Getting the transcript
Reading the captions from YouTube. A video nobody has opened here before takes 10 to 30 seconds; this page fills in on its own.
Getting the transcript
Reading the captions from YouTube. A video nobody has opened here before takes 10 to 30 seconds; this page fills in on its own.

Chong-U — AI Oriented Dev · @AIOriented
This video has no Most replayed graph yet: YouTube shows one only once a video has enough views. These are the moments viewers replayed most in Chong-U — AI Oriented Dev's most watched videos.
Most replayed moment at 14:17
2.8x that video's typical replay level
what you want to do here is just tell your AI to extract out the character from the output image, and then relay it out into a proper grid that you see over here, okay? And once you do that, then you have an evenly spaced out spreadsheet. Okay, so because, this is really important because if you don't
Said at 14:11
The graph counts replays. It does not show where viewers stopped watching.
Words
4,509
Runtime
18:11
Speaking pace
248wpm
Reading time
19min
248 words per minute, above the 201 75th percentile of 349 measured videos. That distribution comes from the 349-video hook study.
Opening (first 30 seconds)
Hey everyone. What you see here is a little pressure washing game. And the aim of the game is to basically get through the entire level by clearing up the entire space before the time runs out And if you make a mistake, such as watering the plants, or shooting the owner's pets you will end up losing the game. if you do a good job, you will be able to get through the level and then continue to earn more upgrades Now this entire game that you see here was built completely with AI. This includes the game logic, the character models, the environment art, As well as all of the other systems that brought this game to life. I posted about this
124 words, the words spoken in the first 30 seconds at 248 words per minute.
Free, no signup. See how the first 30 seconds hold attention, with rewrites.
Sentence shape
| Measure | This transcript |
|---|---|
| Sentences | 199 |
| Average words per sentence | 22.7 |
| Longest sentence | 134 words |
| Questions asked | 14 |
| Sentences containing a number | 7 |
Most used terms
Filler phrases
70 in total: like 28 · actually 23 · right? 8 · basically 7 · kind of 3 · I mean 1.
A literal whole-word count of the same phrase list the Prepublish browser extension uses, so a phrase inside another word is not counted and a phrase used in its ordinary sense still is. It is a count and not a judgement.
Run the check on the words above: where attention is likely to drop, with a rewrite for each weak line. The free check shows the scores and the one issue costing the most.
What this transcript is
Every word below is the caption track YouTube publishes for this video, pulled from the video itself and reproduced unchanged. It is not Prepublish's writing, not a summary, and not a re-transcription: it is the video's own published captions. English captions, published by the channel, in the video’s original language. Source: the video on YouTube. A channel that would rather this page did not exist can ask for its removal through the contact page, and it is removed.
Hey everyone. What you see here is a little pressure washing game. And the aim of the game is to basically get through the entire level by clearing up the entire space before the time runs out And if you make a mistake, such as watering the plants, or shooting the owner's pets you will end up losing the game. if you do a good job, you will be able to get through the level and then continue to earn more upgrades Now this entire game that you see here was built completely with AI.
This includes the game logic, the character models, the environment art, As well as all of the other systems that brought this game to life. I posted about this game on X, and as you can see, it got quite a lot of attention. and just to show you some quick stats over here, since posting this game, I've got nearly a thousand players spending about two minutes in this. Now, this is all made possible thanks to Anthropic's latest release, Claude Opus 5.5, And if you actually just search you notice that people are no longer posting about benchmarks, most of them are just posting about the kind of cool 3D games that they are able to create now with this new model. but rather than talking to you about how it works in game engines, this was made entirely with no engine.
It was just using something called WebAssembly and WebGPU. and this allows me to run this game entirely on the web, so you don't have to do any installs or anything like that. It has also allowed me to very quickly ensure that this works for mobile devices as well. Because it's so easy to distribute, I've been putting up my games up on WaveDash, And anybody can just go on right to the website And play it right in their browser, whether it's on their desktop or on their mobile phones.
