GDC Plays Teardown with Dennis Gustafsson

Hey everyone, welcome to the GDC Twitch channel. My name is Bryant Francis. I am here. The I'm no longer the invisible voice inside your head. I used to be very invisible, but now I am here in front of you with planets all around me on the spirit overlay art. Um uh talking to you today about a neat new game called Teardown. If you're here in chat, and I already see some great folks in chat like 4 Cat RPM and XY01 Design, you probably know what Teardown is. I've We've been watching these clips on YouTube for a few weeks, haven't we? Um uh Teardown is a game where you tear things down. It's uh full of got these neat voxels. You are pushing and prodding things. Uh for for those of you in chat, uh if you're not new here, lately we've been doing this thing where I record the footage in advance so I can chat with the developer who is down below and we'll get to him in a second. Uh so the footage you're seeing today is pre-recorded and you're going to see me like learning the game cuz we're only playing the first couple hours. You're going to see me having some pretty like organic

reactions to things, which we're going to which are going to fuel our conversation. But, enough about me and my mediocre gameplay skills. I am here with a wonderful person, Tuxedo Labs himself, Dennis Gustafsson. Dennis, would you please introduce yourself and tell us how you're doing? Hi. Yes. So, uh I'm Dennis Gustafsson and I've been one of the main developers of this game, Teardown. Uh I've been working on it for almost 3 years, at least 2 and 1/2 years full-time. And uh it started off as a solo project, but uh it um grew into a small team and we're now two main developers. That's myself and Emil Bengtsson. Mhm. And then we have a number of contractors who's been that has been helping out. So, I think we've been seven people involved in in creating this game. And it's currently in early access on Steam. Uh and we're going to work for at least another year, we said. Oh. To

build more content for it and add mod support and fix bugs and whatever the community uh based on the the feedback that we get from the community. Sweet. I uh I uh what I take out of that is oh sweet, we can have this conversation again in a year again and talk about what you've learned in that time. Yes. But yeah, uh Teardown. So, uh you can see me going through the tutorial right here like learning the basic mechanics and getting to the first mission, which I'm going to spend way too much time on uh today. But, uh Dennis, I would like you to sort of tell us um what were the origins of Teardown and how like like I've seen games like this sort of come and go, you know, like the post-Minecraft boom was like, oh voxels and crafting and building like were very big, but this is a game instead of building things up you tear them down. So, what where did this game come from and what were kind of the first things that emerged that made you go, oh we have a fun game here. Uh we should [laughter] keep developing this and try to make it into a full-fledged thing. It's actually a pretty long story [laughter] because this this game started as just a technology experiment. Like most of my

previous games also started with some kind of thing with technology like playing around with with some something new. And uh this was very much one of those games where I started tinkering with I actually started with a rendering like doing it a custom ray tracing engine. And then added in voxels a little later. I was kind of fascinated by art style. Mhm. Uh I'm very inspired by MagicaVoxel community. I don't know if if you're aware of that. It's it's a little voxel modeling and rendering software. Uh and I really like this style of blending very abstract voxels with super realistic lighting and and the combination of that. Mhm. Um and then I also have a history of uh making destruction for games. I've been working with physics for a long time and I've also made uh or been part of doing games with

destruction before. Mhm. Most notably is uh the mobile game Smash Hit. It's also based on destruction. Uh so, I was playing around with this technology and voxels is very suitable for destruction, much more so than than traditional polygon-based games. Mhm. Uh because they're just so much easier to work with when destroying stuff. Uh and also with the physics it really works well together. So, we had this little sandbox playground where where you could just run around and destroy voxel objects. And then turning that into a game was a pretty long process where uh I was prototyping stuff together with Emil for at least half a year trying out different types of gameplay. Uh but we didn't really find anything suitable and then I I continued doing that on my own for another almost half year before finally landing in this uh

two-phase uh heist type game. Yeah. Uh because because what's fun with this was to uh to do to really use destruction for the gameplay where you can do more of a precision work with destruction as opposed to just having destruction as as an effect as you typically see in games. [clears throat] Uh but that also meant it was pretty hard to find a suitable gameplay or game idea uh under the assumption that everything should be destructible because uh a lot of traditional game designs and the tools we use for level design and game design in general, they don't really work. They kind of break down when everything is destructible cuz you can't really restrict the player Mhm. with walls and doors and and all this stuff. So, so it was a long project. It was a long a process to end up in in this type of game. Mhm.

Yeah, uh let's talk about the first level that we see here. We've got folks in chat. Uh I think there's some like some people pointing something out. Andy the Noqui um says they're amazed they could play a game with a ray tracing on their 1060. Uh I'm running on this one like a 970. Uh so, if you think this is the game looks good, uh it's running on my old graphics card. Uh I'm trying to get a 3070, but everyone else is. Everyone else. Anyway. Right. Um so, [clears throat] uh this level's interesting. What I experienced in this level and what what we're watching me do in real time is sort of learning. I'm learning how the destruction in this game works um and how uh how I'm supposed to complete this objective, which is that this building is uh oh it's just 6.3 m too tall. Uh my dumb American self does not know what a meter is, so I don't know that means tear down [laughter] the building. I'm sitting here thinking I need to knock a few inches off the top. Um [clears throat] uh and and it's not once it once you understand it, it is straightforward, but as you saw me do a few minutes ago, I was I tore down like all the walls in the building and I looked up and the

