You play as a ghost that is evading ghost hunters. Phase through walls and gather your belongings so you can pass on to the next life.
Planning out realistic room layouts that create interesting emergent gameplay.
Assembling a realistic house layout with non-linear design principles.
Planning enemy patrol pathing to create memorable scenarios.
Playtesting. A lot of playtesting.
Initial pitch for house layout. Each room marked with a star would have a linear challenge.
This project required me to work in a design space I was unfamiliar with - non-linear stealth design.
The first draft I pushed had a hub and spoke design for linear segments.
However, the team unanimously pushed for a fully non-linear experience. No linear challenges, it needed to be organic. So, I studied stealth games, like Hitman: World of Assassination, to learn how to do this.
In interviews, the Hitman series developers describe their levels as “[...] spirals, or a snail house, we call it.” This means that there needs to be a main, orbiting path that comes inward, as well as shortcuts in and out of that line.
I applied this principle on a micro and macro scale. All spaces should have a main path with sideways to sneak around, on all scales.
There is a main path that snakes around the house, but the player should be encouraged to sneak between rooms.
There is a main path that snakes around the room, but the player should be encouraged to sneak between cover.
During preproduction, I started with rooms before the house.
I crafted spaces that felt unique from each other at this scale, then worked outward. This allowed environments to start doing blockout and concept art while I arranged the structure. For a three month long project, preventing blockers this way is crucial.
I would then assemble these rooms into a house in Alpha.
Vague enough for artist interpretation, while giving a clear spacial layout of the room. (At this development stage, we had traps like an oven or a banana peel. This was cut in Alpha.)
Notice how it's good and not bad?
I made these quickly in MS Paint, as the artistic direction hadn't been settled yet, and it lets me put emphasis on spacial layout and level design.
I got the idea from the Undertale artbook: sending low fidelity artwork to the concept artist and working from there.
In Alpha, development moves rapidly as ideas get fleshed out.
This was my first time on a project with this team size (10 people) and this long (3 months). As team size grows, individual responsibilities shrink, but interdependencies grow. Being level designer in Alpha on a team of this scale puts you into an interesting position.
Systems needs a space to work in locked down, and Environment Art needs some freedom to experiment. How do you satisfy both?
Prototype the house in a rapid way for the systems team, then prepare the actual house for artist's convenience.
Level Streaming is a clean way to collaborate in locking repro systems like Perforce. Having two versions of the same level, one as a playground and one for the artists to flesh out, is similar in concept.
The test level didn't need proper level streaming, so I used the Cube Grid to make one solid mesh. Then, I threw in basic geometric forms with text labels to represent furniture.
In the proper version, each room is a sublevel.
First, I worked out the level in the abstract. How should rooms connect to each other? Shuffling the rooms around like this reminded me of when I studied architecture.
In houses throughout the ages, traditionally the house is split between private and shared spaces. So, having the main shared spaces be hubs, then putting the living and personal spaces away from those helped make the space more believable.
After working that out, I turned that abstract form into a drawing, and that drawing into this general layout map. Over time, the rooms would shuffle around.
In this phase, we moved from the prototype house to the final house.
Now, we have working enemies, beautiful looking assets and proper gameplay.
I used Unreal's Level Streaming to work around the restrictions of TortoiseSVN and enable the rest of my team.
Unreal's Level Streaming is usually used to load in and out segments of levels, like how Source uses visleafs, except manually and with intention. What this means is that I could turn rooms into sub-levels, so that artists can check those out and work on them.
Now, our team of 3 artists don't have to wait their turn to check out the level, and can instead all work simultaneously.
The two house approach let the "real" house serve as a second pass for room and furniture scale.
I hadn't learned how to properly scale stuff. My artists handled fixing the furniture proportionally to the player, but the room scale was off.
I will say that there was a miscommunication that lead to me believing that the house was a mansion until this point. Thicker walls and tall ceilings representing a larger than life sensation.
During playtesting, some people pointed out how unrealistic the bathroom was in terms of layout. I live in a very strange house, where there was a bathroom with two doors on it that I often used as a shortcut in my home. I recreated this in game, using the bathroom as a hallway. (I'm unsure why nobody called me out on it, maybe because the original version of the house was a manor so eccentricities were overlooked?) So I removed a door and redesigned it to be a hiding place, according to genre conventions.
At this final milestone, on my 10 person team, I shifted into a playtesting role.
As level designer, I did this both to make minor, last minute refinements and to support the team. These refinements included starting the player facing the shrine, so they knew to take the items there, and making terrain easier to navigate.
Here's the finished product from a top down perspective.