I really like this workflow because it allows me as a solo developer to be able to very rapidly create games and have a way to host them, distribute them, and allow many people to play it. so in today's video, I'll be walking you through the entire process that I took, starting from how I brainstormed this game idea, all the way to creating all of the different art assets, including this character that you see over here, creating the initial prototype and then finally landing onto this final version that you see over here, which I think is a lot of fun and I'm getting some really good feedback about so if you've ever dreamt about building games on your own and then having a platform for you to be able to ship it This video is for you now, before we jump right into today's video, If you're interested in keeping up with the latest news in AI game development, you wanna check out my free insiders club.
I send out regular emails that cover my thoughts on the latest news and experiments as it relates to AI and game development you'll also get access to free resources accompanying many of my videos, including asset packs, prompts, source code, and various other things that will help you on your game development journey. but if you wanna take things a step further, you might wanna check out vibegamedev.com. This is my all-in-one toolkit to help you to be able to design, build, and then ship that game that you've always dreamed about, thanks to the power of AI it contains various resources to help you on your journey, including this full set of game development agent skills that I constantly update step-by-step tutorials to take you from a blank screen all the way to a finished project members also get exclusive access to resources including full project source codes, prompts, assets, and everything that I took to build and ship games using AI it also includes a Discord community.
It also includes an exclusive Discord community many of whom have leveraged the power of AI to be able to ship their very first games, so head on over to vibegamedev.com. Link is in the description below. And without further ado, let's get started with today's video where you will learn how I went about building this entire game that you see here on screen All right, so the very first thing that I'm gonna talk about is something that a lot of you have asked me about and that is how much did it actually cost to build this game.
And let me just tell you something. I don't really keep track of the cost. Part of the reason is that I've never found myself having to do things like having multiple, Claude subscriptions or whatever. I'm basically on the Claude Max plan. I recently updated from the 5X plan to the 20X plan because I've been doing a lot of experimentation. But, but I do know that this matters to a lot of you, so I asked Opus 5.5 to actually generate this information for me. so from the beginning, it took about six and a half hours in total.
And it's not just me sitting in front. Actually, what I did was that I just fire off the prompts, and then I leave it in the background, and then I check things as they are, And I just come back whenever things are done. and like I said, I'll be showing you the exact prompts that I use, so that if you want to go and replicate it, you can do so yourself. but here's a bunch of things that it generated, okay? And I don't know which of these matters to you, so the information's here.
I make use of several sub-agents, but one thing to note is that you may see that there are a lot of sub-agents here. Now, this does not mean I run twenty-one sub-agents simultaneously. What I tend to do is that For any given task, I'll usually have one main agent and then If it is parallelizable, I will then ask it to create one or two other sub-agents. the other thing that I do is that I have a lot of sub-agents running in the background to capture the prompts that I use, as well as take screenshots This effectively generates this page that you see here, which I often use to go through, these videos when I talk through them. and you can see right here I used the Sonnet five agent just to do the commits, the capturing of information, taking screenshots and stuff like that.
But then for the main task of building the entire game, you can see I use Opus five point five. There are a bunch of other things that I do use. I made use of Triple3D file for my image generation and ElevenLabs, And just scrolling up, by the time it took me to wavedash it cost me about two hundred and thirty-three dollars at API pricing. And then everything so far, because I've been continuing to add more stuff and, do some bug fixes, it's roughly double that price. but then this is just an estimate.
Like I said, I'm on the Claude Max plan. I did not use any resets, and I've still got plenty of usage. All right, so that's really about the cost. I just want to get that out of the way because I get this question a lot, and most of the time I have no idea myself. So how did I go about doing this? I do have a starting boilerplate that I have been using to build all of my games that make use of WebAssembly, and WebGPU, as well as Rust. this is a template that I use that effectively has all the skills that I need, in order for me to build my game.
And all these skills that you see over here are part of vibegamedev.com which I continue to update every time I make new videos or I make new projects and like I said, the thing that I've been working a lot for browser-based projects right now is just using WebAssembly and WebGPU. and a lot of people have also asked why, you wanna make use of something like this. I was inspired by a post by Eric Provencher. he's a developer at OpenAI.
