Want To Make Games? This New Tool Changes Everything...

When Claude and Chat GBT first came out, I like a lot of game developers thought this would be a gamecher for solo devs and also small teams. After all, solo devs and small teams, they have to be scrappy and do pretty much every art form under the sun. AI coding can be a huge help in that regard. But once I started using AI coding tools, I started finding myself spending more time debugging the code it was spitting out than actually working on the game. See, the problem is obvious. These AI coding tools, they don't really understand your project. They don't know how the sound, the music, the 3D models, the 2D sprites, the C# code, the prefabs, how they all work together. In a way, they're coding your game isolated on like another planet while you're here on Earth trying to tell them about a project with a keyboard. That's really not helpful for solo devs or small teams. Really what we need is a team member that fully understands the Unity project. So there's this new tool called

Bezi and it's basically an AI team member. It fully understands your Unity project and my mind was blown once I started using this tool. This really is a game changer for solo devs or small teams and I wanted to show it to you. So what I'm going to do is I'm going to take four totally unique tasks for my game and I'm going to let Bezi help me complete these tasks. And at the end of this video, I'm going to show you how much time was saved. And yes, this video is sponsored by Bezi. I don't usually do videos like these, but Bezi, they came in clutch this year. They sponsored a ton of my podcast videos. And honestly, because of their support, my studio and the podcast have stayed alive during a difficult financial time. So, I'm incredibly grateful and I wanted to share just how effective Bezi is at, well, being an actual team member for your project. Okay. So, Bezi understands your entire project. Somehow, like magic, it'll scrub through and read through your entire project hierarchy, all of your code, your art, your sounds, and with mind-boggling accuracy, it

could tell you exactly what your project is, what the mechanics are, what the loop is, and even like what the characters of your game are. Like, watch this. I have this game called Brave Boy I'm working on. Now, here's the game on Steam, okay? And and Bezi hasn't read the description on Steam. The description says, "Pick your breed and build your clan. An open world adventure about a pup from the Scottish Highlands destined to lead a rebellion against the evil cat king of England." Now, let's ask Bezi just to scrub through our assets in Unity and describe to me what the game is. What's the mechanics? Who are the characters? Let's see if Bezi could figure it out just by looking at the project hierarchy. [Applause] [Music] Brave Boy is a 3D adventure platformer where you play as a brave dog on heroic quests in a charming fantasy world. I mean, how does it even know it's a charming fantasy world? That's what's crazy to me. It's probably scrubbing through the metadata or the names of the assets or something. Based on my exploration of the project, here's a comprehensive breakdown of what this

game is about. This is a demo version of Brave Boy, a 3D action adventure platformer featuring a dog as the main character. The the game combines classic 3D platforming mechanics with pack management, quest systems, and combat elements in a fantasy medieval setting. Here's what's so weird about this description. This is more correct than the Steam one. I need to update the Steam description because it's more of a 3D platformer. After I released the demo, I was like, "Yeah, it's actually more of a 3D platformer than open world." So, this is more correct than the Steam description. The first time I saw this description, my brain felt like it popped. Everything about this description is correct. It understands every asset's purpose, how the assets fit into the rest of the game. So, guess what? With that understanding of the project context, you've effectively got another team member and it's a magic wand for solo game devs. This really does. I mean, it changes everything. And look, I get it. This kind of stuff can be scary cuz, you know, it can take jobs. It can change game development drastically for everybody, but I think

it's here to stay. I don't see how it's not. And it's really, honestly, it's just so effective for solo developers who can't afford a bigger team. And here's the truth. I'm committed to not using AI art tools in my studio like 2D art, 3D art, sound, voice over, music generators. I don't use those things. And Bezi is cool because they've made a public commitment to only focusing on AI code, making them a trustworthy tool for artists and developers in the gamedev community. So, with this magic wand, with this team member, I'm going to quickly show you some insane things that I can do for Brave Boy really fast. And I've been wanting to make these changes for the last year. I just haven't had the time because I'm working on another project. And frankly, they would take me about 8 hours that I don't have to create. And again, I'm not a good coder. Uh I'm more of an artist and a story person. So, this kind of stuff takes more time for me. So, the first thing I want to create is I want to create a potion that you can collect and it will temporarily make the player run fast. Number two, I've been wanting to fix a bug where the player, for some reason,

they can infinitely roll without any pause. I've also been wanting to create an editor tool that will randomly create islands in the sky. And finally, I've been needing to do some research about why Brave Boy is running at 20 FPS on the Steam Deck and how I could potentially fix this. So, let's do these four things and see how quickly we can do it and how Bezi stacks up against more traditional, if I could say that, more traditional isolated tools like Claude or Chatbt. And remember, my goal is to spend as little time on these things so that I can focus as a solo developer on this project on what I'm gifted at and my forte, which is art, music, and story. I want to spend as little time doing code and editor tools and debugging. Bezi is going to handle the heavy lifting of this development stuff. Okay, so let's tell Bezi that we need a potion C# script and we want that to be collected by the player. Please create a potion script that can be attached to a

