Dough Boy - Cute Diesel Punk Infantry Robot
high dynamic range
PROJECT THREAD
Hey Polycount!
Brian here again. Going to be modeling the Dough Boy, a Teutonian basic infantry robot that can transform to higher tier units when need be. Definitely looking for critques. I know there's a lot to learn, and Polycount has been by far one of the few places online to get really thick, juicy feedback to get better on.
Objective: Model and Texture a current-gen RTS Unit, the Dough Boy, for the game "The Maestros"
Planned Workflow
Concept > High Poly Sculpt (Maya, 3DSMax?, ZBrush) > Retopologize > UV (ZBrush UVMaster, Maya) > Texture
Specific Rules I need to hit:
LATEST UPDATE:
WIP 13
So, after a discussion with the rest of the art team, we decided that having our characters be textured in that Van Gogh stylish way was a bad idea for now, so I've "normalized" the textures here, basically doing what I ususally do for texturing this guy.
Does anyone have any critiques for the texturing at this point? I'm hitting against that rendering wall I usually hit, so I'm at a loss as to how to push this further, given the size of the character in-game.
And my current textures:
(The Gloss map. Not sure how to plug this into Marmoset or UDK)
Hey Polycount!
Brian here again. Going to be modeling the Dough Boy, a Teutonian basic infantry robot that can transform to higher tier units when need be. Definitely looking for critques. I know there's a lot to learn, and Polycount has been by far one of the few places online to get really thick, juicy feedback to get better on.
Objective: Model and Texture a current-gen RTS Unit, the Dough Boy, for the game "The Maestros"
Planned Workflow
Concept > High Poly Sculpt (Maya, 3DSMax?, ZBrush) > Retopologize > UV (ZBrush UVMaster, Maya) > Texture
Specific Rules I need to hit:
- Final texturing needs to emulate the Van Gogh mark making style. (This piece is contributing overall to a game project I'm Art Directing on)
- Make better materials. Metal needs to look like metal. Leather like leather, etc. T that effect, I need to spend more time on the Specular Map and actually make a Glow Map (I think that's what it's called?)
- Learn better high-poly modeling techniques. Hard surface high poly mdoeling and Sub D modeling still eludes me. People say to use 3DSmax over my more familiar Maya, but it's really hard to find a good Maya tutorial series on completing a fairly complex hard surface model. Kinda tempted to actually give 3Dmotive money at this point. There's a LOT of free tuts on 3DSmax high poly modeling, though. Perhaps I should just take the time to learn the software.
- Understand what is valuable and what isn't. I'm new to modeling robot characters, but from looking at references from League of Legends or Team Fortress 2, there seems to be a lot of things I can forgo, like accurate joints, to speed up and keep my efficiency high for the model.
LATEST UPDATE:
WIP 13
So, after a discussion with the rest of the art team, we decided that having our characters be textured in that Van Gogh stylish way was a bad idea for now, so I've "normalized" the textures here, basically doing what I ususally do for texturing this guy.
Does anyone have any critiques for the texturing at this point? I'm hitting against that rendering wall I usually hit, so I'm at a loss as to how to push this further, given the size of the character in-game.
And my current textures:
(The Gloss map. Not sure how to plug this into Marmoset or UDK)
Replies
High Poly WIP 1:
Beginning highpoly model in Maya.
High Poly WIP2:
Blocking out arm, hands, wrists, legs.
Really stumbling through this,but fortunately, we're working with fairly simple geo. Just need to figure out how best to do these medium difficulty hard surface forms.
So, I'm still trying to understand the subd workflow in Maya. It's hard. I get the whole concept regarding control edges, etc, but when I hit problem situations like "Oh, how do I add control edges where all these planes meet together in these strange curvatures and angles and at starred vertices?" my mind sort of locks up and has to sit down. Like I need an example solution before I can move forward because internally I don't "want to waste time" experimenting for the solution.
I'm really tempted to drop $40 dollars for the 3Dmotive tutorial on hard surface modeling in Maya, or suffer through badly-mic'd videos.
Keep in mind not everything needs to be connected especially if you're going to be baking this out to a low poly.
Look up floating geometry that will help tons.
Oh definitely, bolts and nobs as floating geo is totally internalied in my head.
This would make an awesome 3D print, right?
right?
Iono, just a thought I had for a friend's birthday coming up.
So, I'm working on making the hands, and I don't have any explicit plans to make them super accurate in terms of how they actually work, but just enough to say that it CAN rotate.
Does anyone have any solutions for the fingers? Or should I just make them nice looking "boxes" and call it a day?
And does anyone know how to make blade ends in SubD?
This reference sketch from someone has been helping me bunches:
Are you going to make a weapon for him (her)?
So, I think I sculpted in MOST of the larger details I need? I think if I hack away at any more details, it's either going to be very slight noise that I heavily doubt will show up in-game or will make the character too noisy. I think I'm going to fix the cloth areas more and then begin retopology.
For the low poly though, I'm literally going to not really bake from it and just make it as high res as they do for the TF2 Mann Bots, since the guy will be holding a gun.