And he recently posted about why you might wanna make use of WASM and WebGPU, given that, most of the people are not looking at the code anymore. If you're using something like Three.js, which, provides a lot of like human-readable APIs, LLMs are really so good right now that these abstraction layers may not be necessary. It's obviously a lot more nuanced, and there are a lot of different reasons why you wanna make use of it.
But for me, it's been working very well, and I'll probably do a much longer video about this entire pipeline and the ways that it works really well for me. But in essence, it's allowed me to get really good results, in the browser as well as it running on mobile devices and the iteration speed is really quick for me. all right, so what did I do here is I said, "Let's make a game that's inspired one of those casual games like mowing the lawn, but I wanna focus on pressure washing driveways." And then I gave a description of, this, being like mission-driven and you're not supposed to damage things, you get paid.
And I said, "I want to focus on a vertical slice on this pressure washing element. I want it to be fun and juicy." And then this is where you can see I said that I want it to be a web browser game, WASM and WebGPU, no need for any Three.js. And then I want to adopt my game dev best practices, which is the process which I always use, which is to not try to build the characters and the environments within the game itself, but to have it separated out such that you have your environment and your props as well as your character in different Blender files and only then do you bring it into the game once you're ready. so the first thing I always do is I always generate different mock-ups. so you can see here, once I've given this prompt, I always add, "Do you understand?" So that the, AI just pauses for a while to make sure that it understands the direction I wanna go. and I say, "Let's use GPT Images 2.5 to create the mock-ups for the game." And you can see here that I did advise it that I wanted it to work for both mobile as well as desktop.
So you can see a bunch of these different, art styles and how it would turn out if it looks on different devices. And I've always emphasized this. It's really important not to skip this step because you want to align on what direction you want to go to before you ask your AI to go ahead and just build it. I'm sure you see all these one-shot videos where people just give it one prompt and leave it running for twelve, thirteen, fourteen hours, or even two days. but the problem is that you don't really, have much creative control over these.
They are really great tech demos, and they serve a purpose, but if you're trying to build a game that's in your head, you always want to make sure that you are aligned with the AI about like what exactly. It's always in the details. Stop one-shotting games. They serve a purpose to show capability of the model, but if you're trying to build a game that you're trying to call your own, you definitely do not want to one-shot it.
So after this step, the AI actually said to take a mixture of some of these styles and then put it together, and then build the game based on that. So that's why I said, "Let's go with your call." And now this is where I do things in parallel. what I want to do is I want to make sure that I start to create the character models and then start to do the environment, and I want it to be done separately because, like I said, I want my character models to be in its own file, so that I can then rig it and then animate it, separately.
And then my environments would be in a separate file because these are more like static props that don't really need much animation. this is the way that I like to set up my process, and this allows me to create and run both of these at the same time. So you can see here, before doing any modeling, given that we have some concept art, what we wanna do now is to create a turnaround sheet for the character and then- use an image generation model to create any, additional props or angles of the environments that you might need.
So this is what, a turnaround sheet looks like. So basically it has a front side and back facing. And then for a different use case this time, I actually make use of Tripo, which is very similar to another tool that I use called Meshy, but I wanted to try something different this time. it basically works the same. You just need to supply a bunch of reference images, and it also supports multi angles. and I don't use the website, I make use of the API to, pass it in.
So you can see here with this turnaround image, it was able to generate the model, and then, I made use of Blender to then go ahead and rig it, so that it's able to animate. and then for the other props with the reference images, it was able to determine that it needed to create the trees, and then to create the cars, and then you can see this is a first pass version of what it looked like, and it's already looking really good.
So I think Opus five point five is excellent at getting this right. If you actually look a couple of weeks ago, when I was already trying to do animations, they required a little bit of fixing. weight painting and moving the bones around. in this case, Opus five point five got the rigging and, weight painting correct, pretty much on its own. I didn't have to do anything. This just, worked right off the box the first time, and it was really good. so let me just show this to you here.
