The Tip of the Iceberg: Storytelling and Reactive Design in Canabalt and Capsule

Author: by James Dilks

Adam Saltsman's day job is running Semi Secret Software, the self-publishing company he founded with Eric Johnson in 2008. However, he's probably better known for the games he's managed to create in his spare time, many of which have been published on mobile devices through Semi Secret. These include Wurdle, Canabalt, and Gravity Hook.

When Saltsman released the flash version of Canabalt on an unsuspecting world in 2009, he had no idea that it would blow up as it did. At the time, he wasn't even sure why it did. Has he figured anything out since? You'll have to read on.

I meet the game designer in a tent at the GameCity festival in Nottingam, England. It's wet and cold, but rather than complain about the flip-flops on his feet, or the effects of late nights and jetlag on his back, Saltsman is eager to talk about the festival's site-specific version of Capsule (which was originally made for backers of the Venus Patrol Kickstarter, and now available for its subscribers), dubbed Capsule Capsule.

The game begins when the player is led through the city's streets by a "scientist", listening to instructions on the control of a spacecraft. Then, they're enclosed in a tiny, pitch-black room with just the radar monitor of Capsule's game screen and a control panel in front of them, and the sounds of space swirling in their headphones.

Let's start with Capsule Capsule: why the repetition?

Adam Saltsman: That was really more [GameCity] than me. Iain [Simons, festival director] likes to call it "the slowest theme park ride ever". It's a game about an enclosed space, so they actually engineered an enclosed space to play the game in to make it extra uncomfortable, which is exactly what that game is about.

Capsule Capsule

Was that the original idea?

AS: It came from a couple of different places. One of the big inspirations was Lunar Lander, which is a game that I just adore. It has a great physics model and a really great sense of scale. It has a really nice intrinsic sense of challenge, difficulty and accomplishment. It's kind of self-directed: you can choose where you want to land and what risks you want to take. And it also has a thing which I decided I wanted to exaggerate, which is that [after a lot of work] you can just run out of fuel. There are consequences; your lunar lander plummets to the surface of the moon and you ostensibly die a fairly horrible death. That's the game; it's classic arcade fun!

To me, it feels really different to Space Invaders. They're all dire scenarios, but Lunar Lander doesn't have catchy music or the [imitates Space Invader's SFX]. Space Invaders doesn't feel lonely to me. Lunar Lander doesn't even have enemies; the only enemy is the horrible terrain and the crappy landing pads. They have a cabinet at the Exploratorium in San Francisco and playing that for the first time in years [I thought], "How do I not own this machine?"

The controls for it -- nobody would design an arcade machine like this now -- the thrust controls are tiny little push buttons, they look like the kind of thing you would have in a spaceship, like a crappy switch that you're supposed to thumb a little while you look at some meter. Then the thrust meter is this big, clunky thing and the contrast adds to this cool, lonely authentic kind of [feeling]. I wanted to take that and really exacerbate it. What if when you ran out of fuel you didn't crash and die? You just landed and sat there, and now you can't go anywhere. You're stuck.

Lots of the games I make are just the makeable versions of more ambitious ideas. Canabalt, for example, was inspired by Super Mario 1 and Sonic the Hedgehog but the game I really wanted to make was a game where you're a guy running through a city being chased by water. It's like Escape from New York. There's a horrible flood, and some particle physics, [you're] surfing on flotsam and jumping up and parkouring off buildings and rooftops. But it would have been so hard, and especially back then, I had absolutely no patience for projects of that scope. So the slimmed down version of that is Canabalt. I had to take out the water, take out climbing up ledges, and focus on something else.

I always wanted to make a game -- that was not even a game. The game would just be: you're an astronaut on a space walk and your tether breaks. You're doing an EVA [extravehicular activity] and now you're floating through space, and that's it. There's nothing to do. You would be in first-person view, you would see gloves, and the astronaut control box on your chest, and you can look around. It would just be that until you ran out of oxygen. For like four hours. It would be the worst game, but a little bit of that ended up expressing itself in Capsule. You're out of fuel, and the consequences of that are not that you die, but that you have to listen to yourself die, which is horrifying and uncomfortable.

At one point I found my breathing syncing up with the breathing in my headphones.