spear in the game. This potion will then double the player speed for a set amount of time. Please use the particle effects being used by other abilities to make the speed ability look cool. I don't need any other scripts. I just need a potion C script and also an adjustment to the the player controller script. So, I'm just letting it know to keep it simple, just change the player controller and also create a potion script. And just as this is being written, notice that we're in agent mode. Ask mode is much simpler. Agent mode allows scripts to be adjusted within Bezi. So, I can just accept the changes. It'll create the potion script for me, put it in the project. It'll also adjust the player script. In ask mode, you kind of have to copy the code over in Visual Studio. So, I'm going to stay in agent

mode here. So, what it's done here is it's looked through the project hierarchy, kind of figured out how it can connect various scripts together and make the player script work with this new potion script. All right, I'm just going to scrub through this potion script. Make sure it looks good. Good, good, good. It's using ontrigger events. Yep. Good. And it's going to make an adjustment to my player script here. It tells me the changes. base speed has speed boost. Good. Particle system, good, good, good. Everything looks good. It didn't really change much. It says it changed four lines. Um, it didn't remove any. So, no big changes here. Really important to scrub through and look at the code. You want to be wise and you want to be responsible with AI code and not just let it add spaghetti code. So, this looks reasonable. It also tells us the various visuals it's going to use. This is really cool. I created these a long time ago, right? And they're used in all my other abilities. It's just going to incorporate those into this new speed ability with the potion. So, I'm just going to go ahead and click accept all and it's just going to add those scripts to my project and also adjust the player script. All I got to do is go

back to Unity. It'll update that. So, now all I need to do is just create, I don't know, a 3D sphere for now. Let's just do a sphere here. I'm going to put it right here. There's my potion. I'm going to make sure that the sphere collider is a trigger. I'm going to add the potion script that it's been created by Bezi. And I can choose a collection sound. Let's just pick a random sound for now. We'll do an arrow shoot sound. Why not? Um, we're not going to worry about any other collection particles or anything like that. Not for now. I think that's all I need to do. Let's hit play and make sure it works. There's my potion. Let's grab it. Got it. And now we're running twice as fast. And you can see the particle effects are there. And it lasted for 5 seconds. My abilities still work. I can still use my sword. Nothing is broken. Now, obviously, it's, you know, just like any AI code, it's not going to be perfect, but in this case, I mean, it was perfect. I can adjust the sphere collider if I want to make it a little easier to collect. I can make the speed multiplier, let's say

10, and last for 10 seconds. And then I can just drag this over into my prefabs. And I've officially got a potion that I can place throughout the game anytime I want the player to be able to do a temporary speed boost. Kind of like a star in Mario. That to me is wild. Like how quick it was to create that. All right, the second thing I want to address is this dash ability. I can spam the dash ability and just roll constantly. There's no stopping. It should just do this. But if I hold shift, I can like spam it. Yeah. So, let's tell Bezi, let's be really straightforward here and really simple and just say, "Please fix a bug I'm encountering with my dash ability." It's called dash, but it's also like a sprint. I can just hold shift and the dash ability will continuously go.

Please make it so I have to release sprint before another sprint can occur and requires a bit of a delay after releasing sprint or shift. So what it's going to do is it's going to look at the context of the player and it's going to adjust accordingly. All right, I'll fix the dash ability by implementing a key release requirement and cooldown system. Let me examine the current dash logic. And so you can see here what it's looking at. It's looking at our player input, kind of understanding how the input is working in the game, and then it's also looking at the player controller. So, it's trying to see what's going on with the input system and also the player controller. Whereas with something like just using Claude or just using Copilot in Visual Studio, it's going to have a harder time understanding how the player input system is used. I might actually have to describe how the player input system is used. In this case, it's just going to automatically know it needs to look at that stuff. All right, the dash bug is officially fixed. At least that's what Bezi is saying. It tells you the

problem. It tells you the solution and tells you the code changes. I can take a look here at the various code changes by selecting the code. Scrubbing through here and seeing where the changes are occurring. Dash key released is a new boolean that's created with also a cool down and also cool down duration which is public meaning I can change this in the inspector. It knew that I would probably want this visible in the inspector. So, let's scroll down and see what the changes are. Dash cooldown timer is being put into the update. It's also checking to see if I've released dash. Okay, good, good, good. And probably somewhere down here is the function. Yep, for the dash cooldown timer. The only thing here is it's putting it in the update. And so, you want to be careful putting things in the update because that can slow your game down. But I don't know any other way to do that. Um, so I would I would have put it in the update as well. So I'm good with these changes. I'm going to click accept all and let's see if it fixed the problem. All right, let's go ahead and dash and hold shift here. Okay, I've got

