Sketchbook: wasabimilkshake
Hi. I'm a student, and I'm currently taking a class in environment art using UT3. I'm actually not planning on going into environment art, but my great-grandddaddy didn't lose a leg fighting the Kaiser with his bare hands so I could slack off and make half-assed art just because it's "not my thing." As such, I'd deeply appreciate any criticisms you folks can offer.
Well, then, let me start with the most boring and conventional prop imaginable.
It's a crate. A wooden crate. The top face (the one with the crossbeam) is unique, and the rest overlap in places since it's only possible to see 4 of the side faces at any given time. I've kind of been struggling with UV space vs. unique texture space, because I'm normally all about efficiency, and because I've been baking normal maps from high-poly meshes, which can get messy quickly if you have to deal with symmetry.
Well I'll be damned, it's another wooden crate. I like the diffuse texture for this one, although looking at it ingame it seems like 512x512 isn't quite sufficient to get the same texture density I have in most of the other props I've made.
A word about texturing: I can't paint well, so there's that. But as for photo manipulation: I used to hate creating textures. In fact, I still do because I'm not that good at it. But I've had two eye-opening revelations that have made me more fond of texture creation since starting this project: for one, creating a high-poly mesh and baking normal/occlusion/cavity maps gives me such a leg-up in conveying that detail in the texture that it's absolutely worth the extra time. Secondly, I only recently discovered Photoshop's "Blend If" feature despite having used Photoshop since I was 11 years old. That thing is a huge time saver.
At least this crate's not wooden. Again, not overlapping UVs is killing me. I'm not sure how UT3 handles mirrored UVs as far as flipping normal map channels, but I'm sure something as boxy as this box could be handled without looking terrible. I'm disappointed by how poorly the baked normal map shows through -- it looks almost completely flat. Also, the latticework sections are just a single plane, so I should probably throw in a couple of edges so they're lit better in-game.
Here's a light that would hang from the side of a building. The texture looks a little plain and not dirtied enough.
This one is baby's first ZBrush model. I think it really shows in the smeared-looking sculpt, but to get the final normal map I combine what I've baked with a CrazyBump normal map to add surface detail, so I think my amateur sculpt was covered up fairly well (the end result in this case being that the normal is probably far too noisy). I'm not too fond of the diffuse texture just because it looks way too chewed up. That probably would have looked nicer contrasted with a smooth layer of paint that's been chipped away.
I have more to post soon; hopefully this thread will keep me motivated and help me improve my work. Please offer any critiques that come to mind, and thanks for reading.
Well, then, let me start with the most boring and conventional prop imaginable.
It's a crate. A wooden crate. The top face (the one with the crossbeam) is unique, and the rest overlap in places since it's only possible to see 4 of the side faces at any given time. I've kind of been struggling with UV space vs. unique texture space, because I'm normally all about efficiency, and because I've been baking normal maps from high-poly meshes, which can get messy quickly if you have to deal with symmetry.
Well I'll be damned, it's another wooden crate. I like the diffuse texture for this one, although looking at it ingame it seems like 512x512 isn't quite sufficient to get the same texture density I have in most of the other props I've made.
A word about texturing: I can't paint well, so there's that. But as for photo manipulation: I used to hate creating textures. In fact, I still do because I'm not that good at it. But I've had two eye-opening revelations that have made me more fond of texture creation since starting this project: for one, creating a high-poly mesh and baking normal/occlusion/cavity maps gives me such a leg-up in conveying that detail in the texture that it's absolutely worth the extra time. Secondly, I only recently discovered Photoshop's "Blend If" feature despite having used Photoshop since I was 11 years old. That thing is a huge time saver.
At least this crate's not wooden. Again, not overlapping UVs is killing me. I'm not sure how UT3 handles mirrored UVs as far as flipping normal map channels, but I'm sure something as boxy as this box could be handled without looking terrible. I'm disappointed by how poorly the baked normal map shows through -- it looks almost completely flat. Also, the latticework sections are just a single plane, so I should probably throw in a couple of edges so they're lit better in-game.
Here's a light that would hang from the side of a building. The texture looks a little plain and not dirtied enough.
This one is baby's first ZBrush model. I think it really shows in the smeared-looking sculpt, but to get the final normal map I combine what I've baked with a CrazyBump normal map to add surface detail, so I think my amateur sculpt was covered up fairly well (the end result in this case being that the normal is probably far too noisy). I'm not too fond of the diffuse texture just because it looks way too chewed up. That probably would have looked nicer contrasted with a smooth layer of paint that's been chipped away.
I have more to post soon; hopefully this thread will keep me motivated and help me improve my work. Please offer any critiques that come to mind, and thanks for reading.
Replies
Other than that it's looking good really. Keep in mind though that re-using the same texture-space for certain parts ain't always a bad idea either. With that I'm mainly referring to the first create, one wouldn't really notice if some of the wooden beams used the same texture really.
This also brings me onto the fact that some of your textures are rather big for the kind of prop its for. Once again, re-using parts and cutting down on the texture-size while still keeping the same texel density is usually quite a good idea whenever applicable.
Keep it up!
With regard to extra loops: would you suggest something more like this, to use the first crate as an example?
Making the extra beams on the top and right side separate, floating pieces would allow me to reduce the overall polygon count by quite a bit. This is a no-brainer, right? It doesn't negatively impact performance to use floating geometry, does it?
Looking back at that light, the bulb definitely seems far too dense.
Thanks for the comments.
