Level Design Workshop: Designing Celeste

[Music] yeah I'm Matt Thorson I'm an indie game developer if you have heard of me it's probably because of this game towerfall Tara Falls a local multiplayer game but we're working on a single-player game now and that's why I'm here to talk about it's called Celeste I'm just gonna show the trailer for it just because the game's not out yet we're still working on it so just want to sort of introduce it [Music] [Applause] [Music] yeah that's less thank you so it's like a hard platform about climbing a mountain it has like I've made about 300 levels for it so far we're pregnant around 400 before it's out so I just want to talk about some like our process and some ideas are like swirling around in my head as I make all those levels so this is how I like to visualize the game I basically think of it as like a fractal story or a story made up of smaller stories so at the game levels lust is about a girl climbing a mountain and like she has these personal problems and she's kind of running away from them but by climbing the mountain she ends up having to confront those problems and then like most platformers Celeste is split up into worlds we call those areas so each area is a smaller story within that story it has its own arc and pacing and we want each one to feel distinct and unique and they all have like aesthetic and mechanical themes running through them and then yeah and then each area in turn is made up of levels and each level is like a really small quick story and these are just like really simple things like I jump over a pit of spikes and dash upward onto like an overhang and climb up to safety that kind of story and visualizing each piece as a story is basically acts as our anchor to like Center all of our design decisions yeah it's really useful for that and in practice it means we're constantly zooming in and out to these three perspectives to reevaluate our work so I'm just gonna start jump right into levels and our approach to making the individual levels so Celeste is about traversing space so I just want to introduce the like basic mechanics of the game first so the most basic move you have besides jumping is climbing so you can climb the side of any solid block in the game and at least until your grip strength runs out yeah and your grip strength refills instantly as soon as you touch the ground you also have this air - ability which you can - in eight directions so Cardinals or diagonals you can only use it once in the air and just like your grip strength it refills instantly when you touch the ground and then you have a wall jump the wall jump is powerful because it doesn't have a limited resource like those other moves you can do as many times in the air as you want as long as you're next to a wall however it's not like like it pushes you away from the wall quite a bit so it's not like the super meatboy wall gym where you can just sort of climb up a single wall by while jumping repeatedly if you want to climb up a single wall like that you have to use your climb and use your grip strength so the moment-to-moment gameplay in celeste is really inspired by a real-life rock climbing and one dimension of rock climbing that i love is the idea of multiple approaches like because rock climbing uses your body and everyone's body is different routes and rock climbing are usually designed loosely enough that they can be solved in a lot of different ways and sometimes those differences are like really big and obvious like if you have like a really tall friend who can just skip like half a route but usually they're really small differences like the way you decide to like move your foot on a foothold to shift your center of gravity or where you like brace yourself against the wall with your hand and celeste like i said is about traversing space and vertical space is very important because you're climbing a mountain and our move side overlaps and lots of different ways like all those three moves I talked about give you vertical progress but they use different resources and have different nuances and trade-offs so you can approach stuff in a lot of different ways and when I started designing levels I didn't really realize how important this was so I would design something like this challenge with a very specific solution in mind like I think for this one my solution was the center one but playtesters never did things the way I wanted them to you obviously and I yeah I just really started to learn the importance of this and a big part of play testing now is finding these different solutions and enabling but so in general we want our level speak permissive because this makes them feel like really dynamic and expressive for players but you have to draw the line somewhere right because if you can approach everything from any angle you're not going to challenge the player and it's really easy to keep like after every play test keep standing off the corners of your levels until they're not so much levels as they are just like amorphous blobs of content and blessed from like a wider perspective is supposed to be a hard game right it is it's about climbing a mountain it's supposed to be challenging so one thing that's been hard for me is making a hard game without it just turning into this like rigid series of precise inputs the player has to do exactly as I tell them to to still let the player be creative but not make the game easy so the way I found that balance was to go back to focusing on the story of a level using the story as an anchor so for example the story of this level is that the player has to climb up the side of this this block but as soon as they grab it it starts shaking and falls so they have to do it quickly and then there's these spikes on the left that's sort of the the two spikes poking out on the top left so if you jump too soon those spikes kill you so there's the tension if you want to jump you have to wait