Thanks for sharing and detailed production breakdown
i thought there wouldnt be anything better than akeytsu for creating easy animations. im happy if i am proven wrong.
Keith, I just wanted to stop by and say: Thank you.
3d artist Ludvik Koutny talked about the way he builds and models beautiful virtual weapons.
Hi, my name is Ludvik Koutny, I am 25 years old CG Generalist currently living in Prague, Czech Republic and working as freelance generalist. I’ve entered computer graphics as an architectural visualization artist when I was around 18, and then shifted my focus more towards Advertising/VFX field when I moved to Prague to join ProgressiveFX studio, where I worked for 3 years, mainly on 3D computer graphics for commercials and movies.
After that, I’ve joined Corona Renderer team where I worked for half a year on training materials and UI/UX design. Now I am still cooperating part time with Corona team on training materials and UI, but I mostly freelance doing what I enjoy the most – general 3D graphics. That means almost all aspects of production, be it modeling, texturing, shading, scene assembly, lighting, render setup, even compositing.
I can’t really say this was a planned project. My ultimate dream is to work as a 3D concept artist. Just use 3D tools to create all kinds of sci-fi and fantasy design in the same way 2D concept artists do. So whenever I do personal project these days, I tend to just go in that direction to eventually be able to do this full time and get paid for it.
Usually, I get the best ideas right before I fall asleep, and past few days I had some trouble with sleep, mostly staring in the ceiling, so heaps of ideas were just floating around. I often mentally visualize shapes, designs, environments and moods I would like to materialize into pictures (or videos), so when I finally gave up on trying to fall asleep after 4 hours, I just got up, turned on my computer, and started some modeling to kill time before morning comes, and they open the bakery so I can get myself some donuts.
There was nothing particular I wanted to test on this project. It did not involve any workflows I do not know how to deal with, but I’ve never made any firearm/rifle/pistol, so in that regard, this was my first. My plan was trying to focus on some internal logic from which some of the design detail emerge, instead of just doing random greeble-like detailing out of boredom.
When it comes to modeling, I try to employ my “3D concept art approach”. What this means is that I try to utilize fastest and most efficient ways to get the idea of design out of my head into the 3D viewport at the expense of topology quality. The models I produce are in no way acceptable for production pipelines or game engines, but they are enought to make nice renders from pretty much any angle.
The thing is, that there are some preconceptions floating around, implying that if your topology is not perfect quads and triangles, then the world is somehow going to explode. That’s not really the case. The truth is, there are no such things as quads or n-gons. In both 3D viewport and renderers, only way to express three dimensional geometry is via triangles. Quad face is just two triangles, except you can not control in which of the two diagonal vertices cut between triangles occur. That becomes even less controllable with n-gons, but even n-gons are just triangles, except they are triangulated automatically by the software, and you have no control over them.
With this in mind, you can simply employ mindset of not giving a damn about topology unless a problem occurs. If n-gon shows up fine in viewport, it will most likely render fine in pretty much any modern renderer.
So to put this into practice, I usually start basic shapes using regular poly modeling or sub-d (Turbosmooth in 3ds Max) modeling, depending on if the shape I am aiming for is more of an organic or hard faceted character. But sooner or later, I just collapse everything down and obliterate the nice topology with all sorts of booleans to add fine details. And believe it or not, it works pretty much always. Sometimes I get some crazy n-gon that stretches across and around half of the model and triangulates incorrectly, but that can be resolved within a few seconds using a few smart cut tool operations.
I basically try to do same thing 2D concept artists do. If you look at their speed painting techniques, they usually produce incredible images in really short time, that appear to have a lot of detail and dynamics going on, yet if you zoom just a little bit in, you start to see all those details are just quite undetailed brush strokes. It’s not about making each stroke detailed like a technical blueprint, but rather a way to get the overall idea and mood out of your head on the canvas.
I try to do the same thing in 3D, where I do not care about perfect topology that looks great on wireframe screenshots, but rather searching for a way to get the idea out of my head on the screen as fast as possible, to communicate it to other people in sufficient detail.
Often, you should ask yourself if perfect topology is even worth it. If you do not work in a strict pipeline, and you know your mesh won’t be animated with some deformations, then often, staying strict with topology may very well be waste of your time. Employing these techniques can often help you become sort of one man army, and tackle even large projects such as short movies on your own, without needing a team of people.
As I mentioned earlier, most of the details on this model emerged from some basic internal logic. It makes modeling a bit easier since you don’t get bored of modeling details that do not make much sense.
