Nerdy Navigation

3D Modeling Process Thread

Home Forums General discussion Arts & Culture 3D Modeling Process Thread

Viewing 15 posts - 1 through 15 (of 41 total)
  • Author
    Posts
  • #37901
    Max
    Participant

    I got a message the other day asking about some of the 3D modeling I’ve posted in the past, and I thought it might be a good idea to start a dedicated thread showing what goes into something like that for anyone who may be interested. I’m pretty bad about taking in-progress screencaps of existing work, so I’ll probably start a completely new model and go through the process from concept to completion with an eye towards explaining each phase of the process as I get there. I work in Autodesk Maya because it’s what I’m most familiar with, but there are plenty of other programs out there. Blender is free and very full-featured, but it is different enough from Maya that I probably won’t be much help if anyone is going to follow along with the home game.

    The basic process I’m going to follow will be:

    Concept Art: this will involve a basic idea for the thing I’m going to model, followed by turnaround drawings of every side. I’ll show you how I go about making sure everything is aligned to preserve proportions as well. (spoiler alert: I don’t always do this part, and sometimes jump right in to modeling with no reference. this is a good recipe for a racecar that has taken over a year to complete)

    Modeling: how to actually turn your 2D drawing (or photos as the case may be) into a polygonal model. I’ve not worked with NURBS, so if you want to know anything about those you’ll have to dredge up an old school movie special effects artist or somebody that does industrial CAD stuff.

    UV Unwrap: This is by far the least fun, but will have a huge effect on the final visual outcome of your piece. What this involves is basically taking all the vertices in your model and laying them out like a bearskin rug so you can apply a texture map to them in an image editor like Photoshop. It is called UV-ing because you lay out the verts on a cartesian coordinate plane with the axes U and V to avoid confusion with your X-Y-Z grid out in your 3D scene

    Texturing: This is where you generate your color maps (the basic color on your model), your specular highlight maps (how reflective a given piece of your model will be when light strikes it), and normal maps (this is basically simulated geometry that makes your model look like it has more polygons than it actually does).

    Rigging: I’ll probably do a basic rig so I can get a pose of the model. Rigging involves creating a virtual skeleton and binding your geometry to it so you can then move it around. It is conceptually identical to constructing the kind of armature you’d see inside a Tim Burton stop motion character, except you have to paint skin weights per-joint to determine how much influence each joint has over each part of the model. If you’ve ever played a game and seen somebody’s elbow collapse like a folded noodle, you are seeing bad skin weights.

    Lighting/Rendering: In order to get the final beauty shot we’ll have to set up a basic light rig.

    In all likelihood, this thread will be going on for a super long time because I don’t think I’ll be able to spend a whole bunch of time all at once on any given step. I haven’t yet decided what I want to model for this. I feel like a character from D&D would be appropriate, but if you do one you have to do all of them. I have some concepts already done for a chibi Rocketeer in the style of Skottie Young that I made for a buddy to model, and I have several vehicles I want to do personally I could jump into. I’ll have to think about it.

    #37902
    Max
    Participant

    Here are some 3d scenes I’ve made in the past so you can see the kind of thing we’ll be working on:





    #37904
    M
    Participant

    Well I’m definitely looking forward to this. If you feel up for it, it would be great if you’d go through the steps here and so on.

    Also I … u… err… is that Mr. Garrison’s “It”?
    Right there, between special needs plane and castlevalia!