AS: We talked about that a lot. When Robin Arnott was doing the sounds for the original version and for this festival version, we had a lot of discussion about whether or not we should put them in, whether it would be too corny. Some people think it sounds like a dude who's getting it on, rather than being legitimately terrified. To me, it makes it sound like an enclosed space; you never hear somebody breathing when it's not reverberating off a helmet.

It's realistic, and it's also kind of a gimmick, but it's actually communicating something about the game's environment that we can't communicate, because the game is the monitor. You're supposed to pretend you're in this capsule, and the sound is something that helps you do that. The breathing tells you something about the size of the space you're in.

How much of the sound was in the game in the Venus Patrol version?

AS: Most of the game was done before I knew Robin was going to come and work on it; I just hadn't figured out how I was going to do all the sounds. Initially, it was a prototype. I liked the verisimilitude of viewing your progress through some kind of radar, so that everything in the world is going to be fed back to you through this one interface, this one believable piece of technology. It's going to have your fuel, your air, your speed, your direction and the types of things you can run into. It was about a single idea that I wanted to test.

In Canabalt the buildings are generated based on the speed that you're running, and each building is generated one at a time. So however fast you're running now is the input that is used to work out what is an interesting obstacle for you to encounter next. It always makes sure that jumps are achievable, but have an interesting level of difficulty.

For me, that's interesting because it allows you to hit a box and slow down, and it doesn't ruin the whole game, because it instantly reacts to the change in your speed.

There was a principle under there that I thought was really interesting: this is a cool thing you can do with a procedurally generated game that's not just to avoid doing level design.

If everything's procedural, it can be reactive. You can have a game environment that responds to the choices that players make. Dynamic difficulty adjustment is sort of an expression of that in big triple-A games, but it's the most boring version. The Left 4 Dead director AI is a pretty good example.

A lot of Capsule's procedurally generated world is actually really Left 4 Dead-inspired. It's actually monitoring how you're doing and trying to create scenarios that are emotionally potent. If you have lots of oxygen and health it's going to start inventing more challenging things for you. As your oxygen and fuel dwindle, it's going to be much more likely that you'll just barely stumble across some small form of sustenance. All that stuff is actually programmed in, which in one way is a horrible lie, but in another way, how else would you do this except by having a scripted level with set distances, which couldn't be replayed?

It's like telling any story that involves a last-minute save.

AS: Yes.

And also, the player might miss that bit of oxygen.

AS: Yeah, and especially if your power's low and you're afraid of pulsing [which reveals objects], because that uses up power. So it doesn't always save the day, but it increases the likelihood that there will be a really exciting scenario. There is a lot of programming in that. To me, it seems to work, and systemically that's one of the core things that Capsule is supposed to do. [The] world generation stuff is more complex than Canabalt but still not as complex as it would have to be to happen in a Resident Evil game.

Why did you decide to give the player an objective, rather than letting them just explore?

AS: Part of it was to give a purpose to the cool looking compass things at the bottom. Part of it was to feed more atmosphere back in. In the festival version, the story gets into the nitty-gritty really fast and it doesn't explain very much.

When you play it at home it's a lot more of a slow-burn type of thing where at first you know nothing -- there's no narrative you listen to on the way to the game [as there is in the festival version] -- you just start it up, and the first part of the game is you figuring out what is going on with the controls and the gameplay, but also with the story. So when we compressed it for the festival we took a lot of that out and rearranged the story.

The only way people can play the game at home is if they subscribe to Venus Patrol. Did you know this when you were making the game?

AS: Originally, I thought people were only going to be able to get it by participating in the Venus Patrol Kickstarter. It was fine because the game didn't take that much work, and I'd already done most of the systemic prototyping.

But did it affect how you made it? That it was only for this small audience?

AS: I was way more willing to make it weirder. We could have made it so that you would use oxygen at a similar rate that you burn fuel, as opposed to being slower like it is now, to ensure that if you ran out of juice you would die of oxygen loss almost immediately afterwards. [But I was thinking] this is going to this weird, picky audience. It won't freak them out if we do things differently than some other games do. You don't consume air particularly quickly and everything is set up to make it much more likely that you'll just run out of fuel and just be stuck there, and just have to deal with it. No games make you do that. It's such a horrible thing to do to a player.