I'm not absolutely certain if there might be close shots of the character in our main menus, etc, so I'm hoping that allows me to keep some fidelity to the hands.
Skipping over to the "Bopper" gun.
Having trouble with the cooling jacket, trying to find the proper tutorial for it. If anyone can help me with that, that would be super appreciated! Otherwise, everything else has been fortunately straight forward Inser Edge Loop or working with very very simple primitives. Forward grips and stock/trigger left to make.
Awwww that is one adorable lil' bastard so far dude.
What're you having trouble with on the cooling jacket? I might be able to do a mockup for you.
Here's what I have now. Made the gun overall feel a LOT more stubby, fixing a lot of the proportional issues I saw relative to the character's feel. I'm definitely getting heavy influence by a couple SMG designs from WWII, though game wise, it's more familiar with Bioshock Infinite's Machine Gun, of course :P
@Hadhijah: So, I'm fortunate enough that the shapes and geometry I'm doing through subD moodeling is mostly just realy simple control edge placements with simple extrusions of cylinders and squares. I still feel incredibly raw doing complex shapes/intersections, and this usually happens when I'm well into the attempt and I don't know how to fix it midstream.
In terms of specific examples:
The cooling jacket examples are frequent enough on the internet that I figured out to just grab vertices in a pattern and chamfer them. It was just when I tried adding thickness to the plane through extrusion that the holes wouldn't be perfectly circle, or the hole isn't a clean cylinder. Unless it doesn't have to be?
What I have here was not numericaly clinical, I kinda just set all the inner loops of the circles zeroed to a common location per vertex and then scaling it by hand. It didn't feel precise, though perhaps it's more important it LOOKS the part as oppose to exactly IS (which I think I have now).
Other places, like the receiver and the charge handle housing and ammo housing (both boxes), had unsatisfactory solutions for me. Right now they're split into three different geos, but I know there's a way to make it all one geo starting from the receiver's base mesh, but I . . . I don't know how to do it. Tried solving it with Adding Edge Loops but it's still hard for me to "flow" through the challenge, and I get stuck. And then I just split the geo and get it all "hacked together." It's overall just learning SubD I guess. I'm still really trying hard to find a comprehensive tutorial online that I can jive with that does teach a good portion of SubD challenges, or maybe I should just buy that tutorial from 3Dmotive.
Regarding some technical stuff though, unless you plan on including those cooling holes on the SMG in the Low-Poly, you'd be better off just modeling them as intrusions, and not making them actual holes. Also, you don't really need them to be perfect circles, since the important part is that they look circular enough, like you said.
However, when I do stuff like that I usually try to get the holes as circular as possible, and I usually start off with a plane, add the holes, using either a script I have for Maya or by just using a cylinder as a guide, and then I bend the plane to create the cylinder shape.
Keep up the good work, looking forward to seeing more!
So . . . not a lot sculpting I need to do, at least, at the speed I'm moving at? If I need to get back into this, I don't think baking specific components is going to be terrible since I think the lowpoly geo will stay stable.
I think my High Poly is done. Shoes I'm probably just going to straight up make a final version in low poly, but it's prolly not going to show up in game.
I don't want to take out the "style card, so don't care" too soon, but do you concur with the proportional issues?
Oh might be cool to think of a neat way to attach that little ball on top of his helmet, that area looks really simple and you could do a little fun mechanical stuff up there. Not too distracting, but just a bit of construction
Hells yeah! Someone I know had access to an industrial 3D printer and gave me a little 2 inch print of the first character model I ever made.
It was seriously one of the best gifts I've ever gotten... so damn satisfying to hold one of your models in your hand.
Anyway, this is looking cool man! I always like your characters... definitely a lot of personality. Oh also:
I watched that one when I was subscribed to 3D motive. It's pretty good, and definitely goes into depth about how to route support edges. However... if you're looking for something that teaches good topology, this isn't it. His attitude is basically "it's a static asset, and you're going to bake it anyway, so eff topology. If it looks ok, then it's fine."
It's been awhile since I' watched it, but from what I remember no it definitely wasn't that bad.
For some reason I remember him using a lot of n-gons just for the sake of efficiency, rather than spending time trying to figure out how to quad it out. Since he didn't plan to sculpt on it, and since it smoothed properly in Maya, it was fine.
It might make an animator cringe a bit, but from an environment artist's perspective it was perfectly fine.
I'll take a look for sure if the arms are too long. Fortunately, I don't think scaling the geometries down after baking won't be too horrendous. I tried using my Selwys cloth brushes on the high poly of the upper arm, but it just turned out so weird. It would severely deform the raised stripes coming down the arm. I tried pinching here and there to give it better creases, etc. I'm thinking any errors may not be seen at game camera angle since this guy will be fairly small, like a WC3 unit.
As an aside: Baking normals down into a 3000 triangle model is really freaking hard. I'll show you guys what have tomrrow, but wow, I'm literally taking to baking pieces one at a time.