    #37905
    Max
    Participant

    So from top to bottom those are:
    ·The opening scene of Space Quest II translated pixel by pixel into 1X1X1 cubes

    ·a zombie character for a never-completed game pitch proposal
    ·Don Karnage’s Tri-Wing Terror from TaleSpin

    incidentally this is also rigged with some controls that allow all the control surfaces, guns, and throttle to function like a real plane:

    ·This was a concept vehicle I did for a proposed World’s Fair level in a cancelled educational game I worked on. The setting was a near-future retro city that had a lot of retro-future jetsons-esque influences. After that project ended I decided to go ahead and model it myself
    ·the reaper fight from the first castlevania game that literally took me 20 years to beat. It was the first of the cube-per-pixel projects I did; there are over 3 million polygons in that scene:

    ·A formula 1 inspired vehicle that was intended to save me some time drawing a webcomic for the folks at http://www.shiftylook.com that ended up taking me over a year to get to that state. When you hear the phrase “a month of sundays,” that is what you’re looking at there. I was going to pitch them a comic based on Rally-X, and I didn’t want to have to draw the super complex suspension and engine elements over and over. The plan was to toon-shade it so it would look like the rest of the art in much the same way almost every cartoon now is handling vehicles post-futurama. Long story short, Shiftylook is getting shut down, and not one frame of the comic ever got made. I’m still working on it anyway, I’ll just strip out the Rally-X elements. I’m thinking of calling it Stoplight Grand Prix (a humorous street racing term for street racing) and having it be somewhere in tone between Speed Racer and Wacky Races.

    #37927
    M
    Participant

    So, how much work goes into a model like that?

    Let’s say you gotta do a humanoid creature from scratch – like the Krang-enabled zombie there, with concept art and everything all the way to rigging and texturing etc (but not things like animating it). For example, how long did you spend on that particular zombie? And how much would it matter if you were making the zombie as a low-detail background character (say, you had to fill relatively far away street with 100 in-game zombies so no details needed and low triangle count and such), and on the opposite end (if the zombie was something you needed to lunge at the camera in a 1st person game and take a bite out of the player, for example).

    And another issue, how much of said work would be recyclable? Like if you had to make another one that was basically a similar zombie but looked like Steve from accounting, something more than just changing the colors in the texture I mean.

    #37929
    Max
    Participant

    Typically in the game industry you’ll have dedicated concept artists who are designing your characters. The last AAA game I worked on (as Art Production Manager) had about 60-ish people on the team, 20 of them were artists. I had an Art Director who had two assistant Art Directors (all of whom pulled double duty as concept artists, and one of whom pulled triple duty as a fill-in modeler). Each of the assistant art directors had an area of responsibility: one was characters/vehicles/props/fx/ui, and the other was environments (there were a LOT of environments in the schedule). We had one guy who was just a concept artist full time. These guys did their own textures and set up the rigs for characters and vehicles that were then passed off to animation. Occasionally they would do hero props (large one-off or limited reuse items for the environment). We had a Tech Artist whose job it was to come up with tools and scripts to help other artists, as well as create tricky effects and shaders in our game engine (Unreal). We had a Lead Animator and four other animators of various experience levels (we weren’t using any mocap for this project) who hand-keyed everything that wasn’t done procedurally. I had two effects artists with an un-filled lead fx position who did all the particles and such. Two UI artists, one of whom was dedicated to my team and one I shared with all the other games we had running at the studio. The one on my team full time handled all the visuals and UI concepts and the other guy did all the back-end scripting. I worked pretty closely with the UI team since the other art directors had their hands full with the modelling and fx guys. The rest of the art team was comprised of environment artists of various levels, broken up under two Lead Environment Artists.

    We were working with an existing IP (it was not officially announced prior to cancellation, so I can’t tell you what it was due to Non-Disclosure Agreements we all signed), so the characters and vehicles we’d be allowed to use were established up front. The way it would go is that Design would tell us what kinds of weapons and behaviors we would need for Character X. Then concept would take a pass at translating the established look of the character into our art style. I would usually budget 3-4 working days for that to get a finished illustration. The rule of thumb is that American workers are productive for about 4hrs of an 8hr day after factoring in meetings, emails, calls, random interruptions (Canadians allegedly only have 2hrs of productive work per day), so we’ll say that it would take 12-16hrs of dedicated concept time to get me a colored, 3/4 shot of a character with no turnaround angles. If there are any tricky details, those bits would be blown up larger. Those then run past the Art Director for signoff.

