How Trepang2 Builds Intense First-Person Combat
How a team of just four developers built an intense FPS experience.
AI and Games is made possible thanks to our paying supporters on Substack . Support the show to have your name in video credits, contribute to future episode topics, watch content in early access, and receive exclusive supporters-only content on Substack.
Episode Bonus: Listen to our complete interview at the bottom of this article as a paying subscriber to AI and Games.
Plus if you'd like to work with us on your own games projects, please check out our consulting services provided at AI and Games. To sponsor our content, please visit our dedicated sponsorship page.
Trepang 2 is a first person shooter that takes no prisoners. It's fast, aggressive, and designed to keep players on their toes as groups of soldiers hunt you down and put the pressure on you. Forcing you to rely on your wits as they wipe them out one by one. You're cloaking, diving, sliding, punching, and slow-mo blasting your way through each set piece as you work to bring down the Horizon Corporation.
For this
case study, strap in for one of the biggest deep-dives into a game we've ever had access to. I sat down with two of the developers behind Trepang2 to find out how it all came together. The games that inspired it, the journey from tech demo to fully-fledged game, and a how a team of just four developers built a fully-fledged first-person shooter that not only pays homage to classic titles of the early 2000s, but carves out its own direction in the process.Follow AI and Games on: BlueSky | YouTube | LinkedIn | TikTok
Introducing Trepang2
In an era of indie studios recapturing the vibe and feel of classic first-person shooters, Trepang2 strikes a balance between over-the-top action and the need for survival of early 2000s titles, with the player utilising increased mobility and superhuman powers to face off against waves of ruthless spec-ops forces across a number of primary and secondary campaign missions.
Described by popular first-person shooter content creator GmanLives as "the FEAR sequel we never got" in his playthrough of the demo back in 2019, the close quarters combat in tight industrial settings, combined with the use of slow-motion is evident that the game is something of an homage to Monolith's classic shooter. But it's equally a lot more than that, you're sliding around, you're sneaking up using a cloak, enemies can become meat-shields, you can dual wield specific weapons, it even has a handful of boss battles. It's pulling from many different corners of the genre and then repackaging it into an experience that feels distinctly fresh, anarchic and above all fun.
But what makes this stand out in a number of ways isn't just how the enemy AI has to be designed to handle all of the weird and wonderful ways players can interact with them, but that the game itself was built by a core team of only four people. And so in the weeks after launch of the Bladekisser DLC in the fall of 2024, I sat down for a lengthy conversation with Wilson Chung, the programmer and designer of the game, and Jacob Andersson who is a technical artist and animator. Not only did we go into a real deep-dive on everything they had built, but shared a collection of behind-the-scenes assets and documentation that we'll see in this video - I mean in all seriousness this might be the most access to the making of a game we've ever had on the channel.
But before we get into the meat of it all, there was one burning question sitting at the front of my mind. What's with the name? And did I somehow miss Trepang1 hitting Steam at some point in the last couple of years? The answer is simple, but... also complicated.
Wilson Chung:
If we go all the way back to 2011, I was the university student, doing a game jam. And the the game jams theme was, loneliness. And, we had to make a game within 48 hours. And we knew we couldn't make, any kind of, like, sophisticated, like, physics or anything like that. So we made the game underwater. And then we're thinking, like, what's the sea creature? That's, like, really lonely.And then we're like, oh, like a sea cucumber. Like, I think they look pretty lonely, right? So we were looking at, like, how do they attack, other creatures, and it turns out they can't, but they can defend themselves by, ejecting some of their organs. And then their predators eat those organs instead of them, and then they can regrow their organs. And then we didn't know what to name this game, jam game. So we went on Google and we looked at, like, exotic, names for sea cucumber. And then I think it was either Indonesian or Malaysian. The term for sea cucumber was, terrifying.
So, that was terrifying. One and then. Yeah, starting trapping twos. It was just me in the beginning, and I didn't have any ideas for the name of the game. So I was like, all right, let's name it Trepang2. And then I guess we stuck with it because this game also involves, you know, ejecting organs everywhere. And so it's kind of similar in that way. I mean, there's no sea cucumbers, but like schematically, it's similar in that organs get ejected.
