Official 3D art thread - Warband

[xenoargh]

Interestingly, we actually can use normalmap compression schemas on this engine to reduce the compression artifacts greatly. 

But it's never ever worth it vs. good ol' BC1 (DXT1), except maybe if you were doing something insane, like a million-tri thing you baked down to a smallish statuette or something just plain nuts like that    The reason?  Size in memory, unfortunately; it takes twice as much RAM to store BC5 NM vs. BC1 color encodings. 

It's a real pity the engine loads textures into its main memory like it does and hasn't ever gotten compiled for 64-bit so that RAM isn't a huge limiting factor; it's kinda fun, though, using the engine to learn more about optimizing assets properly

For a sword like this, the handle's usually the last thing people look at, fortunately.

 

[xenoargh]

...and, done.  Couldn't resist the reference to another angel's blade

512 maps by me, geometry by Dthehun.

Spoiler: Breakdown

1.  The model's a lot better than most weapons I've seen here.  Really!

The thing that's mainly letting it down is the texture; it has good ideas that just needed some polish, is all.

2.  Some polycount reduction was possible early, especially in the blade itself.  Even without the hard edges on the fuller (which could've just been done in the normalmap) there were a lot of faces that could be lost without altering the sword's main profile.  Polycount came in at 516; that's a 40% reduction.  If I'd done the fuller in the normalmap, that'd save me about 30 triangles.  There were excess triangles on the tips of the crossguards that were eliminated, and I rebuilt some triangles in a few places to correct the mesh flow.  I tried to leave the design alone, though (other than my little joke referencing Angel Blade).

3.  I'm mainly showing it with all of these hard edges here to illustrate just how much difference we can get, with the same geometry, by emphasizing different zones.  It took the weapon's look harder and it makes the deep fuller really pop.  A soft, curved fuller just needs to be in the normalmap.

4.  The main things I saw, when I looked at your public version of this sword, was that the jewel doesn't pop, because it's white against a light background- it simply doesn't have enough contrast.  The issues with the handle and the crossguard are purely about giving those areas enough space for details.  I could've fitted this into a 512X256 pretty easily if I was willing to lose a little more resolution on the blade and do a little more mirroring to save room on the UV map.  As you can see, that gave me enough room to put quite a lot of detail into the handle.

5.  If you want to see the mesh, textures, PSD's, whatever, PM me

 

[DtheHun]

...and, done.  Couldn't resist the reference to another angel's blade

Spoiler

512 maps by me, geometry by Dthehun.

Spoiler: Breakdown

1.  The model's a lot better than most weapons I've seen here.  Really!

The thing that's mainly letting it down is the texture; it has good ideas that just needed some polish, is all.

2.  Some polycount reduction was possible early, especially in the blade itself.  Even without the hard edges on the fuller (which could've just been done in the normalmap) there were a lot of faces that could be lost without altering the sword's main profile.  Polycount came in at 516; that's a 40% reduction.  If I'd done the fuller in the normalmap, that'd save me about 30 triangles.  There were excess triangles on the tips of the crossguards that were eliminated, and I rebuilt some triangles in a few places to correct the mesh flow.  I tried to leave the design alone, though (other than my little joke referencing Angel Blade).

3.  I'm mainly showing it with all of these hard edges here to illustrate just how much difference we can get, with the same geometry, by emphasizing different zones.  It took the weapon's look harder and it makes the deep fuller really pop.  A soft, curved fuller just needs to be in the normalmap.

4.  The main things I saw, when I looked at your public version of this sword, was that the jewel doesn't pop, because it's white against a light background- it simply doesn't have enough contrast.  The issues with the handle and the crossguard are purely about giving those areas enough space for details.  I could've fitted this into a 512X256 pretty easily if I was willing to lose a little more resolution on the blade and do a little more mirroring to save room on the UV map.  As you can see, that gave me enough room to put quite a lot of detail into the handle.

5.  If you want to see the mesh, textures, PSD's, whatever, PM me

I was offline just for a month - they've said it will be hard - now I'm coming back to see a ghost retexturing my stuff, and he makes it pretty good. I'm really happy that you're back, Xenoargh! A new armor idea popped up from the Angle's Blade reference. 

 

[смертник]

Hi everybody
Air rifle ,high poly
WIP

Spoiler

 

[Pumpkin Lord]

Hi everybody
Air rifle ,high poly
WIP

Spoiler

Geez. Do you have artstation man?