whole building's standing up and I'm like, okay, well what the heck am I supposed like what am I doing? Like how what is blowing up? Like there's there's a game logic here that is very interesting and ties into that destruction. Um what was sort of your process for build like and this is the level where the player is learning it, right? Like this is the Yeah. tutorial in a sense. Um Yeah. What we got to No, that's um I I see what you're meaning that the destruction it doesn't really feel the weight of the building. It can be held up by very small piece uh of the building. And that is uh first and foremost is a technical limitation. It's just like if I should simulate stress in every part of the building entire time, that would just be uh too time-consuming to do in real time. Yeah. Uh so, that's the that's the main reason for it. Mhm. Um and then the reason we put we talked about putting another type of a house because we can also do other types of of destruction where where this

is is kind of hidden by jointing stuff and having more like wiggly buildings and and but it was sort of honest to be up front like this is this is how the destruction works already in this tutorial level. So that people don't get disappointed and that all the other buildings don't work that way. So, it was a kind of a an informed decision to to make it that way. Yeah, I I I think this tutorial's interesting cuz unlike other tutorial levels that are very like everything is set out and you just do exactly what the game tells you. Yeah. section before this was like that, but this is very like, oh no, you're going to feel lost. If you if you haven't watched any videos, you're going to feel lost and confused for a bit until it clicks. And we're going to watch in real time me processing it clicking. Um Yeah. that's that that's a very interesting way. I know you're in early access, so maybe this will change later on, but um that's kind of an interesting Yeah, I think it's kind kind of like the idea and I I get what you're saying that it

is a very open-ended solution for for being a tutorial level. But, that kind of reflects the whole game, I think. So, like there there is no set solution to any part of the game. Mhm. So, I I feel personally this this worked out pretty well as a tutorial level, especially since now I see you're running out of uh propane tanks here. Yeah. And that's also supposed to happen. Like there's only five of them, so you have them to kind of figure out what to do when you run out of them. And there are multiple ways you you can bring it down in in other ways. Uh there's only five, huh? We're about to find out how uh how I solve that problem. Uh Going to chat, um let's grab a XY01 Design's uh model. Uh I K I IKI was calling me out here like, what? It's pretty straightforward. Just take down the base of the house till it collapses. [laughter] Do I look like a smart person, IKI? I'm here. I'm doing this. I I don't know. I don't know what Anyway, XY01 design would like to know

their question is any insight on the voxel data model. Is there anything you'd like to share about I sort of don't know how to ask that question in a way cuz I don't know a lot about voxels and modeling voxels, but is there anything you'd like to share about how the voxel modeling in this game works? Yeah, I don't know how technical we should we should get in this stream, but it is the whole engine is built from ground up to do just these voxel environments. So, it's a custom engine that's designed entirely for this. And um I think what I do differently from most other voxel engines is most other voxel engines, including Minecraft, has this axis-aligned world where everything is aligned in in straight lines. Whereas I have instead of just one big volume of billions of voxels, I have thousands of smaller volumes that are filled with voxels. So, each object has its own

everything is straight within its own volume, but I have very many of them instead. And I think that's the main difference. Other than that, it's pretty similar to how other voxel engines store their data. It's just like a big 3D texture full full data. Gotcha. For folks in chat, if you have any follow-up questions on that, like I said, this is just something I know less about so we're taking chat questions if you want to know how this game got made. Ask them. Viper actually has a pretty good one. What computer did you you build on to iterate the performance? Did you start on a regular to mid-level CPU or GPU, or did you have a high-end PC and then slowly optimize it? I had a high-end PC when I started [laughter] and I still have the same PC which is now kind of mid-range, I would say. It's It's a I core i7 quad-core processor and a 1080 GTX 1080

graphics card. And I I used that as a target and then trying to get it a little trying to run really smoothly on that one and then it's going to run on a 1070 quite okay, also. But then because you can tweak the settings a little bit, it actually turned out to run on slower hardware as well. That came more as a surprise to me. I I never thought this would run on a 970, for instance. But it it turns out it it actually did. Apparently, according to chat, it runs on a 770, too. So, that's [laughter] All right. Good job. Would you Did that person actually is Josh Daffidy Wong who says they're impressed that the game came with so much of a campaign with a mission progression and update system even though it's in early access. That brings us to Yeah, we can talk about early access. What was sort of your Once you realized that this game was coming along that

like, okay, it's fun. Like the the the core like, you know, you can set up a house and tearing down the house is fun. What was your process for saying, "Okay, we're going to ship in early access, but early access is going to have this much content at launch and then we're going to plan out more for the the full launch." Like, what was sort of your measuring stick for figuring out we can do this and still call it early access? Yeah, it's a it's a good question. I This is our first game on Steam. We've done a lot of mobile games before, but launching something on Steam and for PC is quite different and we didn't really have any experience from that. So, we we kind of [laughter and clears throat] had to to iterate on that a little bit what the early access should be and I think all of us thought that we would release something smaller in early access. We talked about having maybe 2 hours worth of gameplay and something along those lines. But then it kind of grew and then we had uh We also found out during playtesting