Tommy Thompson (i.e. me):
So I sort of read like, okay, it's Indonesian for sea cucumber. All right. And then the thing that blew my mind was then I think the next time I looked up, the next time I played it, I'm like, hang on: this got published by Team 17? You got that past marketing and PR?Jacob Andersson:
Well they tried! (*laughs*)Wilson:
Oh yeah. That was actually really hard. They sent us like a whole spreadsheet of, suggestions. But I think the funny thing is, I think what we came through in the end was we showed them, like, if you type in trepang, nothing shows up other than our game.
From Demo to Production
While the game was published by Team17 in the summer of 2023, it slowly developed a following on the internet courtesy of demos being shared on indieDB as far back as 2017 - with only Chung working on the project at that time. As the prototype continued to score praise for its intense gunplay and fun spin on classic shooting mechanics, the team began to flesh out with Andersson coming on board alongside Brandon Tran and Brandon McKagan. Each of whom taking taking on tasks across multiple disciplines for the project.
But while the team quadrupled in size, it continued to retain it's scrappy indie game jamming sensibilities. As word of mouth grew, with the game repeatedly appearing in the top 100 games voted on indieDB from 2017 to 2019, the team were adding more and more features, finding the fun - and the chaos - with every new addition.
Tommy:
Did the game start with all these mechanics in it or did you start with one thing and then just like, you know what? What this game's missing is a really kick ass slate, or this is missing a cloak like it was. It always kind of envisioned this all these different mechanics and it together or did it slowly. Every time you added something new, was it was like, “oh god, here's a new problem we have to, you know, factor in?”Wilson:
We didn't do this the proper way. So we didn't have a design document or anything like that. We would just get together and dream up crazy ideas and then, you know, a few days later, someone would be like, oh, hey, I added this thing, by the way. We're just like, what? Like, what did you do?Of course, every time it was like, this is so much fun. We can't do this any other way. But, yeah, every tiny thing that we added, like the sliding, the cloaking, the hostage taking everything. Just. It added, like a whole bunch of technical, headaches, that we were fortunately able to solve at the end. I guess since we, we all come from, like, really technical backgrounds, we never really saw it as like, ‘oh, man, this is, like, just a huge pain in the ass’. Like, it was always like a technical challenge that we wanted to, to solve because that that's fun for us.
Jacob:
Like, just in terms of abilities and stuff. Right. Sorry to cut you off, but like, yeah, it did grow like from, like way back when Wilson had like a prototype for, for trade pranks, sort of, general idea and everything like that. That hasn't changed that much since then. From when we released. We just kind of like elaborated a little bit here and there on that and built out generally like, the systems that ended up in the game around that. It was always already kind of there from the beginning, in terms of like abilities and everything like this.Tommy:
Like one of the, one of the things I did kind of take away from it was how much of the game is kind of ‘vibes based’ where, I could actually see - on the development side - you're like, “oh, you know what would make this really kick ass?”.
Actually, I was playing a couple of days ago in preparation because I haven't finished the game, but I was going back replaying a bunch of the missions I had played before, and my brothers are visiting and my brother came in and asked “what is this?”, but like in a good way *laugh*. I cloaked, ran up, punched a guy, grabbed him, held him as hostage, and pulled a shotgun out, and then shot a guy in the head and then pumped it single handed. And he's like, “that's the coolest fucking thing I've ever seen” *laughs*.
Just the sequence of actions coming in. And it was, I think that ‘vibes based’ aspect of it, of just, yeah, oh, I can grab a guy, or I could shoot a guy, and then as you may have noticed I'm a big Terminator mark [editorial note: Tommy has a giant Terminator painting in his office]. So anybody who's doing cool reload animations with shotguns immediately gets like my approval.Jacob:
I mean, so half of that was already there from like from when I came on in early 2019. When I saw the prototype myself before I even joined the project, that's basically how I reacted when I saw the prototype at the time. And I was like, ‘I got to jump on the ship and like, help him out to reach this mission’. We have to make this work, because I want to see this alive and well as well.
So what you mentioned there, about the one handed, shotgun pumping and stuff like that. That's basically where, like, when I came in and I started to build out the more visual aspects of, like, how do we fully, like, properly visualize all of these different mechanics and elements? That's like what what came afterwards. But at its core, it's very much the same as it was since pretty early on in the in the prototyping stages.
The AI Architecture
Now it's worth taking a hot minute just to go over the AI architecture of the game. While the hype surrounding Trepang2 often harked back to FEAR, the AI tech stack has very little in common.