 

[смертник]

artstation ?

yes https://www.artstation.com/otk

 

[NPC99]

artstation ?

yes https://www.artstation.com/otk

Awesome models. 

 

[DtheHun]

Spoiler

Hi everybody
Air rifle ,high poly
WIP

Spoiler

Professional! 
Not that I understand half of the functionalities, but what's the role of the bicycle pump?

 

[смертник]

DtheHun
to create cylinder pressure

 

[DtheHun]

DtheHun
to create cylinder pressure

Oh, I see. 

 

[xenoargh]

Weird experimental thing:  can FPS armors for a fantasy setting be done with 512 skins and look all right by modern standards?

I thought it all worked out pretty well, but the cloth disappointed me a bit; this is where the lack of pixels showed up the most. 

Some of this was certainly my fault.  I planned the UV map pretty poorly, because I thought the problem would be elsewhere... and just didn't have enough resolution to make it work as well as it might have.  But it got closer than I thought would really be possible, so it's a good kind of failure; I'm pretty sure I could've pulled it off with a 512/1024.

Anyhow, I know this was an odd thing to try, but I had a lot of fun getting myself ready to do this kind of stuff at speed again.

 

[Make My Day]

Hey guys, I've created game ready amber ring. Here it is.

https://www.artstation.com/artwork/nQQm11

what do u think bout it?

 

[xenoargh]

Well, that's not bad, in terms of realism

Spoiler: If you're aiming for game-ready, here's some critique:

It's quite a realistic ring; poly-count-wise, it's almost as high as meshes people were using to 3D print wax molds for real rings a few years ago.  Unfortunately, if we're talking real-time assets, that isn't what you should be aiming for.

1.  The model's presented as quad-dominant, rather than triangles.  Game-ready means triangulated.
2.  There are dozens of places where multiple quads could be converted to two triangles, with no noticeable loss in fidelity, presuming the normalmap is built well.  I'd suggest rebuilding the topology.
3.  The polycount is quite high for an object that most likely will be barely visible in-game; you should be aiming for maybe 100-200 triangles for a ring or other small bauble, just enough to convey the central shapes.  If this was a really special object meant to be viewed up close, the polycount would still be higher than ideal; see 2.
4.  I'm not sure why there are two surfaces being used for the amber; I presume that's meant to enhance the optical behaviors of the amber in ArtStation's PBO setup.  This is usually not how to do that; usually this sort of thing is done through a specialized shader.  It also presumes that the game engine you'd deploy in would use rendering-order tricks to render in the way you're expecting.  This frequently isn't really practical.
5.  If you're looking for work, posting tricount, textures used and sizes is a good thing.  I find it pretty annoying that ArtStation doesn't post that data by default.  Based on the UV map scales, though, I suspect you're using a huge texture for this thing; an object that size should use maybe a 256X256, perhaps a 512 if, again, it's going to get viewed at point-blank like it is in ArtStation.

For everybody wondering if I'm overstating the polycount / topo stuff, here's some wires:

Every single one of those squares-sub-divided-into-weird-quads could be converted back into single quads, reducing the polycount to 1/8th without affecting fidelity (each one of those zones is 16 tris, where we only need 2 to have a reasonable normal).  If I was working on this piece, I'd take it into Wings and manually repair the topology; that won't take long and it'll be a good learning experience.

 

[DtheHun]

Looks good. The flower setting sinks too deep into the white gold ring part.

 

[Make My Day]

Thank you very much for advices guys, that means a lot to me. Gonna work a little bit more with that ring!

 

[xenoargh]

Keep at it!  The overall effect was really nice, my specific technical critique aside

 

[Slytacular]

Hey guys, does this look human enough? I'm trying to go for a default human character that is lean and toned with an unbalanced physique. Pretty much going to use this model to dress up in Early Medieval style clothing and armour.

https://pasteboard.co/HLMMAxs.png
https://pasteboard.co/HLMML2I.png
https://pasteboard.co/HLMMWeq.png

 

[Den.Di.]

@Slytacular

https://prnt.sc/lew08l

 

[Adorno]

The arms look a bit short, or it's because the "upper" arms (humerus) are so much bigger than the forearms.

 

[DtheHun]

Super human. The length proportions are good. And yes, calves should not bulge forward beneath the knees. If the clothes won't cover those upper arms, then muscle tones would be appreciated, because now they looks fat compared the flat torso. If you go for a gymn body consciously, make it somehow believable. So, I think it's feasible for dressing. Have fun!