that the campaign was actually took people way longer than than we thought it would. So, it turned into a like a bigger game and more content than we we thought it would at first. But it's not easy. I I think this game has had a pretty big crowd on Twitter for a couple of years and I think we could have put something out there much earlier, but it's also about setting the expectations, I guess, what the game will be. We didn't really want to launch until we could show our vision with the game. And I think that's where early access can go wrong. Like if you if you show something and then find out that this wasn't really the way we wanted to make something try to change that, you're going to disappoint at least 50% of the people and then yeah. But we also wanted to release early because we That's something we decided

very early on in the project that this would be perfect for a really early early access where where we can actually take a lot of suggestions from from people because this is new technology and new game mechanics and it would be awesome to have people's feedback and base the rest of the game on that. So, we didn't want to make too much or have it too finished. And that's when we decided on this. Um releasing the first chapter, the first part, so to say. Gotcha. For folks in chat, there are some folks saying, "Yeah, go drive the vehicles." I didn't know you could drive vehicles yet at this point. So, every time I watch myself walk by the construction vehicles, I just cringe. So, no, unfortunately, you're about to see me get frustrated and reload the level one more time. But you know, that's how it is. We got some other stuff, but we can actually go back to talking about some of the Actually, no, that's what I I want to talk about level design cuz it gets really interesting. But we have some good questions in chat.

Rex Guys, y'all are great with the questions today, by the way. Rex Cortex would like to know what parts of the game run as Lua scripts and is it possible to mod game logic that's not part of Lua scripts? You and I were just talking beforehand about how you're getting more modding available for this game. You're working on getting that ready right now. Why not talk about Lua scripts and then we can talk more about modding? Yeah, we That's something we started adding maybe half a year ago or yeah. A little bit more, maybe, that we moved over a lot of the game logic from C++ to Lua scripts. And that's because we knew that this would be a fun game to mod. We got a lot of requests for releasing the engine and like people wanted to make all kinds of games with it. So, we kind of knew that we wanted to make this a moddable game early on and that's when we started to break stuff out of the C++ and putting it in Lua instead. I think it was a really good decision not just for the modding

aspect, it also helped developing the game a lot, I think. So, we have all all the simulation is running in C++ and pretty much all the gameplay is in Lua. I'm rewriting parts of that now to make it easier to modify. But the the goal, and I think we're almost there already, is that you should be able to make an entire new game mechanic based on the same engine. Like you can make a new game basically as a mod to this game. That's that's the goal. I think it's going to take a few iterations before we're really there and more more specifically, we're going to have to release the level editor for that. And that's in It's not in great shape at the moment, so I have to clean that up a little bit. [snorts] But yeah, Lua is great. I I think it's a a great scripting language and

it's also technically I I've made it so that each script that we have we have dozens of them running at the same time, but they're all kind of isolated. So, it's it's pretty easy to go in and write another one and modify the game a little bit and we're going to release some examples of that when we release this improved mod support. But there are certain things like how things break and how vehicles handle and stuff like that, it's not going to be moddable, but everything that is related to how scores are calculated, what's how the progression works, what the levels look like and what the goal is and and all these things are going to be moddable and scriptable. Um Cypher would like to know why did you decide to use SSR for reflection color instead of full path tracing? It's just a performance um thing. The ray tracing in Teardown is

very special or very different from how ray tracing is normally done since everything is voxels. Ray tracing becomes much cheaper. I I could just do it with some acceleration structure. I think I could do the ray tracing in with the same objects that are rendered to the scene, but that's not the case right now. So, I actually have a separate voxel structure just with the 3D occlusion data, so to speak. Um where I perform the ray tracing. That also means I don't have access to any colors in that data structure. So, so I it's it's not technically possible for me to do full path tracing with the with that method I've chosen. But that also makes it very fast, it can run on older hardware. Uh I briefly had to pause the game here while I was handling something. Uh package person came at the door, and you know how it is. We're all

We're all in quarantine. We have to We had to We had to go uh I had to, you know, UPS comes, they're delivering something important. Um uh I had a big smile on my face cuz I was just watching the house wobble and remembering what it felt like to go, "I'm finally doing it." Like, "I'm finally blowing up this house." Um uh uh what technique do you use for the collision? Is it regular mesh mesh or is it voxel versus voxel on the CPU or the GPU? I can't believe I understand this. voxel. Yeah. [laughter] This is becoming a technical stream. I like that. Uh it's all voxel versus voxel. There are no triangles in this game. Everything is volumetric data. Uh except for water surfaces and these power lines that you see. Those are the the two exceptions. Uh but yeah, both the physics and the rendering is all performed directly on volumetric voxel data, so no triangles or polygons. Yeah. Uh and the physics happen on the CPU, and all the rendering, or actually not all, but most

of the rendering is on the GPU. Yeah. Uh folks in chat, I love your questions. We're going to keep asking them. I'm What I'm going to do though is I want to talk about um I want to talk about my reaction here that we're actually seeing on screen, which is like, "I'm finally getting it, right?" Like, "I'm finally knocking this house down. I finally uh I finally understand what a meter is." So, I'm like I'm eating this uh these dynamite at the chimney to knock it down. Um still not the most efficient way, but like you can see me like having fun in this moment. Like, you can see me like I could totally do this more efficiently, but at this point I'm chopping down the house cuz I'm like "I get it. I get it. I'm so close." I'm like I'm just mashing my mouse. Like, um as a game designer, like uh in a game like this, you know, there there like the player needs to have some kind of understanding about how things work. We talked about the tutorial. Here I am going, "Oh, no, the chimney is like a something too tall." Um what are your kind of thoughts around like designing for like frustration versus aha? Like, we're going to see as the next levels unfold like what the objectives are and like the more aha moments that I encounter. Um but like what is your kind of like like you said