shift pressed and it's not I'm not able to spam it. I have to release shift, wait, and then press it again. Much smoother. Makes way more sense to the player. Feels great. Is it screwing up any of my other abilities? [Music] Nope. What about midair? If I hold shift works good. It's funny because that ability I did not include in the demo because it was broken and I didn't have time to fix it. So, that's fixed. Let's see if we can create an editor tool that allows the the spawning of islands. I don't know. I haven't even tried this yet. Let's see if this works. There's various prefabs called grass hill around a variant. Let's see if we can create an editor tool that will randomly create islands floating throughout the world. Please now create an editor tool that will place ground piece prefabs randomly throughout a

scene. Let's let's do this. Let's incorporate some some logic about the player. Please ensure these are placed at height levels and distances away from each other that are high enough for the player to jump to and midair dash to. I don't know if this is going to work. Let's see if it works. So, basically, it has to understand the player controller and how far I can dash and how high I can jump. Hopefully, it can look at those calculations and say, "Okay, the player can dash this far. the player could jump this high and make these random islands traversible. We'll see. All right, so it's written this script for us. Let's click on the script and take a look at what's going on here. Various ground prefabs we can choose in the editor. Various placement settings, height settings, great distance settings. So, these are all things I can control with the tool. Collision avoidance. Interesting. So, it's going to look for ground layers and stuff. Avoid overlapping. Cool, cool, cool. It looks great. And the cool thing about editor tools is let's say that this is a

really clunky script. Let's say it's spaghetti code. Well, it's an editor tool, right? So, it's not going to affect gameplay. It's it's a one-click thing, so it might take I don't know if it will, but even if it did take a while to create this level, it's certainly not going to slow our game down. So, that's why I love using AI for editor tools because it speeds up development, but it doesn't actually slow down the game. Let's just click accept all. And I'm just going to see what happens when I click this with this current world. All right. And like clockwork with any AI tool, we're getting some errors here. No AI is perfect. Let's take a look and see what this is. The internal editor utility does not exist in the current context. So, let's go ahead and just tell Bezi this is an error. I'm just copied and paste it and hopefully Bezi will be able to figure out what's going on. I'll fix the error by replacing internal editor utility usage with the proper Unity. So, basically it missed the fact that I'm using Unity 2022. So, now it's going to make a small oneline adjustment right here. So, that's what it's going to use now. Tells me the problem, tells me the solution, and let's go ahead and accept all. All

right. So, it should load up and remove the errors. And now the tool should show up in my tools menu. There it is. Ground piece pler. So, what's crazy is, and I feel kind of guilty about this, I don't know, but like you you'd buy a tool from the Unity asset store with these kind of functions, but it was written in like 5 minutes. So, let's go ahead and auto add. Look at this auto add button. I didn't even ask for this. Let's see if it could do it. Wow. So, it already knows. Wow. How did it know that? It already knows which prefabs I'm probably going to want to use. That's wild to me. Let's look take a look at these other settings here. Number of pieces 20. Placement center 00. Placement area size. And look, these values are used to ensure platforms are reachable by the player. So, I could change these values if I want to. Let's click the button. I'm kind of nervous. Oh my word. Look at that. It just created this. What? What are these? Let's figure out what's going on. So, it looks like there's an issue here with one of the prefabs.

No, that's actually correct. That material is missing for some reason. Let's see if that's my fault. Material Rock Hills. And I think that's because the prefab, for some reason, the prefab itself doesn't have that. There it is. Okay, that was my fault. But oh my word, look at what it generated here. It created this random level. I mean, I Let me go ahead and bring the controller over and see if I can traverse it. So, wow. I now have this beautiful open world environment that can be traversed. My goodness, this is incredible. But, I mean, it's using what we've currently created. So, these assets that we created are pretty helpful because you could place them and it creates a really cool looking world. Well, I would say it's a really good start. Okay, final thing. We have a game that I'm getting reports that it's running really poorly on Steam Deck. And in fact, I could show you really quick. So you can see the game right now is running at 22 frames per second. That is

not good for a platformer. For a really simple 3D platformer, I'm getting really slow performance. So we're going to t we're going to try and aim for 60 fps. Let's figure out what's going on. All right, so it's running slow on Steam Deck. I'm going to ask Bezi just random like just a very generic question. Why is this scene running at 20 fps on Steam Deck? That is too slow. I want it to run at 60. Give me some recommendations. And that's it. Let's see what it does. Okay, so it's going to give us a list of all these different optimizations that we should be considering for our scene. Massive object count, generated ground pieces. Now, this isn't true because I'm going to actually remove all those generated pieces. I'm just worried about the demo, right? Which is way less pieces, but that is true. This is a number one cost for sure. We have a foliage issue. 100 plus interactive foliage objects that have individual