So I kicked off these two things in parallel using sub-agents, So these were run in the background, and then I said, "While these are happening in parallel, I want you to work on building a gray box level gym that will allow us to test out the key mechanics of the pressure washing. Use placeholder art." And I said that I want, to focus on the mechanics. I wanted the core gameplay to feel just right. Okay? So I want the exaggerated motion, I wanted the physics of the water, and then I said, "Take a chance to think about how the art direction can be fulfilled using shaders, post-processing, and effects." all right, and so let me just show you where it basically got to.
So you can see this is actually the gray box level. All right? Because these were still running in the background, but it was able to create this, And if you look very closely, it's just using primitives over here, and it was already able to get really far with the, water spraying mechanics, which was the key thing for me 'cause I really wanted it to feel fun and it was able to do that. So this was all done, by Opus 5.5, and having all three of these things running in parallel is effectively what contributes to having all these multiple sub-agents running.
So Let me just show you the gray box that you have over here. So it managed to capture a video of it, and this is all done automatically by my background sub-agents. And it just makes it a lot easier for me whenever I'm doing these videos because I can just, look at what was done. I don't have to record it myself. And you can see the pressure washing mechanics are really looking really good, right? So by having gray boxes, it means that you can get started on play testing the game, making sure it feels right while the art assets are being done in the background.
And this is how game development works, right? You always have an artist, you have a gameplay prototyper, and then you have your environment artist. But now with thanks to the power of AI, you can run all these things. You have this entire team at your disposal. with that done, the next thing I did is said that I want to create a new scene for previewing the imported, models So that it allows us to check, how things are.
So if you actually look over here, you can see this is the mechanics gym that you have over here. And then and so with that, It was able to preview the model, so then it is able to adjust the different, size and weight such that when it animates, it actually looks correct. not only was it able to import, the models and make sure that everything loaded correctly, but you can see here it was really focusing on how to make sure that the spout, of the hose that you see over here is connected to the rig of the character.
And this resulted in a really good result That you can see in the final game is that, when he's walking around without the spout, he's just holding it, but when he starts shooting it, you get the fact that h- the spout is now connected to his hand. And this was not something that I expected it to get right off the bat, but it was able to do so because it had all of these test scenes as well. in fact, one of the things that I actually did was I said, "Hey, I'm actually really interested to understanding how you got the jet aiming to work." And I asked Opus 5.5 to create this page for me.
And this, talks a little bit about how, it works, right? because everything is running in Rust and WebAssembly, this can happen really quickly. It happens at one hundred and twenty hertz. That's one hundred and twenty times a second. And so this allows it to actually compute, a lot of what is required in order to make the models, turn and face the right direction, allowing the water to spray correctly automatically. And you can see it even went ahead and showed me the code snippets.
And I know you're thinking to yourself, "Wait a minute, this is all about AI coding. Like, why am I looking at code? Aren't we in twenty twenty-six now? Nobody looks at code." And yes, you technically don't have to, but I think it's really important to use AI as a tool, not just to blindly, create, whatever is in your head and then, just leave it there. But I think it's really important to sometimes just take the chance to use it as an education tool for yourself.
Because if it were me trying to build this from scratch without AI, I would be, searching and researching, different ways to do it, But this AI helps you to, get the end result that you want and then allows you to then look back at how, it was done, right? It's like a teacher. right? It actually shows you how it does the, calculations, and if you want to go into detail, you can actually look, into this. And it does it in a pretty accessible way.
It says "Hey, this nozzle is actually just a spring, and all it does is that it Tries to swing it from one point to the other where the cursor is, right?" and then because it's a spring, it like replicates what the water would look like. This was actually just very informative for me, and this is something I'm gonna try to do a little bit more because it really gives you a view into how things are done. So that was pretty much it.
The rest of the things that I did was really focused around polish. So you can see, I had, this issue here where the hand and the hose was not completely connected. I had a couple of other issues where I didn't really like how there was just this gray background. So I actually said, "Create a sky box," and you can see here. So it just uses an image generation tool to create a sky box And this automatically gives your game a much better atmosphere.