long enough to jump and then once you get on that ledge you kind of have to pump yourself up for this like difficult series of jumps to get to the end and so just show how that story kind of steered us I just went back in version control this is the very first draft of that level and the pieces of that story are all there except notably the spikes on the top left aren't there because it took me some playtesting to realize that that story of climbing up that block wasn't interesting without that tension so here's the next revision where I added those spikes I also made these falling blocks like more forgiving so making that left falling block taller means that you can get on to it with more - angles and this makes it more permissive multiple approaches right so clinging to the block like it was all right to make that block getting on to it easier because it's get how you get on to that block is not important to the story of the level like onto the block and climbing it has to feel precarious and that's why it's hovering over these spikes but the story of the level is about climbing up the side of it and jumping off at the right time right so making getting on to it more permissive lets us put the spotlight on the part of level that actually where the story we're actually trying to tell and when the player thinks back to this level we want them to remember that story not some like annoying jump at the start of the level and then playtesting this another huge problem comes up where players just try to skip directly to the second block and it's actually possible to get to it and grab the side of it but once you grab the side of a falling block it falls and you just have nowhere to go and you die so of course this shouldn't be possible because it skips that whole story we had set up for the level and it replaces it with just a really lame one just like an annoyingly precise jump but it wasn't enough that it was already impossible it has to be obviously impossible otherwise players just tunnel vision on it and keep going for a surprisingly long time and get really frustrated so the next revision just pull these spikes up this makes this of the player has to dash sooner and they can't even get close to that block so they they stop trying this after like one or two tries and playtesting has also been really important for this for finding these routes that are impossible but not obviously impossible because those really negatively affect the players experience in this kind of game and just for fun here's all the revisions of the level from the very start yeah I do a lot of stuff like I try putting spikes on that second falling block for a little bit but it just ends up being annoying and restrictive I like shave off ceiling tiles and add platforms and stuff but the important thing is that the story of the level never changed if I found that the story of this level just didn't fit in the area it was in or it was just a bad story then I would just scrap the level or try to adapt the story into something else but the story is always the cornerstone of the design this is a screenshot from sgdq the last sgdq marathon we did this four day prototype of Celeste in Pico eight and they did a run of it GQ it was like a two-and-a-half minute run and the way speedrunners play this game is really inspiring to us so it actually is possible to skip the intended path in that level it's just like really difficult and this is intentional this is the past we do on levels where we try to add these it's just really important that no one ever mistakes this for the intended path and like no person on their first time through the game is going to have the skills or understanding of the systems to do this so I just have like this mental list of different techniques that we never teach the player and we never require them to use to beat the game normally but we just like they're fair game for skips like these here's another one where the player has to do this like long dash jump that we never teach them how to do and they also have to know that when you go through this level transition it refills your dash as though you had touched the ground and we never teach that either so only the best players are gonna figure this out and as far as like multiple approaches and rock climbing goes this is like when you go to the rock gym and there's that one person that's just like super ripped and doing all the routes with just their arms and like if someone wants to put the effort to get that good at our game we want to support like reward them for it so using these techniques and a few others you can break pretty much all of our levels and I kind of see this as like a second story to the level as long as no one on their first playthrough of the game thinks this is the way to do it then coming back to a level later and just like breaking it with sheer like skill and determination is a really cool like second layer to the level excuse me another thing I love about climbing is the idea oh okay another thing I like sorry I'm missing a slide okay well another thing I love about the climbing is the idea of safety so in a climbing route you have to do like a series of moves obviously to get to the top but between each move you're hanging on to the wall in a different position and some of those positions are more safe than others and they're also more or less safe in different ways like for example you could be using a lot of core strength to balance on a hold or you could be using a lot of finger strength to grip like a really shallow hold and this is something I think about a lot in Celeste so different positions in Celeste are safe in different ways so standing on solid ground is like typically completely safe as long as you're not in imminent danger and it refills your - in your grip clinging to a wall is usually pretty safe but you'll run out of grip strength or bench eventually and fall off and then there's