Fundamental Flaws in 3D modeling or is it me?


Tiles, you are correct, it was vertex normals. I kind of did ignore your first post, because the example you showed was on a flat surface. I was wrong. This Vertex Normals helps me a lot. I was able to change my shading the way I wanted.

The difficult part is once you have so many verts, and you need to adjust shading on those, it can become a nightmare, but it did work. Thanks for that.

Now, the problem is when I detach something from the main part. It seems the detached vert normals are in a different location then the original vert normals and the shading changes. When I turned on vertex normals for the detached item, I noticed I had 2 verts in each location (which makes sense), but what did not make sense was the vertes normal line was a bit in a different location. This is tricky. As of now I manually copy normals from one to the other after detachment. Lots of work, but it works.

But anyway, thanks guys for the help. I think I got it from here.

darthviper107, I am understanding you, but in my case I have tons of sticking out parts and flanges. It’s not possible to flow every edge through that. I would have to have 1000 edges in the cylinder to get each correct. On top of that I noticed every time I move an edge like you do, I see that in my game engine after zoom in. So it’ s a game of compromise. So I prefer to keep them spaced out even and work with vertex normals. But I do like your work.

I will post this once I finish the entire thing.


Keeping the surface normals consistent and flowing while adding details is one of the major challenges in subd modeling. There are methods, though, and I particularly like the one explained in this car modeling tutorial:

It uses a guide mesh and adds details using the shrinkwrap modifier in Blender. I found this method works exceedingly well to avoid the surface issues while adding as much detail as you need.

If it works for cars, it works for plane hulls. The surfaces turn out beautiful without the need to worry about normal kinks. It just works.

I am, however, unsure whether Max has a similar function built-in. There’s the conform tool, but it is very slow compared to the shrinkwrap modifier in Blender.


hvanderwegen, interesting post. I don’t know anything about Blender, but I did wonder if car is more difficult than an airplane with details. And I tested it out. Modeling the angles of the car was way more difficult for me, but not many parts to extract from rounded surfaces. And the surfaces were way more flat than a rounded fuselage. So shading of extracted parts was not a problem.

I even played with a fuselage of a fighter jet that is not rounded, and there it was much easier as flatter things in Max are simpler. It is only extracting tons of detached rounded parts on rounded surface that became tricky.

I even asked a famous company that developed aircrafts for DCS World and they were not aware of adjusting normals :slight_smile: They told me those errors are impossible to get rid off, but their examples were easier than mine, so they never faced them.


One more thing. If Max was smart, it should have a default normal presets based on flow lines. That was you could simply say, give me shading for point X that is natural between 2 points. Otherwise I have to find the shading myself moving the normals. Nightmare, but I will get it done.


It’s a shame, but this is something that would have been perfect for NURBS modeling then converting to polys or spline modeling techniques to generate polys for a light weight model in BURBS first or straight to poly patches, but because so many people have labeled NURBS “old” it’s impossible to find any good tutorials for NURBS or spline modeling and then people are left to toil away at super dense meshes.

As is without NURBS or Splines, there are a million ways to do it. I would retopo the the front of this into a couple of pieces to minimize the 5-Edge or Star poles when welding/merging them together eliminating the pinching (Cone, Cockpit+Window, Under Cockpit, Top Hood.) Then I would merge the retopo’d front to the back of the plane, assuming there are no other problem areas. It’s a little hard to explain in text, but the main advantage of having separate parts is to be able to control the number of edge loops use far less geo and keep it al quads.


Yeah, its you. Google “smoothing groups” and “T-verts”.


It’s not me at all. I noticed you all never built a anything of the caliber I am building. This tool is one big bug, that’s why it’s not a precision tool. On top of that, the entire approach to shading in 3ds max and modeling in general I think is based on Phantasmagoria (polygon cylinder is a Circle in computer 3d modeling). That will mathematically never work. It is good enough ( I actually don’t agree), but problems will come up when doing any precision based parts.

It is true, that I can adjust Normals for those verts, but it’s a nightmare to adjust them when you have thousands of them, and sometimes simply impossible. This is all due to the fact that ONLY evenly spaced cylinder sides are shaded evenly. The minute you “FLOW” one of the lines like most of you suggest, that side changes it’s shade grading. The entire 3d modeling is based on that.

More, I noticed people that designed this tool were aware of this problem and they did 1 thing to help: they allowed the user to place a vert in the middle of the cylinder wall, without affecting the shading. Anywhere else then the middle and the shading is recalculated. It has to be, in order to “represent” round objects.

All of you were telling me “IT’S ME”, but none of you took the simple cylinder example I gave you and posted the result. I can find flaws in all cylinder 3d models with details on top of it. All I have to do is flip them in a specific way.

All T-Vertices, and those “topography theories” are beautiful on paper, but have almost 0 realism in practice when working on a precision item.

P.S. All you have to do is understand Chaikin’s algorithm to realize 3d modeling based on that is actually silly.
I remember when Google became popular and promised to undo what Microsoft did for the past 20 years. I swear whenever I get a chance in my life, I will undo what all software shops did in the last 30 years.



your post is a perfect example of kruger-dunning.

Is poly/subd modeling flawed? sure, but so is nurbs. even clay isn’t perfect. but the main reason you fail is simply because your modeling approach is flawed, and your techniques are flawed (to say the least), while claiming to know it all.

here is some “on the ground” truth for you: Your hole plane is modeled the “wrong” way. nobody builds the way you do, which led to you suddenly hit a wall, no surprise here. learn how to model properly, your second model (probably ever) will be better, no doubt.


Also, everything in VFX and animation is built on these polygonal techniques so if you think they’re wrong, there’s many experts that would disagree with you.