Not from what I know of. It can sometimes cause problems with shadow casting I think, but in this case it wouldn't be a problem. So ye, floating geometry is deffo a good thing and I use it myself pretty much all the time.
If the vertex lighting turns out to look odd, then some extra segments ain't always a bad thing indeed though. However, to get an even better lighting upon the model I think it'd be wiser tessellating the faces evenly, instead of running a thin strip over a large surface.
And then there's always light maps
Cheers!
Still way too much geometry in places. The pole is 8-sided, but it's probably unnecessary for the globe of the light to be 16-sided, even for UT3. Also, the UVs for this one were maximally unfolded using Maya's smooth UV tool, which results in more uniform, undistorted UVs, but seams become a pain to paint out. They're not hidden well at all in this light (but they're not pictured in the screenshots).
I'm pretty happy with the high-poly, less than thrilled with the texture, and horrified by the the redundancy in the UVs.
I had previously tried this more efficient configuration with mirrored UVs, but I ran into problems when I brought everything into UT3. Basically, there was an incredibly noticeable seam, with each side being lit very differently, and the normal map was not inverted for the mirrored geometry. I'm not sure if it's possible to overcome those problems easily in UT3 (it seems like it should be), but I'd definitely like to figure that out so I can stop wasting UV space for cylindrical objects.
This is something I hope to finish soon, but I had to put it on hold temporarily because it was eating up too much time. I learned a lot about subdivision working on it, but I'm still trying to figure some things out. I keep trying to sharpen edges/corners but having the edges I insert put creases where I don't want them elsewhere on the model. For instance, putting an edge loop on the bottom corner of the windshield puts a crease into the hood, which should be gently sloped.
It seems like the ideal solution in these cases is:
1. If possible, start from a primitive with more initial subdivisions (e.g. cutting into the sides of a cylinder)
2. If possible, reroute edges so that their flow better accommodates your shape
3. Failing those options, separate the area you're working on into a different object
I'm shifting gears now toward building skyscrapers. One thing about doing this for school is that I have to keep moving on, at least for the time being.
I still need to add in some variation with sections that are broken/dirtied. I'm guessing the best way to do that will just be to create an alternate version of my textures and swap out the materials. Changing the models wouldn't be worth the extra overhead since they won't really be viewed very closely.
Today, I kind of got caught up in the details while working on a lower section of the building. I made this palmette ornament in one or two hours. (Still have to make low-poly models for all of these and bake normals, ay ay ay)
I liked what I saw, so I continued to work on detailing this one section:
There's a dilemma here in that adding this detail takes a lot of time. Ultimately, I'm going to end up with a smaller-scale environment than I'd have if I just blitzed through and didn't worry so much about detail. There's certainly a medium to be found between quality and quantity, but as far as I'm concerned I'd much rather have a smaller level with more detail than a big ol' pile of mediocre props. I think the overall visual improvement is more than worth the extra time -- if I had just made a quick diffuse texture and thrown it into CrazyBump, I would have ended up with a far uglier model.
Besides, I think working at it this way fosters improvement far better than does churning out crap. I'm learning a lot as I go, and I can use things like those leaf motifs elsewhere now that they're already modeled.
So where can I improve? A good 80% of my reason for posting this thread is whoring for feedback, so I'd love to hear what you guys think.
I do believe it's time to get painting.
(2 different textures, of course)
And here we have... mirrored UVs! I had quite a time painting out the imperfections around the edges of the arch, but I've yet to touch up the ornaments. It seems like I was just making an uninformed assumption earlier and UT3 can invert normal maps for mirrored UVs perfectly well. Or else it's just not a problem in this instance, whereas the cylindrical column I tried it on earlier was a bit trickier.
There is still one small issue, though:
This seam doesn't show up at all on the front, but on the top and underside of the arch it's very noticeable. It shows up even without a normal map, although applying a normal map does exacerbate the problem. Does anyone know if there's a solution?
To make things clear, I unwrapped one side, mirrored it and merged the geometry, then selected the mirrored side's UVs and offset them all by -1 in U. After that, I exported and baked everything.
I think it's turning out decently, but I'm regretting that I chose to make an exterior environment. I have to have something as presentable as possible in less than two weeks from now, and it's a bit harder to fully flesh out a city scene than an interior, where modular meshes go a bit farther.
I'm trying to work on the sidewalk and street now. I decided to forego hi-poly modeling on this first sidewalk piece, and I'm not liking the results.
Rather than using AO and normal maps baked from a high-poly model, I relied on Crazybump to punch up my diffuse texture, and I don't like how it looks. The bricks are too pink, the concrete too grey, and the whole thing too light. It doesn't match the buildings or anything else in the scene. The Crazybump occlusion makes it too contrasty, with overly dark, spotty cavities. And 1024x1024 seems too small for a model of that size, at least compared to the texture scale I've been using elsewhere in the scene.
I think I should scrap this and try again, this time modeling a passable high-poly and perhaps taking a different approach, making the bricks a continuous BSP brush and laying individual concrete slabs on top.
The pressure is on and I want to have as much of the scene finished as I can, but this sidewalk fails on too many levels for me to consider "screw high poly, quantity over quality" a viable strategy. I (humbly) think there's an appreciable jump in quality between this project and what I was doing a few months ago, and I owe 90% of that improvement to high-poly modeling, time-consuming as it may be.
This new version took 6 to 8 hours to the original's two, but I don't think there's any doubt that it was worth the time.