you said like you designed everything to be so open-ended. Like, how do you like account for players who like might be so frustrated that they have to go look up a guy like like what are your sort of thoughts about how you work around player expectations in that space? Yeah, I I see what you mean, and I think there could be a a certain amount of frustration in this game, or I know there is. Yeah, especially in the beginning. But but I that that's not something we intentionally wanted in the game. It's It's more out of necessity, I would say. We wanted the player to start with a very basic set of tools, basically because we wanted to not overwhelm the players with all the tools, and we wanted to be able to reward players with new tools for completing missions. Uh but I think the frustration mostly comes from a resource limitation. Mhm. And that that's something we also noticed when a prototyping other game ideas

that we had for example a puzzle game that that was also a heist game where but the difficulty came more from getting to the goods and [clears throat] rather than not the the the planning an optimal route. Uh but having And for that to work, you had to have very limited amount of resources, like two bombs, uh four shells in the shotgun, and and the game just doesn't become fun with all these limitations. So, uh so, I think a way to get around that was to pretty quickly ramp up the the resources and hand out new tools uh pretty aggressively to the player in the beginning of the game. Mhm. Uh Barbarossa83 says, "That's how you should wake up in the morning. Just slash the alarm with a sledgehammer." Totally agree. Um uh I K Y would like to say they love that the time game is getting modding support after a director at uh DICE said modding is a declining trend. I have no idea if modding is a declining trend. I

feel like um you know, people have been talking about the Steam Workshop and shop and Twitch. I feel like modding is it's specifically it's on the rise in specific ways. Like, you know, like it's not it's not um it doesn't feel like it's, you know, like Age of Empires mod, you know, like like modding from like 1998 or whatever where we use the in-game editor to make stuff. But at the same time, like um uh I've been playing a lot more games where I pop into the workshop, and I'm like again, I'm not in this world very much, but I uh I feel like like modding has enabled something very specific in PC games. Uh while game development as a whole has had, you know, survive in the market, figure out how to support consoles, you know, like I I play Skyrim on almost every device that doesn't have mods and like in a very unmodded way, but like that's that's, you know, one of the industry's longest modding thing at this point. That is not a question. That was just me opining. Um I love that this next Oh, no, you you're about to say something. Go ahead and say it. Uh yeah, I'm I'm not a modder myself. I never been. I know Emil has been modding a little bit, Mhm. but

uh we're really curious what people will do with it, because we have I at least have this idea that modding for Teardown will not be so much about modifying the game as it will be building new content, new uh mini games, and new challenges with a new gameplay uh [music] for the game. But but we don't know, so I guess we'll find out. Yeah. Um I thought it was so interesting that this level like when I first booted it up, I was like, "Okay, how complicated is this? What is this going to be? Like, what" And I looked at it, and I thought, "Wait, I can do the" And you're going to see me like I'm just going to knock this out in a few minutes. And then I love that you immediately return to this level um and go, "Okay, you know the space. Now you have to navigate it with these limitations." And even then it's not as hard as I as I thought it was going to be when I started. There's just like This is such a good learning space that this level creates. Um what are your thoughts of like like like you said like the the objectives are open-ended, uh

but there's still a teaching component in the design somewhere, you know? Like, what are your thoughts about like teaching players how to understand these very open-ended rule sets? Yeah, I I think we do the previous one you played, this one, and the next one, I would say are considered like tutorial uh missions. Mhm. And may maybe the fourth one also in in in some ways. Um But yeah, I think since since this is new gameplay that hasn't really been done before, I think I think it's quite important to kind of ease the player into especially easing players into the freedom you have, because I've seen a lot of players doing playtesting mhm going up to door and see door is locked, and then start looking for other ways instead of just taking you know, a vehicle and driving through the door. Yeah. [laughter]

Uh and you have you need to think in a certain way that's not usually not allowed in other games. And if you play a lot of other games, uh it can happen that players are a little restricted in their mindset that what's possible. Uh I think that's what we're watching like we're like we're watching me do it right now. Like, I am learning like how to [laughter] how to think cre Oh, yeah, I love this part where I'm like uh like you can watch me like, "Oh, look at Oh, like" Love this. Um uh Yeah, and and I think in a lot of places like uh this skylift is is very specifically put there to help people get up to the platform. Yeah. Uh but it's not it's only in these first couple of levels that that's the case. In in most other missions, we did not put stuff specifically for uh to help people uh perform something in a way that we thought was the right way to do stuff.