Super Meat Boy: you make a mistake, it's fine, go back to the beginning of the level and try it again. In Capsule it's less about punishment, it's more part of the cycle. Part of the fun -- if Capsule's fun -- is, "Oh God, what happened? I don't want to listen to myself asphyxiate. This is making me uncomfortable." That's the "pleasure" of it.

Can you restart the engines? I tried, but couldn't get it to work.

AS: If you run out of fuel, there is a way to jumpstart. Especially in the full version of the game, it ends up being a pretty crucial part of surviving. When people figure out how to do it -- because the game never tells you, steps have been taken to make it discoverable, and not by a clue you find. You'll run out of fuel again, and you'll notice something is slightly different. I wanted it to be like: you're an astronaut -- or a deep-sea diver, it doesn't matter -- and you run out of fuel, and it's actually a problem in a minute. It's not a problem right now, but in a minute it's going to be a big problem, and you have that wiggle room.

I thought about doing something like that and then decided against it, but Robin really, really wanted it to be in there. He made all the sounds for it; then I felt guilty that I had the sounds for a mechanic that wasn't supported, so I put it in.

That was the hardest thing to tune, and to put in the little hints in exactly the right ways so that people had to discover it themselves. When that thing turns back on for the first time because of something you did, it was actually exciting: you just prevented yourself from having to sit there and listen to yourself choke to death. It's a really meaningful thing that you've just avoided.

The game has checkpoints. Did you consider making it so when you died you had to start all over again?

AS: I thought about it, but I wanted to have this element where it would tell some stories. The fancy way would be to call it a posthumous epistolary narrative. You're finding these traces left behind and it's a little bit sci-fi gothic.

Everyone talks about the way BioShock did it, but I like the way Doom 3 did it better, actually: in the first hour of Doom 3, there are no bad guys. It's awesome! It's just you and a flashlight and some audio logs. You're in these dark tunnels but there's nothing there; it's just [id] showing off their lighting engine, but in the process of doing that they actually made something I really, really liked. You're in a coherent space and the light is acting in a really believable way, and you're listening to these spooky stories.

In Capsule, if you could drive around and just listen to audio logs for the whole game, that would be awesome. We could have done that, but it turned out that we wanted to use a voice in a specific way later in the game. There's something about reading something versus hearing someone talking, and if you know that person is dead, that is more lonely. We wanted to find these abandoned places where the only traces of human existence are notes. Lots of games do this. Finding notes that are creepy is a classic horror game thing to do.

Canabalt

Do you think games can move on from using audio logs, letters, and graffiti to tell stories?

AS: It's iceberg theory, right? Hemingway's name for giving people the best 10 percent and they'll just imagine the rest better than you could write it anyway. I think a lot of that's true. That's what a lot of Canabalt's art was all about: I just wanted to show the most interesting thing that I think there is to show about Canabalt's world. It's everything they need to know. I don't care what version of the story they imagine for the rest, because it's all going to be within the same ballpark that creates the right emotional palette.

Isn't that part of the game as well? Imagining the story?

AS: Yeah. I wanted people to play an arcade game a couple of time, say, "I'm bad at this," turn it off, and then later that day, start wondering about what was going on. Or sit down and play it tomorrow and notice there are giant robots in the background. Suddenly it's a slightly different setting than you thought it was.

I think it's good, but I think we will move past it. There are things moving past it that are doing totally amazing things with storytelling and systematizing storytelling mechanics in a way that's really meaningful. And even if it's not happening with sentences written out, there will be more complexity. It doesn't necessarily need to be artificial intelligence or anything like that.

That's the dream, isn't it?

AS: Yeah, and I think we're beginning to see games that are starting to prove out that idea. It's not like a Holodeck that can tell any story, but games that are a very narrow situation. [Interactive fiction author] Emily Short is doing this in a couple of different projects. It's about taking a very narrow slice of a human story and exploring that.

Now that you don't have to simulate the entire range of human behavior, now that you have some constraints, you can revisit the idea of whether these characters have motivations and whether they form their own ideas about things based on what they see around them. As long as what they can see and how they can act is conscribed in some fairly narrow social circle, then you can do it.

And lots of books do that anyway. Lots of books take place in these tiny comedy of manners stories, undercover cop stories, three people trapped on a spaceship, and so on. You can tell multiple stories in these settings that turn out different every time, but that would actually be meaningful and reactive to the player.