these like crumbling and falling blocks where if you if you stand on them you'll get your - back and refill your grip but obviously you have to move quickly before they disappear so using the idea of safety is probably the biggest way I control the pacing of levels in Celeste so a level with hard moves but safe holds between them feels very different from a level with a lot of easy moves but no safety between them and we can you know we can have different kinds of safety and we can give optional safety out of the way like there's a lot of ways to use this idea to make levels more dynamic and varied and interesting and this level is an example of one where there's safe resting spots between every move but the moves themselves are very difficult in the levels long so it makes it really slow tense experience whereas this level on the flip side is very short but has like no safety in the middle like these spikes now they're just poking out a little bit after you touch them a short while after they poke out and they stay out so area you've covered before becomes unsafe and you have to keep moving and the level ends up feeling like one fluid motion even though it's lots of moves and in this level the player has to touch all these flashing switches to open the gate and use these moving platforms they've Springs on them that bounce you up and refill - in your grip so there's like a lot of different kinds of safety in this level and the player can use them in any order and mix and match them and also go for the switches in any order they want and here's one way to do it when I'm prototyping new game objects I am really thinking about how they interact with these ideas of like multiple approaches and safety so most of the levels in Celeste are very short we just want to avoid repetition like there's a lot of single screen stages the game doesn't have any checkpoints really if you die you just go back to the start of the level you just we're in and players die a lot obviously it's a hard game and they'll be repeating stages a lot so we just don't want to take away player progress very often but thinking about length ties into the Edit the idea of safety because any time the player reaches a spot of complete safety where they're just standing on solid ground with no danger we have to question why we aren't just ending the level at that point and if we ever decide not to end the level the reason is to raise the stakes the longer level is the higher the stakes are because you lose more progress when you die and high stakes really change the feel of a level and really amplify the difficulty of everything because of stress we just have to really make sure that long levels can support repetition especially the start of them so in general levels get longer as the game goes on but throughout the whole game we have a few like very long levels that we use carefully and like sparingly mostly for like pacing at like really intense story moments to make it more memorable okay I'm gonna talk a bit about our area design now it feels area arranging the areas and celeste feels a lot like arranging a song but I'm getting ahead of myself I'll talk about that in a bit aries and celeste are a lot like metroid maps so the rooms are all connected and seamless so you can like move from level to level there can be branching paths there can be secret branches you can go back to previous levels that kind of thing the system is versatile enough that different areas can experiment with different structures this is our editor we use to clip click the levels together into our Maps just like with the levels stories our anchor for these so we want every area to have a unique and interesting story arc and we want the experience of the player to match the emotional experience of the protagonist so that hopefully the whole game is always working together toward a singular purpose we're always looking for ways to offload our narrative on to our level design so for example in this area the player is helping an NPC clean out an abandoned resort but the guy you're helping is kind of messed up and he's not really appreciative or like and he's kind of trying to like trap you in the resort as you're doing this and this highlighted part is like you're on the homestretch and the levels are picking up in difficulty and your characters starting to get frustrated with this guy and getting frustrated that she's being roped into helping him and at the end of this part there's a narrative like beat where she snaps at him and our goal is to make that emotional transition happen on the players side as much as possible to make them feel like why am I going to all this trouble for this guy like he'd better give me something good for this and by the end of it hopefully you're you know done with that guy so I start an area with like a vague structure in mind and like a general sense of what has to happen based on zooming out looking at its place and like the bigger game story and I prototype unique game elements and interactions for that area and search for like a concrete story arc for it but at this point there's kind of a chicken and egg problem because we don't know the story of the area yet and we also don't know the story of the levels in the area that really make up what the area is and we want the two to be in sync so I don't really have a solution for this uh-huh so I just kind of used it as an opportunity to experiment so I approached some areas by mapping out like the whole thing on paper first and then just trying to fill it in with actual levels and other areas I just like doodled in our level editor for like three days and then without really thinking about how they would connect and then try to put them together and I also tried to draw on different inspirations for each area so like one area is kind of inspired by undead bird from dark souls or brinstar forms metroid or the bathhouse from spirited away and but of course