Monday, August 12, 2013

Turncoat Dev Diary: Experiments in Environment Creation


Today, I'm going to talk about a little side quest I took while prototyping the game: painting and lighting one of the rooms. When I initially planned the level out in Blender, I did it almost entirely with cubes. I scaled and extruded and adjusted vertices, of course, but when you come down to it, it really was just a bunch of boxes. Of course, a lot of buildings in real life are just assemblages of boxes if you look at them from far enough away, but to try and make an environment that feels real, I knew I needed more than just boxes.



I selected one of the smaller room to paint up - a sort of locker / shower room part of the intake processing area. I chose it because it's small and because I should be able to texture it quickly using readily available stock textures. While the game is going to be set in the future, I want to base the look of it on the real world, so I started looking for reference images as well as tileable images suited to how I was picturing this room in my head. It's funny the kind of little details you don't really notice until you try to recreate something, though. In most cases, for example, doors just don't look right without some kind of molding or frame. It took me surprisingly long to figure out what it was about the doors in my prototype that bothered me. Just cutting a hole in a wall is perfectly functional, but it doesn't feel right because in the real world, doors almost always have a visible frame around them.

Another thing I noticed is that when you start looking closely, even at a relatively clean environment, there are marks, scuffs, and smudges on nearly everything. They're often subtle, but they're almost always there. Now, a lot of that subtle smudge detail in a normal room would fall below the fidelity of the medium, but in this case, I want to exaggerate it and make everything extra grimy. I want to convey not just that this is a depressing, unfortunate place to be, but I almost want a sense of neglect and even foreboding. I want it to feel like the people running the facility don't care about this particular place or the people it contains. Even before props or characters are added or dialogue is recorded, I want this level to convey a sense of desolation. Things are dark, dirty, and just generally unpleasant. This is an out-and-out bad place; escape is imperative.

After spending a few hours with Google's image search looking for inspiration and reference, I started adding some details to the cubes that made up the room I had chosen to paint. I cut out a window between the two rooms, added moldings to the doors and window. I cut out drains in the floor of the shower area and cut vents in the wall. I added planes to hold some signage and also for the ceiling lights.
I really don't know how much of this detail will survive through to the final version of the game, and it really doesn't matter. The goal here is to explore the process and to try and get a feel for whether I need to bring in a dedicated environment artist. If the actual work gets thrown out, that's okay. That's the nature of prototyping.

Once I had the details modeled, I started mapping textures from CGTextures, the Blender Texture CD, and a few other source onto the room's surfaces. I did this to give me a starting point. It's easier to work from something rather than painting on a completely blank canvas. 

Once I had all the surfaces mapped to textures, I "baked" them to a fresh texture map. Baking the textures to a single image will allow me to paint right on the 3D model directly using Blender's paint tools. Here is a reduced size version of the baked map for the shower room (left), along with how the room looks in paint mode in Blender (right):


It's certainly more realistic than my earlier white-walled prototype. Firing up Unity, this is what it looks like now:


It's definitely better, but it still falls way short of feeling real and it's way too clean and bright. On top of that, the current lighting just makes no sense. There's no visible light source, but the room is lit as if there was a light floating in the middle of the room. I'll fix the lighting later, but first, I want to work on the textures.

I went back to blender and started to filthy the place up. The grime will help make the room feel decrepit and it will also break up the repeating patterns from the tiled images I used, making it harder for the eye to pull them out. I loaded up an assortment of dirt, grime, and grunge images to use as brushes and started painting on the room's textures. It's a little weird painting on the inside of an object. Blender's Texture Paint mode works like Unity in that you can only see the "front" of polygons which, in this case, means the inside. Like one of those optical illusions, my eye wanted to think the texture was on the outside rather than the inside, but after a few minutes, I adjusted and work started going faster.

Doing something like this really makes you appreciate the power of undo.  I hit ⌘Z a lot while working on this, and it took me a long time to find an approach that gave results I liked. 

Once I had the new details and grime added to the texture, I pulled the new maps over to Unity to try them out. I also tweaked the shaders a bit, adding a distortion map to the window and creating normal maps for the tiles to give them a tiny bit of depth.


It's definitely an improvement. I might've been a little heavy handed with the grime, but I want to wait until I've done some work on the lighting to decide whether I should pull back on it. I envision this area being much darker and poorly lit than it is now, so I want to judge the texture in the correct lighting. 

In addition to being darker, I also want some of the lights to be flickering or burnt out. A little googling around found me a great light flickering script that gives just the effect I want. In Unity, I started placing lights around the room. Each of the two sections of the room has four light fixtures, but I decided that each section would have one light that was burnt out. The shower room would additionally have a flickering light and an an illuminated exit sign.

In order to keep performance reasonable with so many lights, I decided to bake light maps for the lights except for the one that flickers. "Baking" a light map essentially creates a texture map to store pre-calculated lighting information. They let you do lighting effects that are normally too processor intensive for real time calculation. The downside is that baked lights don't cast shadows on dynamic objects (like the player, moveable props, or bad guys). For a desktop computer or console version of the game, we'll probably want to make more of the lights real-time to get realistic shadows, but for mobile, we really have to be careful to limit the number of dynamic lights to keep the framerate up.

The Beast Lightmapper that's built into Unity is quite good and not that hard to learn, but calculating lightmaps is a fairly processor intensive task. Getting all the lights configured the way I wanted ended up taking a long time. I'd tweak one or two settings, then re-bake and wait five minutes or so. Lather, rinse, repeat.

After a while, I got the light maps baked mostly to my satisfaction. I started liking the overall look much better. The grime, which was kind of overpowering in the brightly lit room, feels just about right now. It still could use some furniture and other props, and the overhead lights could use some more detail, but on the whole, it's not terrible.


I wanted the shower room to be even less well-lit than the outer room, possibly to be used as a hiding place in the final game. This is how it looks after re-doing the lighting and baking the light maps:


It's not perfect by any stretch of the imagination, but the overall effect is roughly in line with what I want to achieve for the final game, and it only took me two days of experimenting to get to this result.

Here's a short video if you want to see it in motion.


At this point, I've done enough to believe we can do the environment work in-house. I understand the technical process well enough, so if I team up with one of our designers who has a better aesthetic sense than I do, I think I can get an environment that looks the way I want it to look.

Well, enough of this side quest. I think it's time to return to prototyping. Before that, however, I want to talk a little about the game's characters.

No comments:

Post a Comment