So this is the other tip that I have for you is, always add a sky box. It just makes your game feel way more polished just by having a single image that's sitting at the back, that is wrapping around your world. and then the next thing I did is I said, "Hey, the world was actually pretty bare." So if you look over here, previously, it looked like he was just living in this random, grass patch. So I said, fill up the rest of the neighborhood," and you can see that this is the neighborhood that it was able to generate.
And I think this just leaves a lot of room for, creativity. you can create different neighborhoods, and I think that's the plan I'm gonna be doing as I continue to work on this game. so that was pretty much what, got me to where we are with this game. The next stage that I did as I prepared to release it was focusing on what you call a first-time user experience or FTUE. And the idea here is that you want to, introduce the game mechanics in an easy way for your users.
So you can see, the difference here is that for this kind of leaves, you can use this fan head of the spout, which kind of sprays around. But then actually I want to teach people that you are able to use a jet spray for slightly more difficult, to remove stains. And you can see, as I try to, switch to it here, it doesn't allow me, and it actually told me that I will need to, pass this lesson. All right, so I'm just gonna quit the tutorial, and I'm just gonna show you what I mean over here.
So you can see as we go to slightly more harder jobs, th- these are oil stains over here. You're not gonna be able to clean it with just this jet. So you have to switch, you have to switch to this jet-powered mode, and this will allow you to work better. Now, these are things that, came to my mind. I just continued to iterate on it, and the idea here is that I said I wanted a tutorial, in addition to just this how to play thing.
I can say we need a, much more tailored first-time user experience. So I want to have a simple first level where they have the fan, and then they have to learn to avoid the stuff. And then after that they have the oil spill, and then the third level of the timer. So that basically just teaches you the entire game, and it does it in a manner so that people don't have to start reading things on their own. So then the rest of it was really about just polishing it So this shows you how I got from this, gray box prototype that you see over here, running All the way to this full game that has a job system, that has level progression, and I'm really impressed.
And if you actually saw, This took the AI basically about six hours running in the background. I was able to do other stuff while this was running, and, Thanks to the pipeline that I have and the skills that I have, I was able to steer it in the direction that I want it and get it to this level of polish. given the attention that this game has had, I'm thinking of, putting in a little bit more content for this game and then, bringing it to completion.
So it's one thing to create a tech demo. Another thing to create like a vertical slice of this game, which already has, different jobs, different upgrades. But I'm thinking of taking this a little bit further. maybe having different neighborhoods, having even more jet types maybe having a lot of different modes. Let me know what you think about how I should take this game further, but the main thing that I want to show you is that Thanks to the power of AI, you are now able to get from an idea in your head into a playable version.
But not only that, into a version that you can then host and then distribute and get many other people to play it. all right, so that's it for today's video. I really hope you enjoyed it. if you found this video in any way helpful, useful, or inspiring, Give it a like, hit that subscribe button, and don't forget to turn on notifications so that you always be the first to know when new content like this drops. Now, if you want to learn more about building games, but not just browser games, you want to use something that's a little bit more traditional like Unity or Unreal Engine, I actually did a whole video about how I built the same game in both Unity and Unreal using AI tools without, writing a single line of code.
It was able to handle everything within both editors, and it gave a really good result. it uses pretty much the same workflow that I use over here creating mockups, generating assets in Blender, and then importing into the game. but there are a bunch of useful tips and tricks that you might want to know about when working with Unity and Unreal. so if you're interested in that, definitely check out this video that you see over here.
All right? So till next time, I'll see ya.
The words are the caption track's own and nothing is reworded or re-transcribed. Paragraph breaks are placed between sentences so the text reads as prose.
Free tools for your own script: paste a draft and see where it stands before you record it.
Paste your draft and see where viewers are likely to drop off, with a rewrite for each weak line.
Paste the first 30 seconds of your own draft for a hook score and rewrites.
Check your draft against YouTube's advertiser-friendly guidelines before you record it.
Read this channel's public videos and transcripts, and download a writing brief for it.