As we know from our various videos on the subject here on AI and Games, FEAR relies on Goal Oriented Action Planning - a hybrid of forward-search planning algorithms with action execution handled using a finite state machine. Be sure to catch the original video, the AI 101 episode, or a feature-length retrospective with FEAR's AI programmer Jeff Orkin to find out more. But Trepang2 relies on many of the ticks and tricks found inside the box.
Developed in Unreal Engine 4, the core of Trepang2's AI is behaviour trees - again a topic we've covered at great length on the channel over the years with my AI 101 video a good starting point if you're not familiar. The sub-behaviours of the tree are organised by priority - with lower priority actions on the right, and higher priority on the left. With the system designed to override behaviours when necessary. Hence an enemy will investigate gunfire or disturbances until they're then shot at, with combat overriding these more passive and exploratory behaviours. This is actually rather similar to the enemies in Warhammer 40k Space Marine 2 where Saber Interactive built a modified behaviour tree architecture that allowed for this reactive behaviour to occur for dozens of NPCs at once.
But another area that of the architecture that strikes me as interesting is that the game generates its level markup dynamically during play. If you're not familiar with the concept, when you have enemies needing to use cover, or finding interesting points to attack from, they typically seek out appropriate mark-up that has been tagged in the level itself for them to go towards. In our artifacts episode on Marvel's Spider-Man, I explained that all of the web swinging and point launching is achieved using markup auto-generated using a tools pipeline. Meanwhile in our episode on BioShock Infinite, I discussed how Elizabeth's behaviour is reliant on the markup not just showing what objects can be interacted with, but they tell her what animations to play, plus in the FEAR retrospective Jeff talked about the team had to go in and do huge passes on each level to ensure the NPCs could see the cover points and interactable.
As Chung explained, Trepang2's markup is generated dynamically through use of Unreal Environment Query System or EQS, given the pace of the game is so fast, the markup is often highly contextual. Unlike a lot of shooters where the player enters through one area, and the NPCs take cover up ahead, you can often enter a combat arena from various angles, be it through a second doorway or you simply cloaked up, ran behind everyone and then opened fire. The highest priority for all of the NPCs is their own self-preservation, so there needs to be a dynamic system to generate the markup such that the NPCs always know good points to move towards.
Wilson:
What we did that was that was different than most games out there. Most games rely on a lot of hand-placed, markups, or even, like, a lot of triggers and scripts. So like a level designer has to put like, you know, if the player enters this room, this guy takes cover behind this couchBut for us, like, everything was so open ended because, like, we had, you know, rooms with, like, six different entrances or like, you know, maybe even there's, like, a second floor and there's, like a staircase and like, there's all these different ways of of entering a room. And the player moves super fast. He moves, three times as fast as as a regular bad guy. And they can also go invisible and, and slow down time. So basically, like, the player could end up anywhere and the enemies might not even be aware of it.
So we couldn't really rely on just like a lot of hand placed, markup, because that'd be really time consuming. We'd have to take into account, like, all the contingencies. Like, for example, what if a player, went invisible and walked into a room and then, you know, unclogged at the other side of the room and started shooting at people? They would need to, you know, reorganize themselves and get to cover. So we had to just do all this stuff at runtime because there was just no way to to like, pre-compute this over by hand, place it.
Hunting the Player
Perhaps the most interesting AI element of the game for me personally, is what happens when the player is being hunted. Given the non-linear, an often cyclic nature of the level design, it's common for the player to enter an open area in one direction, take out an enemy and then run away and break line of sight with the squad of NPCs. Given this is such a core requirement of the game loop, it's critical than when the player breaks line of sight that the NPCs try to chase you down, but it needs to feel reactive, and to an extent realistic. Not only because you could be standing watching them with your cloak active, but also given it feels like the enemies are reacting to your behaviour, and you can then capitalise on it. But if they're too smart, it'll feel like the game is cheating you.
Tommy:
How do you keep track of the player? Do the NPCs ever cheat or did you actively avoid doing that?Wilson:
One thing we didn't really think of when we first started the game. And I think, you know, most game developers never think about this is, so what happens when when an enemy loses sight of the player?So in most games, this isn't really an issue because, if I can see them, they can see me. Once an enemy sees you, they continue to track you, even if you, you know, it's mostly all the way behind them. They'll just turn around and shoot at you. But we we knew we couldn't fake that because, like, you know, people could go invisible anytime. They can zip around to the other side of the map. And, you know, you would feel unfair if if they just suddenly, you know, 180 never skipped you, as you sneak up on them.
