Polycount to Professors: Ngons Are Useful!
polycounter
Today I tried to chat with some professors about how I've seen a lot of industry pros use Ngons in their hardsurface workflows. To my surprise, they blew up on me about it. I was told that no amount of evidence that I "thought" I had would make using Ngons an acceptable practice in the film or game industry. A short scolding later, and I was surfing polycount's Technical Talk threads finding examples, posted by active professionals.
We've all had experiences like this. You were taught in Uni that Ngons are literally satan incarnate, and that if you ever used them that your purity was forever stained and no one would ever hire you, or love you. You were told that no professional on earth uses Ngons because they are sloppy, and they were ugly, and that if you ever showed them in your portfolio you could kiss your career goodbye.
But we know that is not the full story, there's more to it than that. We know that while in most instances quads are desirable, Ngons are a useful tool in places that quads just produce bad results, or just simply increase your workload trying to solve. Ngons are misunderstood, and have their uses, and this hatred and phobia of them needs to stop.
Let's see if we can convince a few educators with a sheer wall of overwhelming evidence and testimony.
Hard surface modelers, industry veterans. Take a stand against half truths today. Share your experience with ngons, and how you've used them in your career; show us work from your portfolio, games you've shipped with your ngon-riddled workflow, and potentially some examples of how they are useful.
We've all had experiences like this. You were taught in Uni that Ngons are literally satan incarnate, and that if you ever used them that your purity was forever stained and no one would ever hire you, or love you. You were told that no professional on earth uses Ngons because they are sloppy, and they were ugly, and that if you ever showed them in your portfolio you could kiss your career goodbye.
But we know that is not the full story, there's more to it than that. We know that while in most instances quads are desirable, Ngons are a useful tool in places that quads just produce bad results, or just simply increase your workload trying to solve. Ngons are misunderstood, and have their uses, and this hatred and phobia of them needs to stop.
Let's see if we can convince a few educators with a sheer wall of overwhelming evidence and testimony.
Hard surface modelers, industry veterans. Take a stand against half truths today. Share your experience with ngons, and how you've used them in your career; show us work from your portfolio, games you've shipped with your ngon-riddled workflow, and potentially some examples of how they are useful.
Replies
those are beliefs from times when accidentally importing a lambert1 would crash your engine.
Ngons can create bad tesselation. On the other hand, they're better than quads in some sub-d topologies. Your engine may, or may not care about them... it depends
I say it's fine really. I usually have an N-gon mess of polygons when I am bulking out forms that I eventually pick at to get some smoother edge flow which may or may not end up with ngons in the final high-poly. I'm sure people that model cars have had to use it.
There are a lot of useful cases to have planar ngons but I can't think of many for non-planar ngons since they can get pretty hard to "control" at times due to how implicit and unpredictable they can get depending on the context topology.
It's for sure a very useful intermediate shape for any workflow though. Pretty much just as useful as curves or splines might be in a sub-div workflow.
And if, in spite of that evidence, it's still suggested n-gons are bad for some reason they can't articulate then you're talking to someone who, unusually for someone in an academic role, isn't interested in learning or critical reasoning.
Ugh - Yay!
But in high poly stuff? Do what the hell you want so long as it smooths well.
lowpoly meshes are something entirely different, and i agree you should triangulate them, but quad only for lowpoly makes even less sense than for highpoly, as in engine all will be tris.
don't make it a rule! 5sides ngons can work great or can be horrible.
Any N-Sided polygon with more than 5 sides can easily be cleaned to quads, or quads and tris, or quads and 5sided Ngons. But on flat surfaces it doesn't matter much.
If you want a rule...
As long as it shades fine, don't bother cleaning it.
but this also only applies to bake meshes, as soon as you want deformation in your highpoly (say it's a subd mesh for film) a whole new level of complexity comes into play.
For games we can fake a lot. For Film for instance, unless you just bake it in, floaters we can use HEAVILY are a no go.