Please, learn to use the software and be a bit more humble, probably all the people is answering you has made models way more complex than your low poly plane, that BTW I don’t know what kind of engine are you using, but today’s engines don’t need to have a plane made of 3000 polys :stuck_out_tongue:

Those two are the same cylinder, it’s duplicated to show you the result and the wireframe, you have to fix your normals as nearly everyone told you, the problem is not the tool or the technique, it’s the person who uses it, specially if that person denies that the problem is that it’s not proeficient using the tool, something normal when it’s the first time using it, but at the same time a little bit of humility and thanks to the people who try to help comes well.


“I noticed you all never built a anything of the caliber I am building” -Hang on there pal. I see you’ve had replies from industry pro’s and I’m not that bad a modeller myself (10 years since I started learning Max at uni -and I’m still learning to this day).
Two possibilities exist: either you are right and everybody else is wrong, or vice versa. I may only be 40, but even I’ve learnt that when you face this question in life, its invariably you who is at fault.

Max works just fine. You can’t cut an unconnected line into a curved object and not expect shading anomalies. 3d modeling requires a three-sided poly at the very least (the clue is in the name). Once your line is connected to other vertices it becomes a poly and you can adjust its shading via smoothing groups.

You’re like some guy who’s miss-spelt a word, decided that its the language that’s at fault, and has vowed to create a whole new language because of it.

PS: I’m curious as to how a model like this compares to the caliber of model you’re working on? (its less than 30k quads btw):


I fully agree with this.


I don’t think you read my post, or understood me. Maybe I was not clear.

  1. The fact I have to fix my normals is a proof I have a point. The only reason we are “fixing” normals in 3D modeling, is because of the shading error on rounded objects.

  2. Also, I said it is possible to fix the issues by adjusting Normals, but once you have thousands of verts it becomes impossible. Not to mention what I said in #1.

Finally, your cylinders, even though I believe you fixed the Normals for those 2 verts, are positioned incorrectly to prove your point. You need to rotate them in such a way that the light shines along side. That’s how shading works in computer 3d modeling. You are proving nothing in your image, but i have no doubt your example works once you rotate. The problem is once you have more than 2 verts. With multiple verts close by it is a nightmare.


Nice Ship, but has nothing to do with what I’m saying. It is not even rounded. Give me the max file for that ship and I will find shading problems all over it. Zoomed out item like that from that angle does not count.

And last but not least. Don’t think I’m some kind of arrogant know it all person. I do appreciate your posts. I simply stand for what I believe is correct. If I caved in to you, I would learn nothing, and things would never improve for any of us. Knowing history and myself I know very well I have a very good eye for when things are wrong, and I’m not afraid to question it. That’s all.


No such thing as wrong 3d model. “Flowing” lines does not matter in 3d modeling the way you think it does.
Actually “flowing” them changes all the shading. Cylinder walls once adjusted will create irregular shading. You can’t adjust all Normals, as you can’t even know what shading to give at that point. You can only use your eye. Not to mention everything is a triangle once the graphic cards process it. Shading and final 3d result is what matters.

I did not hit a wall, and like I said again, give me those “good” 3d models by “experts” and I will show you even more errors.


You don’t get it, and you need to practice 3D modeling more

You need the topology to line up with the details that you’re putting in and if it doesn’t and if you have shading problems most likely you’re trying to work with a surface without enough polygons. You need to make your topology line up or you need to add more loops/subdivisions until it does.


Based on your posts and the topo mistakes visible in your screenshots, i really think that a correct understanding of geometric continuity would get you past most of the problems that you are currently fighting.

Here is someone with decades of modeling experience (link to correct timecode hopefully):

Please take my word for it: his models are bullet-proof, even zoomed in with a microscope.

You are correct that you shouldn’t have to ‘fix normals’, but you won’t have to once you figure out better workflows.


this is probably the most moronic logic i’ve ever witnessed. So, you say… more experienced people try to basically educate you on this issue, but if you would listen to them you wouldn’t learn anything?

uhm so why did you post here then? just to vent, because your way of modeling (and thinking btw.) doesn’t work and it’s all the computers fault?


If you know to handle your software it’s not impossible at all, and in any case you are not working with that amount of vertices, you came here asking for help, people gave you help, and you did not liked the answer, probably because you think everyone else in the world must be wrong, because there cannot be an answer because you cannot be wrong, then you started to tell people that they don’t know what they talk about and that they are wrong and “you have a point” and you are right.

In the end, you don’t like polygonal modeling? why don’t you use Rhino instead?

Ok man, here you go, really, show some respect, don’t try to give lessons unless you really know what you are saying:

No, that’s not how shading works.

Once again, you have to know how the software works, you have to learn to model, it’s not a nightmare at all, it’s pretty easy in fact, and something quite standard for real time work.

Hey stand for what you believe is correct, ignore facts and others that know much more than you, that does not make it correct, but you can stand by whatever you want LOL

As I said, quit modelling with polygons, go to rhino and enjoy nurbs, but keep in mind that you don’t have a point at all, you are simply wrong due to your lack of knowledge.


Polygon modelling is inefficent for stuff like this, but it is possible. Pixar are not the best poly modellers with regard to resource efficiency because they dont need to be. Some people at Pixar model with NURBS and then convert it to edit poly. This is an easy way to maintain curvature, but the mesh density can be extremely and unecessarilly high.

Believe it or not, but planes and cars for that matter are not made from a single piece of material. Not only can it be easier to model panels separately as they would be in real life, but as a by-product, the model automatically looks more realistic.

Your main problem is that you are cutting holes and then smoothing when it should be the other way. Smooth the curved surface, then cut the hole.