    Once you have concept, it goes to the modeler. Artists have specialties, so if you have a person that is better at hard surface than cloth, or monsters than humans, you might try to account for that when handing out the work but you can’t always. Usually what I would do here is try to get the modeler to put together a proxy model with a rig for animation to get started with, that way Design could have something to work with for game balance prior to having a completely finished asset. If that wasn’t practical, I would encourage Design to just substitute an existing asset to prototype gameplay (they were already doing this usually, it wasn’t often I had to suggest that) until we could get them something viable. A character like that zombie might take a professional game modeler a week to knock out, give or take. How detailed you need to make it is a loaded question. A game like Dead Rising that is all about swarms of cannon fodder zombies will probably have a much more generic, low-rez character than a game like Resident Evil that is all about crafting a more intimate encounter for several reasons. Having a bunch of characters on screen at the same time is more gpu intensive, so you want them to be of minimum viable complexity. Smaller they are on screen, the less detailed they need to be. Many game engines automatically handle LODs (Levels of Detail), but sometimes the results are ugly and someone will have to manually create a lower resolution asset for use at a certain draw distance. It is substantially cheaper to instance textures and assets than it is for the GPU to draw a bunch of new things, so especially in horde type games you tend to see little variation in enemy models. You also try to build your environment pieces for maximum re-use for the same reason. Textures are a lot more expensive to render than polys, so if you can get a bunch of things to share a single texture that is great. Your mileage on re-use will vary, but you want to get as much re-use as you can out of everything to save you production time as well as rendering time. If all you need to do is swap out a head, or a shirt, or lop off an arm, of course you will start with an existing model and make edits. That will take a fraction of the time it would to do the whole thing from scratch, unless the geometry in your model is terrible (this should not be the case on a professional project, though sometimes you will need to re-work art that you got from an outsource company). I’d estimate a model like that zombie would take somebody 20-30hrs, and I would say it is really only suitable for a fodder type enemy at that level of detail. A hero zombie for a big scare close up would probably be double that.

    Animation and FX are their own body of work, some of which can be done concurrently to the model and some not until it is finished. You can get a bunch of work done with a proxy that has the correct proportions, but you’ll need to take a final pass on all of the clips after your finished art comes in to make sure there are no clipping or other issues. You make game animations in little clips, so a character jumping may consist of as many as 5 separate clips. You have the jump anticipation, the push off, the transition through the apex, the fall, and the landing for instance. Each of those may have variations (is the character falling for an extended period of time? Are they running when they start the jump or stationary? these kinds of things). I think our main character on that last game was getting close to 200 distinct clips. An average repeated-use enemy had about half that, and special case enemies half that again give or take. As with LODs, your game engine will usually have some provision for blending between animations but it can often be terrible as well. That means you end up with a bunch of specific This-into-That animation clips over time that can be hard to account for up front in your schedule.

    #37934
    Max
    Participant

    Here are the options I’m considering for this project:


    @M
    seems interested in characters, I have a simple Rocketeer concept in the style of Skottie Young I drew for one of my animators to model. This is attractive to me because he doesn’t have a proper face and will therefore take much less time than a character that does. Also a mix of hard and soft surface elements:

    I’ve been more interested in vehicles lately. Here is a whimsical mars rover I want for another comic project. I have already started it, but it isn’t that far along:

    This is a different car that is a companion to the previously posted racecar that I think would be cool to work on too:

    #37936
    M
    Participant

    I’m starting to see why even big indie games like Amnesia had like 2 character models…

    #37939
    Max
    Participant

    We really didn’t even get into what goes into environments. 3D games are non-trivial endeavors. It is possible and not uncommon to purchase pre-made assets and just make changes to them, particularly through the Unity asset marketplace to cut cost and get around your team limitations. I typically encourage aspiring indie devs to figure out what they can make, and build a game around that. Robots don’t need animations as smooth as living characters, vehicles need even less and can often just be physics based. Unless you can hire talent, you should work with your limitations instead of fighting against them.