It's a complex topic and has many facets
Of course, the use of pentagons and triangles in subdivision models are dependant of the flow of curves/edges. There are some basic rules to use them. Once you understand all, you'll see that all those people that are "sick" with quads, in their meshes... they have pentagons and triangles subdivided (pentagonal and triangular topology).
now if you would only translate it to english to make it valuable for a much broader audience, that would be great :P
http://community.thefoundry.co.uk/discussion/topic.aspx?f=36&t=26239
Pixar says shut the fuck up.
Seriously though, i think maybe your professors just suck a big one at getting their point across. I personally believe that the practice of "quads only" is an important first step for aspiring artist, as it can help more easily visualize edge looping and good flow. It also (in my experience) makes UV's easier to read.
However, triangles/ngons are awesome.
This reminds me of that one guy... "i don't believe in triangles".
Ngons are ok in a certain context.
The danger in saying they're OK, is many students will use them in an inappropriate context, which will cause artifacts, then point to threads like this proclaiming them to be ok.
:P
I stick to "Low = No. High mesh = Hell Yesh"
well plenty softwares basically couldn't handle non quad/tri geometry for quite a while.
I think that context comes back to the point we were discussing in skype chat about fundamentals. If you teach students right off the bat that n-gons are totally fine, we might see more abusive uses and less quality results.
There is definitely merit to teaching quad-only. But NOT as an absolute. What this all boils down to is: Learn the fundamentals before you start breaking the rules intelligently. And that goes for any tradecraft; Writing, cooking, engineering, etc. Maybe it's good that students learn this on their own. But I feel that for the good of the industry as a whole, curriculums need amending to add a clause at the end of the course about ngons. Give students a nudge in the right direction. Let them experiment and find out what works on their own.
But profs going out of their way to convince students that n-gons are the devil? NOPE.
A 5 sided polygon is a hexagon but if you have trouble counting above 4 you can call it a ngon, just like a octagon is a ngon.
The n in ngon is technically a variable. n could equal 3 or 3 million. 3-gon, 4-gon, 5-gon, ngon. After about 5 sides you just throw in the variable, N, for "No idea how many sides" or "Not going to bother counting that mess", heh.
If you can't easily count the number of sides you MIGHT have a problem. That doesn't mean you actually do. You MIGHT need to fix it IF it causes a problem. That fix might include the use of other non-quad geometry. That doesn't mean you need to freak out and create a giant quaded nightmare.
If they tell you to stop using ngons tell them to take it up with autodesk and get them to change the programs to exclude ngons since they're so much trouble.
A 3D package with no polygons, good luck with that.
The fact that your teachers are sorely lacking in basic mathematics and that its the foundation of 3D, worries me...
theoretically you are right, practically you know damn well what Ngons are in our field and when people refer to polycount that they mean trianglecount
tris, quads, ngons, I don't think it has to do with the missing ability to count further than 4
Mathematicians typically call anything past 6 an ngon. Artists seem to cap out at 4.
It's not really an inability to count. It's an inability to recall the actual names of the shapes.
3 triangle or tri-gon
4 quadrilateral or tetra-gon
5 penta-gon
6 hexa-gon
7 heppo-gon? 7-gon
8 octa-gon! Nailed it!
9 umm... nine-o-gon? 9-gon
10 10-gon. WAIT I know this one... frick... its a... deca-gon!
11 screw it, n-gon.
12 n-gon
13 n-gon
14 n-gon
15 n-gon
16 n-gon
The n is a variable and it could mean anything, mostly it means "I don't want to or I'm unable to, count the number of sides". No shame, it's not a failure of counting past 4, its more not needing to.
The point I was trying to make, was that triangles and quads are technically ngons. If they say no ngons I guess he can turn in nothing and get an A?
The common advice is that ngons are fine when you have perfectly flat surfaces in sub-d modeling, while this is certainly true, you can also use ngons and especially pentagons in many more situations, including complex organic/curved geometry without issue.
Highpoly model from Brink, see how many ngons you can spot:
High render of same model.
Those evil ngons are in there somewhere, I can feel it in my bones!
^ Finished low, again, no virgins were sacrificed to the gods because I used ngons.