like whatever approach I took it never matches up like I always discover that the level drafts and the area drafts just are incompatible in some way and just we just kind of have to like mash them together like slowly and painstakingly like for example when I'm map yo nari I'll say like and here's ten levels where I just have bouncy cloud levels like they used this bouncy cloud mechanic but then when I go to make those levels if I can't get ten good levels that fit they're out of bouncy clouds and you know we have to rethink things or maybe I'll get 20 levels out of bouncy clouds and I'm like what do I do with these levels and we never really got anything right on the first try everything changed and but the areas and the individual levels had to adapt nothing came into this game fully formed and we throw it a lot of work huh like just last month I threw out 20 levels from our latest area because they just didn't fit the mood so our approach kind of feels like brute force stubbornness here I guess it doesn't sound clever but we just keep like zooming in and out to these three different perspectives and just to see if our if what we're doing is connecting and just kind of doing our best to redo things until it connects at all three levels and we always end up with extra levels so we added this hard mode where we can just put we the difficulties hires we typically make levels harder for this but it's just a place to put like cool mechanical ideas that don't fit into the story of the main game so we can just fully explore the design space so when I'm arranging the areas really thinking about story of each level and where it fits into the area story and because our areas can be nonlinear we have to think about you know bottlenecks for narrative like cutscenes and also for teaching and it often feels like arranging a song because so much is about pacing and framing and you want to do something cool with like the overall structure of the song but you always have to be aware of whether your melody can support that structure so an example the initial concept for this area was just a simple boss sequence were you being chased by like a dark version of yourself you go up the left side this tower and back down the right side and I designed this really cool chase sequence and these are actually some of my favorite levels in the game but what surprised me here is how as a story when it was done it just felt very flat so to make a satisfying story out of this chase we had to add all of this framing around it and in the end most of this area is just there for framing just to like foreshadow and ramp up to the chase and give it context and then properly cool down from it move the narrative forward in a satisfying way so at the start we have the slow exploration section it's very easy and freeform we want to feel like dreamlike and mysterious and the challenge here to progress doesn't really involve mechanical skills about like perception and exploration then there's this narrative beat and these mysterious blocks change so that you can surf through them with your - now all the side paths open up and you can do all these optional challenges we just wanted this to be like really exhilarating like you just like realized you could fly in a dream and then the dream kind of morphs into a nightmare and we have that a chase sequence that the whole area is based around and it climaxes with this like long brutal level on the right side and this is the first time in the game we have continuous time pressure like this so it's very stressful and it's meant to be like a panicked feeling and we're setting up the main conflict of the story here and then we have this cool down section with easier levels ending with like cutscene where the main character realizes she's dreaming and kind of resolves the chase for now it took me a long time to realize that this area shouldn't have any risk of death like I one-point like a couple weeks ago I just went through and deleted all the spikes from these levels and they got way better and just to keep the player engaged after such a crazy like challenge we had to like dial down the difficulty way down to zero and then finally your character wakes up and this is just a really simple section where you run through a bunch of the same levels but there all subtly different like none of the dream elements are there and this gives a space for some like optional NPC dialogue and also gives us a nice like calm ending to this like rollercoaster of an area so teaching is also relevant really relevant to how we arrange the areas in Celeste this is a level late in the first area there's these blocks that move when you touch them and the player has to figure out that they can grab the side of it and like ride it and jump off at the last moment to boost across this gap this level illustrates our approaches to teaching obviously if you just drop the player straight in to this level they would have no idea how to do this and we'd never get past it but with Celeste our goal is to teach implicitly with the level design as much as possible and the only text we use to teach is the text to tell you the controls at the very start of the game so the first way we do this is by priming the player this is the very first time you see one of those blocks and if you just do the obvious thing and jump off of it it flings you into these spikes and kills you so killing the player is the most obvious way to teach people in Celeste because you lose so little when you die and death is also like so jarring that it forces the player to confront what happened but we also have a lot of levels before that point or before we require them to use it where the momentum isn't an obstacle that is just relevant to navigating the space and they might not consciously be aware that it's happening but it will be logged somewhere in their subconscious and then after all that stuff the player reaches this