Mhm. Um [clears throat] because we wanted to [laughter] to have everyone come up with their own ideas, so to speak. Um going into chat, uh ExtraMurda would like to know any features you wanted but already have had to cut? Uh sorry, what was that question? Oh, excuse me. They want to know are there any features that you wanted to implement, but you've already had to cut cuz either they weren't fun or they took up too many resources? Uh there's two I come to think about. One is that uh for a long time we talked about an inventory that you should be able to pick up stuff, and you had this uh while you were out on a mission, you could pick up stuff, put in inventory, and then when you got back to the hub or where you live, or like your your workspace, uh you could use that to decorate the hub, and that would be persistent. Uh and I think we got a lot of requests for that. Um but uh we never implemented it because there

were so many issues with having a persistent hub that if you something put something on fire, or you tear down the whole place, and just how that works, and then we also had this idea about a progression in the hub. I don't know if you played so far that you really noticed that things are kind of happening in the hub while you're out doing your missions and then come back. To visualize that the time has passed, we introduce things in the hub, like small little details, and that would also be pretty pretty tricky with if the hub was persistent. Mhm. And the other one is you mentioned that uh you come back to the same locations a couple of times, or we have multiple missions in the same locations. That's something we initially wanted to do much more of. Um like we wanted a bigger changes and especially more of a progression. Like you can see here. And now that you came

back, you had um uh the gate is locked for instance. And there were some signs up and you can see that there is a little truck over there delivering new um bars for the windows that were previously open and stuff like that. And And this kind of environmental storytelling is something we really wanted to do much much more of. And also involved uh instead of just time of day, we wanted to change the season and have much more uh things happening and more of a progression to the environment as you come back to them. And And also an initial idea was to have some procedural um changes to the environment based on what you did in your previous um visit. So if you if you tore down a wall, it would still needs to be a wall for the game to function, but you could see that they had kind of you know, patched it up and and stuff like that. And

Uh and I really like that idea, but it was just not possible. [laughter] It's too too just time time constraints and it would require a lot of new technology also to to make that possible. Chat, you're still asking some great questions. I will get back to them. Uh I I think we can talk about about narrative design here. This is a really interesting game. Like I said, I don't come to games like this to you I'm a very like I like storytelling games in general kind of get person. Um so like when I got into this game and and the simple stakes of oh you accidentally did a crime, you must now do more crime. Um and there's like a fairly involved uh for some reason I felt like I was in the UK when I was playing this even though like that map is totally fantasy, fantasy. But maybe that's cuz uh we had the Crown season 4 on in my house and maybe I'm just in that space. But um uh I there's like a just like the the the lightest flavor of narrative design and world building that I really enjoy with this game which is um you know, like it's the missions you get. I I appreciate the little detail that your

computer screen is not voxels. Like it's this very straightforward interface that feels approachable in this world. Um it you know, when you're like oh everything's voxels, you're like okay, I understand this. I understand text messages like um uh and there's all these like little like you both pop into a room. Like I remember where uh where when I returned to the factory, I'll see like I don't think I pointed out again, but there was like, you know, beakers on a table at one point and I was like like like there's stuff like these places feel lived in. Um I could see a lot when I play a lot of voxel games and they're doing environments like this, I feel like either unless there's like something specific like like a modder who remakes a specific city or um or like you know, like there's an element of like let let the proceduralism or let the chaos kind of define the space. But here like things feel so curated and I'm curious what you what you and your teammates thought about like making these spaces that feel lived in. Um cuz they really do. I I heard that comment before and I'm actually quite surprised because we are not particularly happy. I think we

anticipated a lot more of that. So the uh it was kind of rushed the whole put it putting up props in these environments and making it look like someone works there or lives there is something we would have wanted to spend at least another couple of months on to really And especially progression you can do a lot of um storytelling and humor I think with that these small little details. Like if you visit a room and then two levels later you come back to it and something happened and you can like start reading these little descriptions about what stuff happened. That That is something we wanted to go crazy with, but didn't really have the time for. But we'll see during early access if we maybe can spend some more time on that. But I'm glad you you enjoyed it the way it is because it's something we wanted to do, but we're not particularly happy with. Yeah, maybe maybe I'm also just being fooled by the ray tracing. Like I'm watching it now and I'm like

maybe it feels lived in because the lighting is so good. Um uh uh for let's grab this like there's a just brief discussion in chat happening about uh vehicles driving uphill. Um I K Y points out that in the villa level when you drive out of the uh racing garage, they uh they get stopped by the angle of the road. Um and other people go by and go oh yeah. Like so vehicle how do vehicles work which are you know, on wheels which are rolling? How do vehicles [clears throat] and voxels work together? You said vehicle logic was the one thing people can't mod when that happens. So I'm curious maybe we need to break down how vehicles work here cuz they're so crucial to the game. Uh so vehicles uh the wheel physics are uh just based on spheres. So the the wheels are actually spheres and they but they do collide with each voxel. So And that's why this um driveway to the garage it is hard to get up on because you're actually driving across these bumps all the time. We have some ideas on how to smoothen that out a little bit in the physics,

but we we haven't really come to that point yet. But it it will get better. But that's also the reason you can't really uh destroy the wheels of any vehicle. I don't know if you tried, but if you try shooting the wheels, they're not going to fall apart because we need them to be uh to be not broken because they're simulated with with spheres. It would be fun if if that that actually worked, but I think the vehicles would be massively undrivable if we tried to to simulate or have the wheels as boxes. I guess are there are there any other like like the notion of shapes here is a very fun thing. Like how do you use voxels to simulate shapes? Are there any fun instances when making this game where you're like okay, we need this thing to work, but it has to be the shape and that's kind of weird. Um are there any other like instances like that? Um visually there are lots of things like that. Like how do you make a door handle