We looked at a whole bunch of different, solutions where, like, you know, where should they be looking at? When the player disappears? Because at that moment, like, it's really quite like when you cloaked, like you're obviously not shooting and you're just running around so you can clearly see what the enemies are doing. And if they are, they're frozen or they're staring at the wall or if they're like, you know, just kind of psychic and somehow you can see all those totally break, the immersion.
So we handle this with the Overwatch target system.
The overwatch system is sort of the inverse of the markup system, in that rather than finding a generated point in the level to break the players line of sight, instead the NPCs are now finding a point in the game space to target after you lose line of sight with them. When the player breaks line of sight with an NPC or makes a loud of enough sound stimulus, it generates a LKP or Last Known Position marker in the game world. From this the Overwatch system generates a target location, which is found by generating a navigation path from the NPC to the LKP and running line of sight checks, with the farthest point on the line becoming the target. Given the current orientation of a character and the level geometry, there's no guarantee that the NPC can see the LKP at this time, hence they look at the Overwatch target instead.
The NPCs will then move towards the Overwatch target and often lay down suppressing fire and even throw grenades, to give the impression they're thinking the player is just hiding around the corner. Meanwhile the overwatch continues to update to the most farthest point on the navigation path to the LKP that has line of sight.
Once the LKP can be actually be seen, and the player isn't found, then the system runs a new EQS query to 'guess' where it thinks the player has ran off too, and it will generate a new Overwatch target. This whole process essentially repeats itself until the player generates a new sound or visual stimulus for the NPCs to react to.
But once the player is back in the fight, there's an opportunity for flanking manoeuvres and other team based behaviours. Ensuring that while your current target might be fleeing for their life, everyone else is putting on the pressure.
Tommy:
When you're moving around and they're flanking, how does does that work. Are they all whenever they flank, did they just wait until there's another viable time to flank or do they just try and flank as soon as they know you're not looking at them?Wilson:
Going back to the behaviour tree, the self-preservation is always placed, at the highest priority in the behaviour tree. The NPCs have a variable called, suppression [which goes] higher when dangerous things happen. Basically, like an enemy's looking at me or they're shooting at me, or they're throwing grenades at me, etc. The result of this is basically like if the player is zeroing in on a guy and just bullying one NPC, he's obviously not going to run out and, you know, shoot me in the face. He's going to focus on hiding. this is just a simple way to manage, like, you know, who should and who should attack and who shouldn't.So if they're in danger and don't attack, And then they'll just, you know, try to run away. So it looks like they don't want to die, and then everybody else is going to try to flank or find some way to sneak up in and ruin your day.
Tommy:
Does it have then a squad management layer because like some of them, because you have these bigger encounters and sometimes they clearly come together in squads of units. Like is that something that's actually modelled internally in the code base, or is it just like purely like an aesthetic thing? Like a presentation thing when they appear in the map?
Wilson:
in every encounter we group all the NPCs, of each faction into a squad of their own. So, all the task force guys have one squad in an encounter horizon. Guys have their own squad, the cultists have their own squad, or the monsters have their own squad, and they share sensory information and memory.So if a guy sees you, then everybody in his squad will eventually. Yeah. Notice that you too. But actually, we added a delay, so it's like, if a guy sees you, he has to finish his dialog before, the sensory info gets passed to his teammates. So if you happen to snap his neck before he finishes the sentence, you can actually, yeah, get away with that.
And then on top of that, it does split the, the squad into a cover team and an attacking team. It cancels out part of the behavior tree. So, the cover team will just try to, to shoot and do suppressive fire and grenades, at the player, but trying to keep their distance, so they won't like rush or flank.
But the attacking team will try to aggressively flank and get close to you. And then, this constantly gets reorganized. So the people who are in danger, their suppression value is high. They get put into the cover team, so they're not trying to rush into danger. So this insures basically half of the NPCs you're fighting. They're going to be trying to find some way to to flank you while the other half are trying to distract you or keep you pinned down.
But other than that, we don't actually do a lot of, squad behaviour. Like, most of the eye is just on the individual level.
