Courtyard Environment WIP (UDK)
polycounter lvl 6
This is just the first piece and the central/salient feature of the Courtyard itself.
The gazebo is a whopping 22470 tris which does not, btw, include the creeping vines or the candles. Seriously though, I don't really think that I can get it down further without skimping on the chamfers needed to ensure that the normal map looks halfway decent.
Maps are 1024 Diffuse, Spec/Gloss, Normal/Height (for Parralax) and a 512x512 detail normal map.
The gazebo is a whopping 22470 tris which does not, btw, include the creeping vines or the candles. Seriously though, I don't really think that I can get it down further without skimping on the chamfers needed to ensure that the normal map looks halfway decent.
Maps are 1024 Diffuse, Spec/Gloss, Normal/Height (for Parralax) and a 512x512 detail normal map.
Replies
You might want to try a some better lighting so we can see more details.
To heck with it. I'll take the poly-bloat this time and chalk the matter up to a learning experience.
pf
Oh, you mean by using the little edit button?
(laughs)
I'm blind, sir. I'm old and cranky and blind.
pf
It looks okay, but deffo a few probs:
The brick section near the top looks too low-res, and the bricks really don't fit their sections, the ultimate approach really is to model each brick seperately and then sculpt them, the normals would show MUCH better, and you'd have an easier time creating the diffuse from the norms and AO bake.
There also seem like excellent areas for sharing texture space, like the fence-like areas at the bottom, or alternatively, sculpt two variations and alternate them, it would certainly make it look more high res. You could also do this with the pillars (just rotate each one 90 degrees as it goes around...
Where exactly are you using parallax on this model, as it doesn't seem evident.
The copper cups look okay, but maybe a bit too mossy.
It does look good in its entirety and deffo wouldn't look bad when in the final environment, keep pushing!
In the future, I'll seriously consider your advice, m1neh and sculpt individual bricks in Zbrush.
pf
Also i can see plenty of places to save on polys, but it is dependent on the viewing angle, that sphere on top for instance, if the player is only seeing this from below, the only time he will see the top is when he is afar, so you can shave some polys off of that for starters.
Its more dense than the dome it sits on, even.
So you're using the vast majority of your tris on those intricate patterns at the bottom, which do look nice, but you could very easily shave quite a few off without much change at all...
You also have way too many chamfers in the lower part generally, especially on the chalices! Could ofc get rid of that edge loop that goes round the middle of the columns, and the tiny chamfers that follow the edge loop of the little protrusions at the top of each pillar won't be noticeable in the slightest.
Top sphere has too many tris, try creating it as its own mesh, and create a geosphere and choose the different types, think the middle choice is the best (icosa maybe?), i forget the names. It is far more optimized.
Could also probably bring down the amount of sides you have on the columns in the middle.
If I were you I really would go in and take out a bunch more before moving on, must be able to get quite a bunch off that 22k count in about 20 mins of deleting loops
The gazebo will be able to be viewed from on top. To bad for me I guess.
m1neh,
If I delete edge chamfers, won't that make my normal maps struggle and do a shitty job with sharp angles? I've read in lots of places that normal maps really hate angles approaching or greater than 90 degrees.
pf
you don't need any of the chamfering (Chamfer? Bevel.) Beveling on the battlements up top, or the bricks that stick out of the battlements. It doesn't look like there's any kind of rounding on those edges, so the extra geometry is wasted.
But yeah, it looks like the majority of your geometry is going into those detailed shapes in the fence on the bottom. You should definitely be able to skim some geo out of those.
Thanks for your crit and recommendation...I will reiterate however: I've heard that normal maps don't like sharp angles. If I get rid of the chamfering, then I'm creating right angles (or even acute angles in a couple of places) in a lot of places which I'm not going to do until somebody can come in here and tell me why it's okay.
No offense sir, honest.
You can also do this for the rampart-style wall piece or any part with repeating geometry - try cutting it into 4 or 8 equal pieces and overlaying them so that the texture repeats.
Optimise this smartly and you could probably scale down to 512x512 and still have greater texel density than you do now.
Make hard edges on the model wherever you have a split in the UV map - It's good practice and gets rid of (or at least reduces) normal maps seams.
On a side note, do you have much padding? For a 1024 map I would use 8 or 16 pixels.
...lots of overlap. No mirroring, but lots of overlap.
Only 4 pix padding (roughly).
Per rendered frame, is it better for the engine to call fewer but larger textures or smaller textures but more of 'em...The smaller textures would lay out and blend etc. with masks an extra uv channel or something?
pf
If someone's got some additional info or correction, I'm always eager to learn about this as well. It's hard to find good input about this kind of stuff.
That could assist you particularly in the fence where you have those complex designs. It looks like you've unwrapped and mapped each one of those diamond designs individually. Much more economical to only create and unwrap one, then duplicate it as needed.
As for the turrets on the top, normal maps don't like right angles at all, but the only time I have problems with them is when I'm asking it to bake a rounded corner onto a hard right angle.
If all you're doing is displaying stone grain and seams on the flat faces, there's no need to bevel the edges.
If you've tried to round off the edges at the top with normal mapping, I don't think it's working. The turrets in particular, and the outcroppings to a lesser extent, don't _look_ rounded at the edges. They appear to be simple hard-edge shapes. Professional opinion, wasted geometry.
Especially in a game engine where geometry count is a consideration, you need to pick and choose what you bevel, because that adds a _lot_ of geometry. Priority for edge-beveling goes to features that a player will see clearly and spend a lot of time looking at. Which generally means, the higher up a structure you go, the less need there is for heavy bevel work.
Three are unique, the rest are overlapped and dupes of the three unique ones.
*
Thanks a whole lot for the answers, wright.tom85!
pf