    #37941
    M
    Participant

    I’ve messed around with Unity a lot, I absolutely love how much you can do with nonexistant resources. Just using assets is rather hard for me, I mean, I can do it with backgrounds and things like that pretty easily and still make something relatively unique.

    But no matter I/we think about, there just isn’t anything that would need at least SOMETHING in it, be it at least 1 or 2 detailed characters (robots, drones and vehicles are also hard enough), or couple of dozen of really plain ones. And if I just throw in a knock-off Gordon Freeman who shows up in 3 other indie games it’ll be quite the distraction.

    Then again, other than buying those or going the ol’ pixel art route, there really isn’t anything really feasible for a micro-project.

    #37942
    Max
    Participant

    Yeah man, you can get some skills, or get some employees. It’s not going to make itself, you know? But I assure you that a prototype that has nothing in it but boxes will be all you need as a proof of concept.

    #38097
    Max
    Participant

    In the interest of getting some momentum on this thread, I’m going to go with the Rocketeer. Since @M seems to be interested in characters, and I already have turnarounds of this guy drawn it seems a good one to jump onto.

    A little background on this character: the Lead Animator on my last project was super into the Skottie Young kid-avengers designs, and wanted to know if I could do up a Rocketeer that he could model in his spare time. In the world of AAA games, you tend to have a team made up of specialists. This guy was is a great animator (he did the Geico gecko for several years before joining our team), and he was doing some practice modeling on the side. Ordinarily when you are doing concepts for characters or vehicles, you’ll do a series of thumbnail sketches to nail down the silhouette before committing to one. This one was a little different, since really what I was doing was style-matching an existing character. I did some research on Skottie Young’s work, and I looked at some Rocketeer reference, and got to work. That resulted in the image I posted up earlier.

    Next I drew the character in a T-Pose. A T-Pose is typically how you model your character, since it makes it easier to check proportions and ensure you have enough geometry in the joints to stretch correctly. I drew the front view first, then got a straight edge and chose some reference points on the character (top of the head, eyeline, bottom of the chin, etc.) to run lines down the page. Then I did the side and back view, and did the best I could to align them to the guides by hand. The results of that are here:

    You can see the guidelines there that I’ve erased. You’ll also notice that there are some left to right symmetry issues that need to get sorted out. To resolve that, I’ll bring the image into photoshop.

    #38098
    Max
    Participant

    The first thing I did was to cut the background out and do a bit of cleanup. This makes the images easier to work with. I also cut each view out and pasted it back to its own layer (appropriately named for easy reference in the future).

    After dragging out some guides, it was painfully obvious how out of true the different figures are. So starting with the front view, I’m going to do some image correction. I used the Skew transform to get a quick straightening, then I had to cut the fin and slide it over to be centered. I also re-drew the mouth holes since they were totally out of whack.

    #38100
    Max
    Participant

    Here it is, all fixed. Once I got one side of the front sorted, I deleted the other side, copied the good side, and did a Transform->flip horizontal in order to have an image with symmetrical details. I copied the layer as a reference and played around with the opacity of it a bit to make it easier to overlay on the side and back view. I used the fixed front view to make sure all the proportions of the side and back agreed with it, and here is the end result. You can save yourself a lot of cleanup time by not having such a jacked up original drawing:

    The next step will be to save out 1024 X 1024 squares with each image centered on it so I will be able to set up reference planes in Maya.

    #38110
    Max
    Participant

    If anyone wants to comment or ask questions in this thread, feel free. If you want to save out the reference and follow along in the modeling program of your choice to post the results here, that would also be awesome. Blender is free, and you can get trial versions of Maya and 3D Studio Max from Autodesk (I’m not sure if you get hamstrung versions or if they are time limited though).

Viewing 15 posts - 1 through 15 (of 41 total)
  • You must be logged in to reply to this topic.

Random Nerdy Show Network Posts

Skip to toolbar