Hello fellow polycounters,
not much work these past weeks, so I decided to make a quick piece for the portfolio in the hope of improving it a bit. Also needed to take my mind off a bigger personal project.
I decided for a real life weapon, since I have never done one and the complexity seems interesting and challenging. Huge quantity of reference images on the web, so that's great!
The choice fell on the Kriss USA Vector (second generation). Still not sure if I'll add any accessories like light, or holo. This is the weapon:
After spending a couple of hours gathering references I started modeling the Barrel Shroud (I thought it was a suppressor, but this instead is used to avoid burns from overheated barrels, the more you know...):
Part of the Frame and the Charge Handle:
The Cartridge, with Magazine, and the Bolt Latch/Release
Stay tuned for more
Replies
1. needs beveling to make softer
2. Same needs some kind of bevel
3. is this a block out? if this is going to be low poly geo it looks ok, but if its going to be high poly again softer so the normal map gives you bevel info
Overall at the moment everything is to sharp.
And this is where I stopped for today.
Added Top and Lower Rails, Side Rail Mount Points, Trigger, and Grip.
Current tricount is around 15k.
Current tricount is 27k
Indeed they added a lot of geometry (around 2.5k additional tris), there are few more insets on the backside of the grip that can't be seen from the screenshots I posted.
I'm considering whether to remove them and just bake them from the highpoly but I'm not sure the normal alone would do it justice, especially if rendered in engine without pom or tessellation.
I mean I could leave them modeled for quality purpose since this is not going to be a production piece and hence it doesn't have stringent polycount limits but I guess I'll decide later on.
Thanks for the feedback
The stock connector and the top piece of the stock itself were the most complex parts to model, and I resorted doing them with modo mesh fusion (subd+bools). Still got to retopo them though.
Also have decided to go with iron sights for the accessories. The most common used with the Kriss Vector Gen II are apparently the Magpul MBUS gen2 which I find very sexy in design, so I'm currently finishing modeling these, then retopo the few pieces that need it, then highpoly, and so on.
These are the bad boys going on the top rail:
http://community.foundry.com/discuss/topic/94400
Also halfway with the highpoly:
I'm enjoying this project so much, can't wait to start texturing it.
Anyway I just did a test bake of the barrel shroud keeping its size in uv space small as it would be in the end, and I think the bevels might even be a little too much on some edges.
So, been very busy lately but I have managed to clean the lowpoly a bit. It's roughly 30k tris for the main weapon, 40k with full accessories.
As it's evident, I have been very generous with curves and cylinders to get as much quality as possible remaining within AAA games requirements (I think modern fps have 40-50k tris weapons?).
I'm now working to add the final touches to the highpoly and then I have to finish the UVs and get to the baking. Still not sure if I'm going to be able to squeeze it all in a 4k map or if I'll need two, but we'll see!
Here's the lowpoly current form (between the various changes I also removed the grips insets as @Polygoblin recommended). Also I hope I have not been making too many errors given this is my first project of the sort (you guys are so mean jk)
there's not too many errors, this is how you become a stronger artist. I've been roasted here plenty. the direct crits are normally just because of limited time
Mspaint is the shit, yo.
The weakest part of this is the optimization. I'm no fan of cutting down so much that it looks like it belongs in 2003 unless that's your budget, and I have no issue with a 50k poly FPS weapon as long as you're gaining visual fidelity from every poly used.
That said, you've got some areas that are so geometrically dense that the edges render into eachother on screen(red). These polys won't render at more than 1px on screen, so I have to ask if they are necessary.
You've also got some edge loops still in there that don't add anything to the shape or help with the shading at all (blue).
You've got some many-sided planes inset into your model that can be detatched and replaced by a simple quad that covers that same hole (orange).
There are bits that honestly should be floaters. There's no gain in having these (yellow) bits modeled into the lowpoly. Highpoly, yes. Lowpoly, no.
There's also some bad triangulation on the pistolgrip that will mess with your smoothing that can be ezily fixed (purple)
I would also remove the green bits as they won't be seen in first person view, and don't add to the silhouette. (greenygreen)
And lastly every one of those bolt lips can be detatched from the main body, so you can optimize the flat side of the gun that has a bunch of 32sided holes cut into it without it affecting the holes going through it. See below.
On your current setup you've got a bunch of 32 sided cylinders modeled into a quad face on the side of your weapon. This generates 36 triangles on the quad per bolt hole. If you split the edges going around the cylinder and optimize the 32sided hole into a hexagon, it'll generate 10 triangles instead of 36. This will also smooth and work better with the normal map as you've got fewer and shorter/fatter triangles.
On the conical shapes that don't feature a hole through the model you can split and collapse the hole, then remove the vertex in the middle.
If you end up doing all this, saving 10k triangles and decide to respend them on chamfered up-close-details or spending an extra 10k on an optic to make it look smooth af, I'll be the first to applaud the optimization.
Good luck and good hunting, or ignore me completely. That's your prerogative
@Thanez Wow! Thanks for the thorough feeback!
100% agree with red and blue. I'll see to optimize it.
Orange, it would not save more than a few dozen triangles, and the front oval insets aren't flat, they have nice a curved bevel in the inner face too, so it would only apply to the stock circle piece.
Purple, I don't get any smoothing problem on the grip (it's a completely flat ngon), and doing this would add more triangles instead of saving.
Green, I was tempted to remove those too when I removed the main grips ones, but from the reference images I noticed that those insets contribute to the stock profile, hence I decided to leave them in sacrificing a bit of geometry. (reference image below)
Regarding the bolt lips, I totally disagree, I made them like they are now not just to avoid interfering with the holes, but because they are one piece with the main body, they even have a nice bevel near the connection that would be lost if I were to make them like you suggest. (reference image below)
Check the lip below the slide release here, it's a floater on the LP but modeled in on the HP.
You're getting that error because you're capturing the entire edge with both the sphere's and the flat's cage.
On the first example i've split up the wall's cage and sphere's cage. Notice how they both encompass the highpoly's edge completely. When bakey this will give you the entire edge on both the wall and the sphere lowpoly, making it render twice on the lowpoly and leaving you with that pinchy effect on the seam.
For optimal results, try manipulating your cage so that you're capturing half the highpoly's edge with the wall's cage, and half the edge with the sphere's cage.
Edit: clarification.
Ok, I understand. Thanks for the explanation! In my experiments I always used the automatic cage that the Substance Painter's baker uses, I'll make sure to try it again as soon as I have a moment.
Anyway, I just finished uving and did the first test bake, and apart from some small errors it seems successful!
I have learned a ton already and I think there's still more before it's done.
Here's the first passage I did this morning in Substance Painter. I can see many problems that have to be addressed. Some are even caused by things you guys mentioned earlier on, so I might be going back to the geometry before going for another texture pass.
I like 25-45.
So you weapon doesnt look so long on the renders.
And here's a quick render of where I left things today. Still lot to do.
I think it is
Also have been playing with different liveries, which one would you guys put first?
https://www.artstation.com/artwork/ZV13m
Fixed a few errors, remade/retopo'd some pieces, made a cleaner bake, and redone textures from scratch.
Total tricount went down to 37k (from the original 42k).
More here: https://www.artstation.com/artwork/88E9R