WIP 9
Spent four days figuring out normal bake issues and figuring out dDo2.
This is a rough material pass. I have not touched it with handpainting, but I feel like, after learning how to work dDo, this has given me a good base to start from. All the scratches and scuffs procedurally generated
We're standing on about 3k tris. A little high for my tastes, but . . . maybe can take out the extrusions from the diaper. once again, this is supposed to be an RTS unit, hence the ridiculously low poly.
Texturing is 80% done I believe. Barring any additional polishing, I'm hoping my whole assertion with trying to integrate a Van Gogh mark making to the diffuse is coming through.
Actually. I'd like some thoughts on this from Polycount. Does it work? Does this handpainted style make it feel at least slightly unique?
Work process wise, had fun and difficulty learning dDo, but it helped a lot to give me a solid base for what is essentially a metal character. And nDo helped me do some realy simple normal map additions.
Standing on around 2800 triangles. A little high, but realy freaking low for an RTS character in UDK, right?
doughBoy_tPose.obj (click to view in 3D)
I'd recommend getting rid of the spike on the top of the helmet and bring it closer to the typical pie-plate helm. For the other side, I'd model something similar to a German pickelhaube; the two should be visually distinct from the top view even without dramatically different colors. (Of course, adding a Kaiser-esque mustache to the other guys couldn't hurt...)
Also Starcraft 2 "most units range from 1000 polies to 2000. However, unit like the Thor have 8k polies."
His large hat covers too much of the character from an rts view.
I suppose I could just straight low poly it out.
Looking at the model in sketchfab, you might want to reconsider your poly distribution on the low poly version though. There are a couple of areas where you spent a lot of tris that do not add to the silhouette at all. Removing these will reduce your tri count a fair bit. Although 2800 tris is not an outrageous amount for a single RTS unit nowadays, you still need to make every tri count.
Some pointers:
Don't be discouraged by my words, I am an awful nitpicker.
Looking at a model full screen that is actually seen at a fifth or thenth its size in game is always making it feel blocky.
My first advise would be to put a pitch black material on the low poly, zoom out to about the same distance the unit will be seen in game and check the silhouette from all angles. Now duplicate your model and start reducing tris, checking back on the original model whether or not it actually makes a significant difference in shape or silhouette. Be brave and bold, everything that does not add to the overall silhouette has to go.
Now, reducing this guy to somewhere around 1.5k tris should be doable. After checking back at the sketchfab model and talking from my experience, there are a couple more areas where you could cut a bunch of tris.
For starters, lose that extra loop that gives thickness to the trim of the helmet and round out the trim with some of the polys you just saved. His eye/lense can be optimized by just using one tapered tube for the whole thing instead of all those bevels wich can hardly be seen (since they are mostly covered by the helmet). Go over the ball joints, more tris can be saved there. A small thing but worth doing: remove the thickness of the iron sights on the gun, you certainly won't need it.
Now I don't want to disect every inch of the model making you feel bad about your geometry. Over the years I worked with many artists that felt the urge to "round things out" even when it is not necessary to do so. I've had trainees completely losing it over the fact that I still found plenty of room for optimizing their models after they where SURE they had nailed the geometry. I AM a frigging nitpicker. :poly121:
All of the above is meant merely as pointers. I am certainly NOT saying that this is THE way to optimize things.
In closing: be bold and brave as I said before and check the silhouette from the actual in game distance. That alone will help you a lot.
This is strictly a topology and mesh update. Apologies to Dinh
@some3Dguy Hopefully this is bold enough. Got this cut down to 2100 tris, something a tons more acceptable I think? The engineers aren't yelling at me anymore, fortunately. Besides any UV adjustments I've made, this model iteration has already been sent to our Rigging Artist, so I definitevely cannot make any more changes. At this point I'm going to go texture polishing.
doughBoy_tPose_bChoi_06032013.obj (click to view in 3D)
And don't worry, I won't pester you with more optimizations and paintovers, it was just that I did not want your engineers coming after you with torches and pitchforks. :poly121:
A failure with the first texturing "scale" has been fixed. Hopefully that Van Gogh emulation is coming through, though understandably it's mostly on the helmet. The small and medium marks, instead of a WoW approach to it where the blends are more explicitly cleaner. But some stuff still carries over like edge highlights, etc.
Any critiques from anyone?
Btw, how do you make cloth/leather looks more like it's supposed to be across the Diffuse, Glow, and Specular maps?
Here is the link to the PDF. Page 3 is the one I'm referencing. I think it will help.
http://media.steampowered.com/apps/dota2/workshop/Dota2CharacterArtGuide.pdf
So, after a discussion with the rest of the art team, we decided that having our characters be textured in that Van Gogh stylish way was a bad idea for now, so I've "normalized" the textures here, basically doing what I ususally do for texturing this guy.
Does anyone have any critiques for the texturing at this point? I'm hitting against that rendering wall I usually hit, so I'm at a loss as to how to push this further, given the size of the character in-game.
And my current textures:
(The Gloss map. Not sure how to plug this into Marmoset or UDK)