It happens in TV, too.

AS: There's a Star Trek: The Next Generation episode where Picard gets trapped in an elevator with two children. He hates kids, it's part of his character, but in the end he decides these kids are okay. You get people and you isolate them, then you don't have to worry about simulating a whole world at once, because it's just so intimidating to even consider how you would start trying to do that.

I think we'll get past it, but in the meantime, epistolary narrative is still totally legit. Dracula is an epistolary story, right? I love it. I love reconstructing worlds from bits and pieces that I find. It's human archaeology; you're given a series of clues and you have to fill in the blanks. I love that process, but I want the other thing to happen too.

Is the way you've presented Capsule here something that can only happen every so often, at a festival like GameCity?

AS: Building a Capsule Capsule is something I could have done on my own. I could have hired someone to build an even more awesome and elaborate setup somehow, but where would I put it? Who would play it? What would I get out of that? Even though Capsule was only originally made for a very small audience, only a couple hundred people are going to get to play it this week. And that's way more than would get to play it not at a festival setting. It's not like [Johann Sebastian] Joust where it can be put into a briefcase.

In my excitement I imagined it as one of those moving, virtual reality rides you might find in a science museum.

AS: There were some ambitious ideas being thrown around, but we buckled down to the necessary constraints of reality. There were versions that involved refrigerated trucks; there were versions that involved actual holes in the ground where you had to climb in from the top. We had to make some compromises, even though burying people would be awesome.

Have you ever played anything similar? I remember one that was pitch black, with no screen --

AS: That's Deep Sea, which was made by Robin Arnott. It was one of the coolest things I've played in all of 2011. The reason that Robin worked on Capsule is basically that game. I was thinking that it would be great if it sounded like [Deep Sea] because I loved everything about that game. Then Robin approached me and said, "We should do a thing together," and I said "I have this game that's a huge mess right now because I don't know how to do any of the sound for it, and it's thoroughly inspired by your game," so it was a good fit.

Was Canabalt a game you worked on entirely by yourself?

AS: I can only think of one game that I worked on completely by myself. For Canabalt, I did all the programming, all the game design, all the artwork, all the sound effects, and Danny Baranowsky did the music, which was awesome. It was essentially made in three weekends, and the middle weekend was when all the important stuff was built. That was done game jamming at a friend's studio -- Flashbang, which doesn't exist anymore -- so there were guys there and we were talking about what we were doing. The big turning point was the morning of the second day. We were having breakfast, and I was talking to Steve Swink, who was doing Shadow Physics but is now doing Scale -- he wrote this book Game Feel, and is a super-thoughtful, interesting game designer.

We were both really in love with Farbs' game Captain Forever, which has a very self-directed nature. The Lego ship-building thing is the cool Captain Forever hook, but I think in a lot of ways the genius insight of Captain Forever was the bit where he doesn't force you to fight progressively harder ships. You get to decide. If it's not your thing, and you're able to get away, you can get away. That's totally fine. You can go and look for somebody to pick on that's more your own size. It goes against some traditional game-making wisdom, but it's obvious that it totally works. There's no pleasure in beating up tiny ships, but it feels so good to take out a giant ship, and become a giant ship.

That was the idea for crates in Canabalt comes from. And if crates can slow you down, it can't be a prescribed, pre-generated level; it has to be reactive. That one morning chat with Steve is where Canabalt's reactive generation came from, and where Capsule's reactive generation came from.

You said at the time you didn't really understand why people loved Canabalt so much. Have you worked out why they do yet?

AS: I'm hesitant to give any specific reason. But my general sense is that there's a clutch of different things that it does that are good: It's pretty awesome right away. One of the things that doesn't hurt it is that the game turns on, the music gets super creepy, things start shaking, you jump out of a window, barely land on your feet and take off running. That's just good. If you just started out on a rooftop, it wouldn't be as good. You need to bowl out of this window, first. Maybe that's a Prince of Persia II homage -- which it is.

I used to have this fantasy when I worked in an office building with this long, long hallway with this glass window at the end that looked out over a river and a cliff. You'd be in this office and a two story party boat would just creep down the river, filled with people partying. What a beautiful thing to feel -- if you were invincible, how beautiful would it be to take off down the hallway at top speed and physically and metaphorically explode out? I quit that job within a couple of months.