level and here we're asking them to have like a realization about the game system right like they have to become consciously aware of these momentum mechanics and then they have to use it in a new way so when Plato says get here and figure this out without any additional help they really love this level because they had to think creatively and the game trusted them to solve it but most of our play testers get stuck here and this less like our goal is less isn't like no player left behind like it's a hard game and it's okay for people to get stuck like where you draw that line of how much you're expecting another player is in the end arbitrary but I just like for the first area I just didn't want everyone to get stuck here pretty much and like Celeste is a hard game but I also want it to be easier than it looks so the solution I arrived at here was to add this left branch if you go up the left branch there's just these two levels the whole point of them is just to teach you very clearly the steps you need to beat that level and most players go through this and then it plops them out right back where they started and then they have that light bulb moment the second level here where you have to do the sideways boost that one literally locked like that platform locks you in see there's no other move you can do than do the boost you have to do and one really important part of the solution is that it never gets in the way of the good players like the players who could solve it on their own can still do that it's completely invisible to them and ideally players from a wide range of skills can play this game and it'll teach them all how to play it silently like without them noticing and through the course of the game you learn a lot of things this way right and at the end of the game we have callbacks to all those things and hopefully the player will look back and see how far they've come at that moment because that's a powerful feeling like it's not unlike looking back from the top of a mountain you just climbed and yeah that is my top I think we probably have time for questions so if you guys wanted to do that there's two mics did you see I take herons talk I'm probably pronouncing his name wrong but I know I didn't it was really good it went through something similar Thanks hey great talk I was wondering how much of the QA you kind of like rely on yourself and going back to the levels again after you've had time to think about it and how much you give to like other people for testing yeah like playtesting is really the center of like it's so important to me like if like I can go back to levels and see them sort of with fresh eyes sometimes but for the most part like when just the act of sitting down and watching someone else play the game shift something in my brain and I just see it in a new light like instantly and then also obviously you just see them keep messing up on the same thing over and over and from yeah I think on playtesting it's all of your place testing in person or do you catch replays or analytics or death counts like with people that aren't actually in front of you yeah it's all in person so far it just like seems so much there's so much you miss from just like analytics I think like you can like when you're in person you can see like my favorite way to play test list is with a couch full of friends all like it has their undivided attention and their maybe they're trading off the controller or maybe one of them's playing and they're talking but you can just overhear them talking about it and like going through the mental process of like see the gears turning in their head I make me think of your first game drum poetics like it's like you it's been a long time you did it and it's still the same themes in the same way I mean it seems very evolved but is it theme you you plan on on changing I mean is it really something that counts for you I have very poor vocabulary so sorry but like a what was what was the start of your question sorry the like is it a theme that really is important for you like this same as genetics oh it's jumper yeah John oh yeah yeah jumper was like a first game I made when I was a kid and it was pretty similar to this so the question was is the theme important to me yeah yeah yeah yeah I mean I grew up playing a hard platform like donkey on country Mario 3 like those kind of games are really near and dear to me and yeah identify a lot with the genre I do this game is kind of like going back to my roots I guess before I go off and try something new after towerfall hello at some point you talked about making the long levels like repaid replayable at the beginning because if the player is going to die a lot of times and they will have to play a lot the first part how do you do that do you have like formula for that I'm always looking for more ways to do that when one way is like be aware of cycles like don't make the player wait at the start of the level obviously another thing is Celeste has collectibles there's these strawberries but like in a lot of platformers when you get a collectible you have to beat the level as well to keep it that's not the case in Celeste so you can get permanent progress so a lot of the longer levels have a lot of collectibles or like more than normal so that even though the level can be a bit of a grind like you're still making this permanent progress you know so it's like a consolation prize and then also multiple approaches is very important for that too to have like the early parts of the levels so that there's like an obvious way to do it but then there's like a faster way that you can kind of discover after you've done a level 50 times yeah thank you yeah okay thanks [Applause] you

Channel: GDC

Views: 234,622

Keywords: gdc, talk, panel, game, games, gaming, development, hd, design, celeste, level design, nintendo switch

Id: 4RlpMhBKNr0

Length: 31min 33sec (1893 seconds)

Published: Tue Feb 27 2018