I wanted the rifle to be somewhat sci-fi and have original details, but at the same time be based off some reality. Basic body modeling was nothing special, it was just experimentation with the shape until it looked right. Once that was done, thinking about internal logic kicked in, and details started to appear on their own, for example:
It is a railgun, therefore it will probably have a big muzzle blast, so I will need some prominent muzzle brake. The brake should be able to also hold some compensator at the end, so it will need some screw ridges
Gun will need to be taken apart, so I will need to cut some panels apart in a way they would allow for disassembly. So that’s how cuts in the body appeared. Then the different parts need to hold somehow together, so that’s where screws appeared.
Railgun will have a big recoil. That should be absorbed by something so that it does not injure the soldier, and that spawned the idea of a spring pushing against buttstock to absorb some recoil. Buttstock should be adjustable, so that’s where the button on it came from.
Railgun will be long range precision rifle, therefore it may need a bipod… So I added one.
Long range precision rifle that’s powered by a high powered pulse won’t probably have automatic fire, so safety will need to have just two states – safe and armed. Since it has really strong blast, it should be more obvious once the gun is armed – so I added small red LED light on the safety trigger itself
Long range rifle will need scope. Any modern technology is usually made as modular, so scope will be mounted on some rail system. That rail system should not look like every other rail system on all other guns, so I came up with my own design.
Scope should have some zeroing and zoom controls, that’s where buttons on the scope came from.
And of course, scope should be able to talk to the rifle to control various systems in the rifle for perfect shot, therefore it should be wired to the rifle. That’s where the idea of that wire came from. And since it has modular rail, then there may be more electronic smart attachments, so it would be appropriate to add a few move attachment device ports.
I could go on and on with this, but I think you get the idea. Basically, once you start thinking about how it works, the details emerge on their own, and you minimize experimentation with random non-sensual shapes to minimum.
Part of the thought process is already described above, so I will just add some details here. Again, to accommodate modularity of the scope, it needed to be detachable from the part that fits on the modular rail. Once I knew it had to be two pieces, it was not hard to figure out they will probably be bolted together. Scope itself would be some resistant plastic covered by metal shell, so that detail emerged easily as well.
For the glass, I cheated a bit. I knew I wanted something good looking. There’s pretty much only one recipe for correct glass material, so what gives glass its appearance is mostly geometry it is applied on. Therefore I went to google images and quickly googled for cross section image of digital camera with lens. Once I found cross section of all the lenses, I traced their halves with splines, and then revolved splines into geometry using Lathe modifier. Once I had geometry of the realistic lenses, I just encased them in some primitive tube, threw glass material on them, and just like that, I had lens array that created a lot of interesting reflections and refractions going on.
All of the materials on the gun are mostly procedural. All the wear, tear and dirt is done using ambient occlusion/dirt maps, that are modified and mapped in a various way. There are no unwrapped UVs anywhere on the model. I just throw a box mapping on everything, and then let the procedural shaders do the rest.
The drawback is that you can’t really control exact occurrence of the effects (well, you can, but that’s whole another complex story), so sometimes you will get those effects on the places where they do not belong. Although in general, those small mistakes will get lost in the sheer amount of appealing detail.
So once everything is shaded through procedurals, all that remains is decals. To place decals precisely, I just select only the faces they are supposed to be on, and apply separate planar mapping just to those faces. That way, I can get them right, and still avoid unwrapping.
This is again not an approach that would endure production pipeline, at least not without baking the result down, but for the purposes of my “3D concept art” workflow, it works just fine.
Corona is currently simply the best renderer on the market when it comes to look development. It has very advanced flexibility and support for native 3ds Max features, comparable to V-Ray or Mental Ray, while retaining speed, simplicity and interactivity of the state of art GPU path tracers.
Interactive rendering is best in class, so you can develop all your shading and lighting in realtime, to avoid syndrome of working blind until you hit render button, and awesome VFB post processing features allow you to make pretty much final image without leaving 3ds Max. But not only that, you can actually render interactively and view your scene in real time through all the color corrections, so you can basically develop your shading and lighting while looking at already post processed, almost final image.
Throughout the entire process of shading the gun, not once was I forced to deal with any kind of technical setting, like sampling or anything like that. When using Corona, it’s more like just switching viewport to a different mode, from mode that you model in to a mode in which you make images.
There’s a few things that Corona does not do, like hair, skin and volumetric simulations, but anything else that Corona does, it almost always does it better and easier than all of its competitors.