When it comes to lowpoly modeling, I use ngons extensively as well. However, when I finalize the lowpoly geometry (after my uvs are done and I've ran some test bakes to ensure my model is good to go), I usually go in and cut ngons into quads and triangles to make sure that I'm not counting on the automatic triangulation to handle it. I also occasionally manually triangulate quads so that the automatic triangulation doesn't create convex shapes that should be concave and vice versa, so my mistrust of the 3d app triangulation is not mutually exclusive to ngons. I'll be the first to admit that I do this as much for my own OCD as I do for any technical reason.
Also, keep the end use for the HP mesh in mind. If there is a small smoothing error but it will ultimately be baked down to a 512 normal map and only take up a few pixels on screen, your time will be better spent on something else.
Lol Mark, this is a pretty tangential rant. There's another equally plausible situation: "There is no need to count the number of sides", because the computer does that for me, and/or because it has no relevance in most situations. Modo doesn't crash if I create a 13 sided face instead of a pentagon.
The reason n-gon for is used to describe any face with greater than 4 sides has absolutely nothing to do with math skills or lack thereof. On the contrary, it has everything to do with how 3d meshes are actually exported in various file formats. When you go to export a 3d model from Max for instance, you have three options:
1. Triangulate entire mesh
2. Export triangles and quads (all n-gons, that is, faces with more than 5 sides will be converted to quads and/or triangles)
3. Export all faces as they are (no conversion)
Older applications and file formats did not support n-gons, some didn't even support quads. This is the root of the common usage for these terms today.
Additionally, n-gon is used as there is, as far as I am aware, no other English language word that describes numerous polygonal shapes with greater than 4 sides. If you want to refer to both hexagons and pentagons, the most accurate term would be n-gons or simply faces/polygons (but those two are generally too vague). Though if you want to get really anal retentive, even triangles and quads are n-gons, but I'm you sure, and everyone else here, understands the meaning in this context.
Basically anything that might wind up with unintended triangulation, and ngons tend to be worse for that.
Quad only (or mostly quads) is simply handy because it's easier to manage edge loops.
+1 to this. I use ngons aplenty in the process of making, but I'll manually break them down at the end to be sure. Convex/concave quads, too.
http://www.scriptspot.com/3ds-max/scripts/angle-loop
http://forums.cgsociety.org/showthread.php?f=7&t=108412&page=1&pp=15
A lot of the images are broken, but in the ones that remain up you can see an assload of n-gons on a character model he was working on. It subdivides and deforms cleanly, and the pinching/bumps from using quads, tris, and n-gons together on a high-poly model actually adds important surface details that would normally be sculpted in Zbrush these days.
So, not only is he using n-gons, but he's specifically taking advantage of the fact that they subdivide weirdly to make his model better. Steven Stahlberg's a madman.
That seems like a pretty useful script, but the download is broken
Try this: https://mega.co.nz/#!HtYwlBRR!lRDhoRrQD8_O_-zWvCAb0QTXId9hqmazSPZGf2H2v1A
Now that you mention it, he posted some nice wires of his works.
In fact 90 per cent of the good working practice that polycount has taught me over the years has been been totally ignored by most of the companies I have ever worked at
This thread has been a great read. Feel more comfortable using them now
Just for the record I wasn't imagining artists struggling to count past 4 which is probably what triggered you and Neox, heh. I had in mind something where the geometry makes it hard to figure out what n is, as in geometry obscures the sides and it's hard to count, crap like this.
Not knowing and not caring what N stands for, are the same thing in this case. If you physically can't count past 4 for whatever reason, apathy, crazy geometry or I guess extreme stupidity like you thought I meant, the N in ngon means something. It comes from mathematics. Its grounded in tradition.
If you don't care to specify a number its just ngon to you. To someone else, they might need to describe in more detail, like they were writing a tutorial that centered around using a 20 sided polygon. They would say, 20-gon, 20 sided polygon, icosagon if you're a polygon-nerd. Someone not knowing how the word was put together might mistakenly say "20 sided ngon".
Ngon is not something that someone just pulled out of their ass to mean "more than 4 sides".