in voxels? It sucks. Two voxels. And how do you make Also like you see the brick wall you're standing in front of right now? It has um it you can see the individual bricks. So it's not just a texture, it actually has depth to it. Um [clears throat] those voxels are 10 cm or like 4 in in in American units. Um which is totally unrealistic, but it's still kind of works. Uh in many instances we wanted the boxes to be uh half the size, but then the levels had to be much smaller. So we wanted to keep keep this voxel size for um memory reasons. But yeah, I I think the it was mostly just visually trying to make stuff recognizable in this resolution has been has been a challenge. Yeah. Um chat would like to know uh we

have one from Emmet help. Are you mostly CPU or GPU bound and which part exactly takes most takes most of the power? It depends entirely on the hardware of course, but typically you are GPU bound if you're on a good computer. Uh except for when you cause a lot of destruction and then in all instances when the frame rate drops to like five frames per second, then it's it's always CPU because then there's a lot of physics. I just noticed but but I also like to add that there is a lot of potential for optimizations that we haven't done. So I I think it's a pretty good chance those massive dips will actually be a lot better going forward. Yeah. Uh I also just noticed there's like a little smoke effect that comes out of the vehicle like that vehicle when you're you know, like just a standard like a diesel exhaust I guess. Um like like that's such like a little detail. Like what's generating that? Is

that That's cuz that can't be voxels too, right? No, the smoke is is also more traditionally rendered. Um It's actually quite funny because smoke is one of those things that people typically do with voxels even with polygon based games. And in this game where everything is voxels, the smoke is actually polygons. So Oh, deception. Lies. Yeah. Um but it's rendered uh the smoke is fully lit uh and it's using a very untraditional uh method for transparency. It's just stochastic transparency. So there's there is actually no transparency. It's just like either pixel or not a pixel and then kind of dithered the whole thing to to make it look transparent. Nice. Um yeah. So for folks in chat, we've only got about 20 minutes left. I will desperately try to get as many questions in as I can, but we're also going to ask some ones from the the GDC angle cuz GDC covers a huge number of

topics that Teardown weirdly fits into almost every single one of them in a different way. Um from graphics programming to uh narrative design to marketing. Um which I'm going to ask about right now before we get back to more fun voxel questions. Um this game is doing pretty well. Uh like you know, like just glancing at the Steam reviews. Um seeing like I see videos for it pop up. You know, you've got Chris Person's uh highlight reel which is usually my fair weather mark for like okay, how much how much is the game getting played? Like how many clips get submitted to that uh to that series? Um uh uh disclosure. I guess I have to disclose claim that I'm a patron I I back it that that highlight reel on Patreon. So I guess I shouldn't just mention it without disclaiming that. But uh yeah. Um Yeah. So you said before there's already like a voxel community, you know, like there's like people who have there's a community of people around the world who want to play games in the space, make games in the space, who want to experiment with it. But but there's kind of a mainstream appeal that you're slowly getting right now, which you know, which kind of on the chaos that this game invites. Like

what are your kind of thought like suddenly, you know, you're making games, you're trying to sell a game. What what has been your experience, you know, trying to get people to notice and pay attention to your work without fully turning your job into a game seller when you know, when you're when you're a game developer? Uh I think we for this game we actually had the luxury of not having to market it at all. We didn't really spend any time or money or effort marketing this game at all because when I started doing these experiments 3 years ago, a crowd was building up on Twitter. And that's where the whole thing started. And by the time we came up to early access, there were multiple fan Discord servers Reddit community and they were just communities everywhere around this game and we thought like any any marketing traditional marketing we

can do on top of this would be totally [laughter] it wouldn't really make any difference. So we we just we just didn't bother at all. We we did a couple of trailers trailers ourselves, but but that's that's it for marketing. And it is it is of course a really really nice and good position to be in and I I realize that it's it's not something that happens to everyone, but but it's um I think for this game it stands out a lot just graphically. That's probably what appeals to some people and also this enormous freedom you have. There's been a lot of games promising um the fully destructible worlds and then most of them failed in one way or another to really deliver on that. And I think this is something people have been dreaming about for a long time to have something

that is actually fully destructible. Uh that's also been uh appealing to a lot of people. Uh to on the on the I think that fully, you know, like fully like do whatever you want logic is actually hitting an interesting place right now, which is I'm trying to solve this puzzle. I'm trying to plot out a path, you know, grab everything, get out in 60 seconds. And right now my brain was going as I was up there, wait a minute, could I just cut the power to the whole facility and turn off the alarms? Uh so I Oh, I see. So I came over to the power generator to try this out. We'll find out how it goes. But I I think this is one of the fun moments that I had recording cuz I was like, I have a brilliant idea. And as you're about to find out, it doesn't work, but I gave it a shot. Um Uh we got folks on stream who would like to know Merly Maple would like to know what are your thoughts about multiplayer and co-op? Are such things possible? Will that have to be left to the modders? The the discussion in the chat is lending itself towards it might not be possible with all these voxels. Like I guess like how does multiplayer factor