Tommy:
Interesting. Do you ever balance the the shooting or anything like that?Wilson:
We never do anything like that. Because I think when we escalate things, like, we really want the player to feel it, and we basically put, like, a hard job at, like 21, we're like, we're not going to have more than 21 guys. I think there might be like 1 or 2 arenas where we have like 23 or 24 or something like that.But we knew like at some point it's just going to get insane. And we didn't want them to do the ‘kung fu’ behaviour where like, you know, half of them sit around and then the other half attack you or something like that. If there is an NPC and they're trying to fight you, they're putting 100% of their effort into it. They don't slack off at all.
Chaos Theatre
After running through several hours of Trepang2 in preparation for our interview, the thing I really took away from the game is how readable it is. The game does a great job of making the world legible to the player even in the thick of combat, animations help communicate what enemies are trying to do, and the audio barks and other foley effects help express to the player what is happening in ways that you only appreciate once you stop to think about it.
This was, as I discovered, a huge amount of work on Jacob's side of production...
Jacob:
We tried to properly represent, visually, what the eyes were doing and thinking at any time. There's a lot of information that the player should be able to derive from, not just looking at the battlefield, but also hearing everything that's going on.But so for the visual aspects and animation side, there was a lot of effort put into actually properly showing, what the AI, the brain itself was, doing at any time. So, for example, knowing where the enemies were looking or like focusing. We have like a dual layer system that separately turns the torso and the head. So, we can have a guy facing a direction, then starting to spot you at the corner of his eye. So then he actually turns his head first to signify very, very quickly that he's actually noticing you. And he's like about two engage you basically. But then as he's doing that, his body is coming to turn to be able to shoot you.
Since we have slow motion as an ability in the game, any piece of visual information gets, broken down in a very digestible manner. So whatever we can properly, communicate to the player, visually is fantastic in, in a state like that. So animations that are critical to, you know, like show the player basically when you're about to get shot or when it's relatively safe to be around this guy or something like that.
You know, like someone firing a shotgun and then going for the pump in slo mo. All of this information is, like, digestible. So it's essential when you're engaging people, like, which am I, which game I aiming right now?
Which one is safe to engage while he's doing that thing? Like, you can make split second, decision making in a scenario like that. So we try to really not just make it look cool in slow motion, which is very important as well, but properly communicating the mechanics and what the brains are doing, that we can't see, at any time. But that was very difficult.
Tommy:
I mean, you've got the technical side of that. So building up those animations. But then critically is something that Wilson said in the email, like you were doing a lot of motion capture for this project on an indie budget during the pandemic as well. So like, let's get into that. That well, so stressful.Jacob:
There wasn't a lot of motion capture done, to be honest. A lot of it was, pre-packaged animations that then got repurposed. So I spent a lot of time manually hand keying a lot of animation, assets for the game. But then at the tail end, we started working with a external motion capture studio. Over in Vancouver, where they actually helped us to make some cutscenes and fill in some of the holes that we had in terms of, properly polishing up these NPC, mechanics like them getting up, for example, from being knocked down.
It's a pretty complicated animation to, to do. And we wanted to have variations, for it. So we had, I think about a dozen made. There's, there's quite a few that we have in the game. But are we doing that by hand would be way, way, way too much work for me to do singlehandedly at the time. It was mainly just them, the guys in the studio, wearing masks during recording sessions. And I was just sitting in a discord call with them, trying to direct them over my phone that they had put up in the studio.
Some pretty interesting little setup we have there, actually. But yeah, it was it was a lot of fun working with those guys. You should you should have seen some of the guys. They were they were putting in some awesome, performances. So that was great.
Wilson:
Yeah. I just wanted to say like, so before we had all these transition animations, like the, the enemies or the NPCs can move just like bots and so like, they could turn around or crouch or encroach or, you know, they could do anything on a dime. And it kind of felt like you were a super soldier, but they were also super soldiers because like, they would one ad, you know, scope you or like, you know, crouch and then change direction and then, you know, sprint off in the distance or something.So I think a lot of the, the transition animations like it, it sounds like that that doesn't really add a lot, but it actually made them feel a lot more like grounded as humans because it takes some time to like, crouch and, and run away from a grenade or turn around to shoot you in that kind of thing. It added to the power fantasy.