Seriously though I've noticed that people seem to prioritize quad meshes over everything else, as if it is the most important aspect of acheiving "good topology" which no concern for edge loops or distribution of geometry. Even on low poly models. It's easy to find people here on polycount who have really bizarre topology for no reason other than they have some sort of phobia of using any other kind of polygon.
Maybe it's just because "all quads" is an easier rule to follow than "use good judgment"
This is not necessarily true.
I'm working on a character that I could, in theory, just Zremesh and then worry about topology once I get to the low-poly. The problem with using this method is that I don't always the have polygon density, edge flow, or general overall layout that I want/need to get the cleanest results possible.
In this case, I'd rather fall back on modeling at least a rough version and then import that into Zbrush knowing that I have detail and general shapes where I need them instead of letting an automated solution try to work things out.
Also, there are other bonuses to modeling things out by hand. For example, now that I have a high-poly with good topology, I can just use one of the lower subdivision levels (with some tweaks) as the low-poly. That's not necessarily something you can get with Zremesher, especially when you need things like support loops or very specific topology.
In all my school programs I've been in, I've never been taught edge loops. We're always told to just "maintain quads" and avoid "n-gons." The rest of the course was just modeling different objects every day with a goal of "efficiency" and "different tools".
I have to look on Polycount wiki when someone tells me on here "your vertex count is wrong" or "your geometry looks weird" because honest to god, I was never shown this stuff some of you guys are talking about in school.
Even now, I'm still kinda confused when certain terms are thrown at me. I have no idea if what I'm doing is really right or wrong in terms of modeling without having to come back to Polycount and ask.
The ironic thing is if I do ask for a professor for help, they do start to show me how to model with edge loops or distribute polygons in a certain way, so it's not like they don't know. It's just that it's never officially taught for some reason.
Quads are very pro, you know...
Edit:
If the topology is generally good, then it shouldn't be much of a problem. Fine, it's a pain in the arse, but if the structure is there it's quite easy to try and optimse - high poly that is.
Looking at Mark's illustration, they are all techniques I've used in the past, it means that there needs to be an amount of tensile strength that is also working with the rest of the mesh.
The fact is that Zbrush only admits quads and triangles. Any pentagon or n-gon is triangulated.
In contrast, mudbox admits meshes with pentagons or n-gons and it works perfectly, like any major 3d app.
Neox, translating all my tips and tutorials into english would suppose a pain in the ass, literally. My native language, the spanish, is very difficult to translate into english due to a different grammar and expressions. When i readed an english translation by google i ended o.O
Anyways, i must say that i only use: triangles, quads and pentagons (the three pinnacles of 3d modelling). Any n-gon with more than 5 sides is forbidden in my workflow/technique.
pretty useful for shaping landscape quickly without sculp tool
sometimes uv Quickpeel doesnt want to calculate n-gon properly, so I have to turn it into triangles or quads.
for zbrush , n-gon will be turned into tris or quads. when you export it.
if it bad enough, like bad triangle, it often create pinching , when you dont have enough polygon to sculp.
bad ngon will create nasty pinching. and ghost edge crossing concave shape.
you can fix this by adding/cut edge between concave.
if you do nothing but sculpting maybe, my experience is that i would like to kick any artist using it as a cheap shortcut and i have to make this crap mesh production ready. hooray for holes or endless spirals
for zbrush it is tru, ngongs will not qork it will triangualte the ngon into one tri and one quad (if its 5 sided) if <ou have a symmetrical mesh might mean both sides are different topologywise. yes for sculpting try to avoid anything but quads and tris. but one subdiv on you ngon based mesh and you have a nice quadonly mesh
all I can imagine are the ngons that were probably there before it was subdivided
For creating high poly objects who cares how you make it as long as it looks and beaks great. Make it anyway you want, in any program you want with any plugins you want.
For in engine game assets of course use only quads and tris. Some engine will subdivid your ngons for you, but some wont and can give you undesired smoothing/normal issues.
Thanks a lot for that! :thumbup:
A: 37 Polygons, 70 Triangles, 37 Verts
B: 26 Polygons, 68 Triangles, 36 verts
C: 32 Polygons, 68 Triangles, 36 verts