into your plans and what are the challenges in implementing it? Uh it's probably the most asked question we get if this game will have multiplayer. And the sad answer is that we don't have any plans for it because we are not convinced that it will work technically. It's it's a lot of physics to synchronize and the engine is not written with with networking in mind. So it would be a huge amount of work to to even try that. So I I don't think that is on the agenda for at least uh anytime soon. It would be very fun to experiment with and I think if we did multiplayer or tried out that we would kind of go the same route we've done with the technology so far to kind of limit that to people with really good connections and low latency and not try

to do something that works for everyone. Uh but yeah, we'll see. I agree and I think everyone agrees that this would be a lot of fun with multiplayer. Yeah, but But we're just not sure how to do it. Making [laughter] games is hard. Uh that's what that's what that's what I Uh that's what I generally I come into every one of these streams going making games is hard. How the hell did you make this so good? [laughter] Uh yeah, let's see here. Um I uh um man, I love like this section. I guess we haven't talked yet about like the the uh like you briefly mentioned earlier like it was more fun instead of limiting players, you know, with the resources to get to the objectives. Like make it easy to get to the objectives, but make the challenge in getting out. So we're hitting the point where I'm like planning a run for 60 seconds. And let me let me iterate on something. I do not believe in practicing video games. If I need to practice something in a video game, I'm like this is work. This is a

job. I I practice martial arts. I practice like I technically practice X-Wing miniatures, which is again, but I have fun doing it, you know, like but yet you're I just did what I just did on the loop here was a run where I said, okay, I need to kind of know if I can do this in 60 seconds or do I need to do it? So congratulations, you tricked me into practicing. Um [laughter] What what do you what is kind of like you've been your thoughts about making objectives where player like you want players to like like experimentation Yeah. comes with kind of like I am anyways, yeah. I think you maybe noticed when you tried here that you got you cannot pick up the target without quick saving first. Yeah. And I think that's for a long time it that wasn't the case. But then we kind of forced it on people to have that on this specific level because we wanted to teach people uh this is how you're supposed to play the game. Some people consider quick saving cheating in a game because They're wrong. wouldn't do it. They're totally wrong.

And and for this game it's it's kind of crucial not to be overly frustrated to use quick save. It's it's part of the game play to do that. And we actually encourage people to use that a lot. Like if you want to try out something, quick save, try it out. It didn't work, you can just go back. And uh yeah, I I think I think that that is the the key thing that makes it work and and not being overly frustrated when you practice stuff that you have this option of always going one step back again, one undo step to the whole to the whole thing. Yeah, uh I care what we'd like to know. When I think of quick saving, I think of the classic classic, you know, like F12, F11 or whatever, F11, F10, I forget what it is. Um have you thought about adding a single button so that players can like do that like classic combo? Yeah, we thought about it and we might add something. I'm not sure because we also get

uh some comments from people who think the the quick save and quick load buttons are too close to each other on the menu because they if it's very frustrating if you do something wrong and hit quick save instead of quick load. Oh, yeah, yeah, yeah. And you got screwed and then having buttons for it would be probably disaster for some people if they press the wrong one. Uh but but it's very intentional that you can only go back one quick save cuz otherwise it would be kind of trivial if you can just back rewind as as far as you want. So it's it's kind of important for the game play I think that there is only one step. Uh and then we also have to be careful of how we present so people don't make the mistake of overwriting their own save. I've I'm caught struck right now as I'm navigating back to like execute the to execute the heist. Like how there's something very interesting about this game world that because it's so open like there's a natural human behavior like like I'm I'm doing I'm going for things that are familiar as opposed to things that are uh

like like oh, I know this path where I know I can just walk up to this building and execute and be on my path. Why am I re-stepping the entire route I took to get here? Um there's something very fascinating about human psychology and in this kind of space that like instead of finding the most the most efficient path, I am just like like relying on the familiar to get through it. And by the way, go me. I got this done in one go, but I imagine it's meant to be easy even though the alarms are blaring. Um Uh That's actually a point quite interesting because this level is the first one where you do this time-based heist in the game. And I don't think I've ever seen two people completing it the same way. Yes. [laughter] Everyone has a different route and that's that's a lot of fun to see cuz that's what we wanted with this game. So even if your path looks maybe complex and not optimal, it's it's definitely the case for everyone who plays the game. I totally missed I I take continue right there. I had missed this when I was playing that like it's re-showing my route there a little bit.

Totally missed that. Uh um when I was playing this. Uh yay, I'm I'm an amateur. Yeah. [laughter] Okay, we've only got about uh 10 minutes here and sadly we won't get to see me light anything on fire. I've got to wrap up a little bit early in a few minutes cuz we have a building inspector coming through to make sure I have not torn down the walls. Fun fact, though it did not involve tearing down walls, we did have kind of a very tear down thing in my apartment this weekend. They had to tear up the floors of the living room and I had to leave my house in the middle of a pandemic and go hang out somewhere else while they did all that cuz it wasn't safe for us to be around cuz turns out floors shouldn't be bouncy. You know that a hardwood floor should not be bouncy. So we're going to spend the rest of the time grabbing questions from chat and then I'm going to duck out so I can go meet the inspector. Uh but Marley Maple I like your what Marley Maple would like to know they've quick loaded so many times when they're trying to save. It's possible that they're