physics scripts running every frame. LOD groups, use LODs. We are using LOD groups. Reduce shadow quality. This is probably one of the major issues. Shadows are very very problematic for Steam Deck. So, it wrote a script for us called performance optimizer. It's going to find various objects that are tagged as foliage and it's going to call them out. Interesting. Add performance optimizer to scene. Um, reduce URP shadows, mark all objects as static. Okay, let's do this intermediate one, this 30 minute adjustment, and then we'll do a build for Steam and see if it fixes our frame rate. I'm going to disagree with Bezi here and I don't think that the foliage is the primary issue. I actually think that Bezi is correct that shadows are probably our biggest concern. And if we take a look at our URP asset, we've got a max draw distance of 600. So, we're going to reduce that to Bezy's recommendation, which is 50. So, we're going to set that to 50. And you can see if you set it to

zero, there's no shadows. Okay. So, we're going to set it to 50. and see if that helps us. So, we have our shadows, but you can see they're not drawing until I get pretty close. No shadows over there. As we move closer though, you can see they start to appear. I feel like that's going to be a pretty good strong performance uh increase for our game. Now, I also have a suspicion. So, I'm going to get um Bezi to get some more research for us. And I've seen this problem with my other game, Twisted Tower. I'm going to ask Bezi what I should do. Should I export to Linux or PC for Steam Deck? Let's see what Bez's recommendation is. Oh, wow. For optimal performance on Steam Deck, you should definitely export for Linux. Okay, good to know. It even created a Steam Deck optimizer script. What is this doing? This Steam Deck optimizer is really cool. I think it overcomplicates things.

So, with any AI code, sometimes you're just not going to do everything that it's trying to do. I'm just going to enable VSSync and that's basically all I'm going to do and then also export to Linux and see if that solves our performance issue. All right, I'm I'm kind of impressed. We didn't hit 60, but with the changes recommended by Bezi along with a few others that I came up with, and we'll talk all about that in just a sec. Same scene, I'm hitting 53 frames per second on average. So, not 60, but almost almost. Now, sometimes it'll dip down. Right now we're at about 46 46 frames per second, but this is a huge boost compared to that 20 frames per second average I was getting in the previous build. So I implemented obviously some changes from Bezi. The shadow draw distance, the shadow resolution, the amount of lights available in the scene, I reduced that from five to two. I also per Bezy's confirmation of one of my suspicions, we

need to build to Linux. I also implemented occlusion culling. That was dumb on my part. It wasn't in the scene. So, that helped a lot. So, Bezy's recommendations resulted in, I would say, 20 frames per second boost along with my occlusion coloring changes, which brought it up another 10. So, overall, with just some recommendations from Bezi, I'm getting a major boost in frame rate. Now, it's obvious. It's obvious you got to have a brain as a developer. You can't just rely on AI. So, I used some of my knowledge, but also used Bez's recommendations, which that's the whole point of Bezi is to work with almost a team member, an agent that's going to help you out and understand the context of your project. So, my brain and then also Bez's brain working together, we're seeing a 20 to 30 frames per second boost in Brave Boy. Pretty impressive. [Music] All right, so what were the final results? How much faster was all of this

with Bez's help? Well, I've been making games for more than half of my life. So, I feel pretty confident just estimating how long these would have taken if it were just me. So, coding the new sprint potion, that would have taken me about an hour. Fixing the rolling bug would have taken me about an hour. Building the custom island generator, that would have taken me probably 3 hours. The optimization stuff would have taken me another 3 hours to figure out. All in, that's close to 8 hours. with Bezi. This took me about an hour. That's wild. That's an 87.5% increase in speed. Insane. Look, I I get it. I really do. You You got to be careful with AI tools, but I think they're here to stay. I mean, that that was so much faster than if I had done it by hand or even by using simply Clawude or Simply Chat GPT. And if you're a solo, look, this is a game changer for you. But does that mean you shouldn't learn to code? No. Honestly, if you don't know how to code,

I don't see how this can really help you. Because you're just going to get a bunch of code that you can't understand and you're going to try and put it into your project. That is not what I'm saying. I'm saying you need to fully understand how to code. And when you plug AI into it, it's a multiplier. And when you plug something like Bezi in that fully understands the context of your project and you plug that into your knowledge of how to code, it like I said, it's a magic wand. But it's important to I don't know just like with a team member look over the code. You got to really understand it before you commit it to your project. With great AI tools comes great responsibility. I think that's a quote from Spider-Man, right? So let me know your thoughts, guys. Let me know in the comments about what you think about these kind of tools. And if you're interested in checking out Bezi, check out the link in the description. As always, super duper thankful for you guys. Super thankful for Bezi for sponsoring this channel and the podcast. It just they came in clutch this year and I I really do appreciate it. Um yeah, let me know.