he is here for.

VFX/Film meshes, on the other hand, are either dense polygonal meshes(like those in Avatar) or - as @Oglu explained - well modeled quad sub-d meshes that will receive render-time subdivision displacement, rendered offline - as opposed to in real-time.

Before I got into the games industry I spent years modeling purely sub-d. I used to obsess over quads to the point of ocd. I just could not have ANY non-quads. Looking back on it now I laugh at my foolishness, all those hours I wasted. With the amount of time this we spend staring at vertices, picking your battles is something I've learned the hard way.

These are 2 different pipelines and workflows, with different goals, considerations, requirements, and end results.

sacboi · Feb 2019

Take a look at the kitbash from Andrew and study how he modeled stuff.
You will learn a ton im sure.

https://www.artstation.com/artwork/nrqDX
https://www.artstation.com/artwork/ynde5
https://www.artstation.com/artwork/6q03n

Yeah, nice one oglu! - promptly bookmarked : )

cptSwing · Feb 2019

In this case, lazy is good.

Topology standards for hard surface modeling?

Replies