error. Maybe it's worth adding a toggleable are you sure to quick save that they can turn on and off in menus for advanced players. I think if there was a toggle people would turn it off and then still get angry. Oh yeah there you go. But I'm not sure it's it's kind of hard we we might experiment with separating them more on the menu so you kind of get that you know muscle memory which like quick save is left and quick quick load is right Ooh yeah. Like a menu that takes up the whole screen and yeah I I like I dig that. That's a that's a UI art thing. Jazzy QQ would like to know about primary rates. How do you find the nearest intersection for each pixel? Do you start by rasterizing a bouncing bounding box for each object and march the voxel grid of that particular object in the shader. I don't know if that was mark or marks or march. They spelled it march. Um yeah march yeah it's Ray marching is a technique where you just walk along the voxels voxel space to see when you hit something.

And yeah I think it what he just described is pretty much how it's done and drawing a bounding box for each volume. As I said before there are thousands of these individual volumes. Like the safe you're dragging now is one is one voxel volume. All the little debris from the wall you're you're hitting is also individual volumes and for each such volume is just rasterized bounding box and then ray marched in a fragment shader to see which voxel you you hit and then there's a material look up like an index palette for for each object. Gotcha. Josh Duffy Wong would like to know what is the coolest clip you've seen of someone playing the game and why is it the gameplay we're watching right now? No that's me. What is the coolest [laughter] you've watched? No I'm Uh that must be there are two levels in the game that we it wasn't really intentional but we we added more optionals to the game

to those levels than what we could take ourselves and that's kind of evil because we didn't know if it was even possible. Uh and some people hate those levels. One of them is called fine arts and the other one is called uh insurance fraud. Mhm. And they are so ridiculously hard that none of the team was ever even close to completing them. But it took just uh less than a week I think before the first one posted a movie movie clip completing completing that. Um so that that was very inspiring to see because we we didn't even know it was possible. And then there there was one uh other clip that really caught my attention and that was a completing uh I think it was actually on the insurance fraud level but not completing all the optionals [clears throat] where it was slingshotting uh

the objective of this mission is to collect uh put cars in a transport truck. Mhm. And there are like seven of them I think or eight seven seven of them in total. Uh that's why it's so hard. Uh but he built a slingshot out of uh these electrical wires uh and strung them up with a huge excavator to build up the tension connected with a plank and then set fire to the plank to get like a fuse for that to burn off and when that snapped off uh the wires can slingshot the whole vehicle into the transport truck. What? that was kind of kind of a creative way to [laughter] to steal a car I think and that was really fun to see. I I wish we had that video to run right now. I trust me I would queue it up. Um I Yeah there there is a YouTube video with

it [laughter] where I I unfortunately can't remember who posted it. Man okay. Oh god. We got to start wrapping things up. I apologize although there's still some fun stuff happening so enjoy my enjoy my teardown footage. We're going to we will grab one more question before we get out of here but before you go folks if you've enjoyed watching the stream today thank you so much for coming. Thank you for talking about teardown with us. Dennis thank you for making teardown cuz this game is just so neat and so cool and I cannot wait to have some more fun with it. Um Yeah if you are new to GDC and the what we do here at the GDC Twitch channel we talk to developers about the games they make if they're interesting. If you click the follow button you can get notifications for when we go live. We'd love to have you back. I'm trying to get us some more at least one more cool interview for 2020. Um I can't say what it is yet cuz we don't have a hard yes but we want to have you

all here talking about cool games with cool people. If you're interested in game making in general uh you can go to the GDC YouTube channel. You can join us at GDC. If you're a game developer game maker yourself we'd love to have you. Obviously it's a global pandemic. We're still trying to figure out 2021. We've got some cool stuff in the pipes right now. You have a few days left to sign up for our master class program and also if you are a game maker yourself and you ship something cool this year you can submit your talk to the GDC core concepts for 2021. Like say if someone made a game called Teardown and they should submit a talk for GDC 2021. I would urge you to do so. for my part is there anything else we need to cover? Only that thank you. We loved having you here and I hope y'all come back and I and Dennis I hope we get to have this conversation after I get I hope we get to talk again after this exit early access cuz Yeah let's do that. Thanks for having me. I'm so excited for what's going to happen for this game in a year. For our

last question today man y'all y'all are so nice in chat. Ken142 would like to know how did you make the fire extinguisher smoke so satisfying? They can't stop spraying it all over the place. That's a question and a compliment. Thank you. I don't know it it it it is actually based on fluid simulation that I did for another game called Sprinkle. It was a a mobile game just based water simulation game. So it is actual fluid dynamics simulating the the smoke in the extinguisher and I guess that's what makes it look the way it does. Right on. But it's also a lot of a lot of it comes from the lighting that is from the stochastic transparency. Right on. If folks have other questions I urge you to check out the Teardown Discord. Dennis is also on Twitter at Tuxedo Labs. If we didn't get to your question today my apologies. That's my fault not Dennis's. Blame me. Make me the object of your hatred. But then just go ask your question and

forget that I exist. It's fine. Anyway with that that's all we've got for today. Have a good one everyone. Stay safe. Be well. Keep those masks on and we will get through this nightmare year together. Bye. All right.