Jacob:
But it also came with its own set of sort of limitations or difficulties when it came to, the mechanics of the NPC. So instead of before we had sort of built the game or the combat mechanics around them being able to do that so well, I mean, to a degree at least. But now that they were performing these sort of things like turning and, things that technically took more time than just straight interpolating or not even interpolating their motions. It made them pretty, easy to deal with.You could say, in the early stages of adding those types of things. So we spent a lot of time, sort of polishing and adding layers to it. So, in the beginning they couldn't turn and shoot at the same time, for example, which was detrimental to, to keep them sort of competitive. So if they're doing a simple turn, they could shoot later on. And, if they're doing a full on 180 flip, turn, that's when they like, pause for just a split second. But they're very quickly back to shooting again. But that was, another one of those things where we had to spend time, tightening things up just to ensure that they still felt, good to play against as competent, NPCs.
Tommy:
I think the one thing to, to add on the top of that is, is also the voice work you've got as well. So all of them, they have all the additional barks, like, I remember chatting with Jeff or at this one, fear. And just like it'll be a different scenario, of course, because he worked in a, he was working in a studio with a whole bunch of people and like, I think the sound guy just kicked his door in one day and said, we're recording voices, I need voice, I need voice prompts.And he was just like, ‘shit!’ and literally opened up like a word document and wrote like 50 lines of dialog or whatever. Like, what was that process like? Because again, you've also got the audio layer on top of it, and the combination of them sounding like they're trying to react to you in a tactical way, but also like, I just love when they're always just panicking like, that's some of the best.
Wilson:
Yeah. So I mean, yeah. Credits go to be a Brandon Brandon trained on our team. He directed all the voice actors and and wrote most of their lines. But I didn't work with him. Pretty closely on it. And I think one thing, he had a lot of fun with that once I gave him the power to, because he, you know, he's a creative guy.He likes to abuse. The power that I gave him was, So I basically originally it was just barks, but then I was like, well, what if, you know, we can make a system where we could, you know, grab nearby NPCs and then make a series of dialog between them. So, originally it was just started out with, like, you know, an NPC saying, I'm moving out.
And then he would move. But then we added a system where, we could grab a nearby guy who is laying down covering fire and make him say, like, “I'm covering you, move out”. Or vice versa, where he's like, “I'm going to move out, cover me!”. And then the other guy said, yes, I'm covering. And then I think at some point we went a little bit too far with this. Like we had everything turning into like a 3 or 4 line conversation. And then I was like, all right, Brandon, can you like, tone this down a bit? We don't need, like, you know, this Shakespearean play as he, you know, reloads his gun or whatever, *laughs*.
Tommy:
Literally be dead halfway through the animation, you know?Wilson:
Exactly. That's what I was always having. Like, they would be like, “should I do this?” And he's like, “yes, you should”. And he's like, “okay”, and then, you know, they would never be able to to finish their conversation. They get killed every time. So, I mean, we did have fun experimenting with that, but we did end up having to shorten it a bit because, yeah, like the game is just too fast.But yeah. I think another really important thing with, with you is just like, like when the player is doing all this stuff, they're not really aware. They're like, you know, they're running around or like punching walls and stuff. So we had to have lines like there are varying degrees of, of, suspicion for the enemies, when they hear something or see something.
So like, yeah. Brandon wrote like I think probably hundreds of lines for these where it's just like, I think I heard something or it's like I probably heard something and it's like, I definitely heard something. Or it's like, you know, that's a tense time. I've heard something I'm very angry at.
Because we had to make sure, like, like players would understand that, like, if they're running around and the enemies are angry that there's something actually happened that they didn't just, like, see you through the wall kind of thing. So we'll help, like, a lot with this kind of stuff.
Anecdotes
Of course, it's all well and good us sitting here and celebrating how great everything turned out in the end. So I asked Wilson and Jacob about all the hiccups, the happy accidents, funny bugs, and bits they smoothed over to make sure the game got over the lined. And like any game that successfully shipped, there's a bunch of fun stories to wrap things up for us.
Wilson:
So I think one thing is, originally, like, there was one time where I tried, I found all these, like, military field manuals on the internet, and I was like, oh, man, what if I, like, implement all these, like, realistic, infantry tactics?But yeah, it goes back to the Mike Tyson thing where, like, everybody has a plan until they get punched in the face, kind of thing. So, it just ended up being these really cumbersome, you know, behaviours that we never actually get to execute. And, it was just more fun making the AI react rather than than doing these really elaborate, schemes.
And then, yeah, another funny, bug that I guess. Yeah, a funny bug that we shipped with was, we accidentally, made the enemies interpolate their aim while they're aiming towards you, but that interpolation never actually reached its target. So that actually created a paradox where enemies could never actually hit you while you're moving. Attributed to them. But that just makes it more fun. So we just hit it with that bug anyway.
Jacob:
It’s kind of Doom-esque. I think that mechanic. That *mechanic*, it's not about…Tommy:
Yeah, it's a feature! It's… it's like anime running, I think. *laughs*Jacob:
You're just invincible. As long as you're got those arms out doing the [mimics anime running]. Yeah. That's it. *laughs*I mean, it's just, fun fact, I guess, that, in the late, mission, horizon HQ, where you're, you're assaulting the bad guys HQ. There's a boss fight at the end where you're fighting a helicopter. That was actually supposed to be scrapped. We felt that it just didn't hit, the right sort of design philosophies and it just didn't feel good.
To play, but as we had scrapped it and already come up with, like, a good sort of fitting solution, internally, we had a discussion with our publisher, and we sort of, reintroduced it after the fact that we had already found a solution. So, we sort of went back to the drawing board and just kind of like, what should we do with this Like, how can we make this work? So we ended up merging the different mechanics of putting the helicopter back in the fight. But we put one of these. We we have, rings that pop out of the ground that are offloading the CEOs and the Lazarus brain. To the cloud. And, I just took one of those, and I slapped it on the helicopter, and I made him bob and weave like crazy, and, thanks to the brain now being on the helicopter, we constantly had, a UI element pointing towards it, so we would never lost it.
And we could run around the arena and do other things as we were fighting it at the same time. And it's pretty intense, like compared to what was there before. So that was like one of those cases where, this like, pretty awkward situation in production actually turned into something that was pretty cool in the end. So I'm actually very happy with how that came out, even though it is kind of, a little bit different from what what's happening in the game normally.
Wilson:
Nice. Yeah. Oh, I just wanted to add, I mean, speaking of, I guess awkward fixes is, so we have this boss, called the Flesh Golem. And he's a really big guy, and he can only melee attack, or he can do a grab, but he's mostly a melee enemy. But the arena we had him in was really cramped, and we were struggling to because he wouldn't fit through these doors and we were having all these issues with navigation.So one thing we decided to do is we just made the entire arena destructible. And so when the, when the golden spawns, we set all the arena geometry to not affect navmesh. So the golem just tries to walk through it, and we just made it. So when he touches any of these, the wooden panels that make up the level, they just explode.
Tommy:
Which looks awesome, by the way!Jacob:
He's just, like the Hulk, smashing through the level and just crashing everything. It's awesome. And he also grabs you through the level as well. Like, he's got a tentacle attack to mitigate the player's movement abilities because the player can just jump up and balance on, like a thing thing that he can't reach.
So we added an ability to him where he got a little tentacle thing that comes out of his hand, and then he launches it towards the player. It doesn't even care about the the level geometry that's in front of it either. So it just goes through like a bunch of walls, grabs the player and pulls him towards him. So that was a fun little thing. As well to like, mitigate, or fight the player's abilities head on like that.
Closing
Trepang2 delivers on its premise of being an explosive and chaotic shooter that harkens back to an era that many of us remember fondly. A 6-year journey that started out with one person and a fun idea. But like anything it requires a lot of smart ideation, and a lot of creativity to work through the many challenges faced. After launch in 2023, the ‘Bladekisser’ expansion launched in the summer of 2024 added new missions and new weapons to great acclaim from players. I have a sneaking suspicion this isn't the last we've seen of Trepang Studios.
As we wrap up, a huge thank you to Wilson and Jacob not just for taking the time to sit down with me and discuss Trepang2, but for all of the additional resources they provided us. They recorded the in-engine footage for us to use in the episode, and I it really helps bring the whole thing to life. So once again, a huge thank you to the team - and be sure to go and check out the game itself with Trepang2 available now on Steam, PlayStation 5, and Xbox Series S and X.
Premium Unlock: Full Interview
Becoming a paid subscriber to
to watch the complete 50-minute interview with Wilson and Jacob, going into much more detail on the various topics covered in this episode.Keep reading with a 7-day free trial
Subscribe to AI and Games to keep reading this post and get 7 days of free access to the full post archives.