I made a post about balancing my game earlier, and everyone was like "you didn't make a game, loser, you're just lying". I guess you guys haven't got the memo yet that you can make a game really easily as a non-programmer now. Here's a quick tutorial.Things you'll need:>claude code subscription ($20 to start, $100 if you enjoy it; never ran out of tokens on the $100 tier myself)>an IDE like cursor(you don't need a subscription to cursor, it's just a convenient way to organize and view which files Claude has access to)>a subscription to an AI image generator(I use Leonardo AI, which gives you tokens you can use with any of the major image generators)>a hand-written game design doc describing the type of game you want to makeOnce you've got your materials together, you're ready to go. Next reply I'll lay out steps for organizing everything.(you can play my game here if you want an example: https://obligate-brahmin-games.itch.io/factions-of-ek-a-mausritter-fangame)
>>3997396The first step, once you've got your Claude Code subscription: ask Claude what to do any time you don't know what to do. It can walk you through what UI buttons to press, what tools are available, what subscriptions are worth the money, anything. However, there are still a few good things to know:Chunks>the fundamental structure of your project comes in 3 pieces: A handwritten design doc, a large number of AI-written design docs, and a changelog>split all the pieces of your game into chunks and subchunks. Chunk A is combat; chunk A1 is the to-hit resolver, etc. You don't need to know how the game will function code-wise, you just need to know what each system will look like to the player. Write your design doc by listing all the chunks your game will need for every single piece of it you can think of.>the basic procedure will be that you'll say "OK Claude, let's do combat" and then Claude writes a doc titled "combat design" , copies over the chunks from your handwritten document without editing it, and starts working on them one at a timeClaude procedures>Claude will have access to all the contents of whatever folder you point it at in the IDE; your whole game lives there. If Claude says it needs a tool like Python, let it install it.>When Claude adds a chunk, it needs to do 3 things: save the feature as a git commit so you can roll it back if it's broken; update the changelog so there's a running record of every edit; and update the docs to reflect the new feature.>Claude won't remember the whole design of the game- it can only hold so much in its head. Your job is to remember how everything fits together into a game, so Claude can focus on the technical stuff. >tell Claude to design your game in a bunch of smaller files that it can easily fit into it's context window rather than one godfile that will swamp it out>run Claude in "auto " mode by clicking shift + tab in the editor window so you don't have to mash spacebar to confirm
>>3997398Designing content>this will vary based on what engine you decide to use; but it's best to have a separate part of the game (mine is called the map editor) you can use to edit human-authored content without launching the game itself. I bundled all my content authoring tools together in this program that runs in the Godot engine as a scene>for art, you'll need to generate images of your sprites yourself using an image generator, download them to a folder, and tell Claude you want a procedure for cutting them out with Python and wiring them in. (whenever I tried to automate this with API keys the results were a waste of money)>for animations, you can either use in-engine animations (usually very primitive) or use Google VEO to generate a video, slice that video into keyframes, and recompile those keyframes as a spritesheet. I've never tried this, just heard about it.>Claude can make you any tool you need. I never edit text files myself, every change I make to the game is filtered either through Claude or through a GUI tool I had Claude make.Keeping Claude sane>the most important thing here is making sure the docs have a master index doc which will allow Claude to see where things sit in the code without reading the entire codebase. The chunks system helps with this by giving everything letter and number labelsTesting Content>this will be your biggest bottleneck. You'll need to launch the game and test almost every single feature Claude adds. There's no getting around this; Claude can boot the game to see if it launches or crashes, but it can't play the game for you.>one thing I've learned is that designing the game to have clear limits on the interactions between systems is essential to avoid making the testing stretch into infinity with every feature you add (combinatorial explosion)
>>3997399That's honestly just about everything you need to know. It's dead simple desu. I'm making a tile-based Qudlike, which is super complicated, so it's taken me 6 months or so to get the two dungeons of my Mausritter alpha; but if you wanna bang out like a jrpg, you can whip that shit in like 2-3 months and have a solid beta. There's all sorts of individual design decisions and such you'll need to figure out, but all of that depends on what kind of RPG you're gonna make, so I can't really help you there.I will say, anything with 3d or animations is much, much harder and more expensive. So that may still be a barrier; but RPGs are uniquely well-suited to this tech because so much of the gameplay is just math. Claude can implement a combat algorithm in like 10 minutes- it's the design work that's really the barrier, not the coding.I expect to see all of your brilliant never-before-seen games posted here by Christmas.
>>3997402>Qudlikekinda gave away the joke here
>>3997435Are you that terminally irony poisoned poster that wrecked this board by shitting up every new thread for years and years?
>>3997437>that terminally irony poisoned poster that wrecked this board by shitting up every new thread for years and years?You’ll have to be more specific than that
The only reason I play around in game engines is to make maps so this seems like it would remove the fun
>>3997442There's a guy who pops into every thread, ignores 3/4 of what was written, and shits all over everything the OP is trying to talk about. For instance, you implied that my post is a joke and I haven't really made a game, conveniently ignoring that I linked to the game in the first post. This poster always replies to new threads in the first few minutes and has been active since late in 2021. One of the main reasons I quit using the board desu.
>>3997443In what sense? I hand-painted all the maps for my Mausritter game. I'm using procedural generation for my next game though, I don't like designing levels that much.
>>3997437no, but please point to examples for me.>irony poisonedanon, you're using "Qudlike". i think the irony poison is in thine own cup. this whole thread is you attention seeking, trying to bait people into being bothered by "aislop". if someone wants to use an llm to make a game, they'd ask an llm.
>>3997483>accusing me of attention-seeking in the OP because the thread doesn't interest youYeah, you're that same guy. He used that phrase all the time. Unbelievable you've been haunting this board for so long.
I think ChatGPT works great for developing as well. It is great in the hands of the right mind. And worse than useless for those who cannot see.
>>3997509I've never tried it; it's probably roughly comparable though. My tech worker friends cycle between various ones at work depending on which update is best
>>3997511I haven’t tried Claude. I’m sure it’s fine, maybe better idk. I’ve been happy with ChatGPT. But I know how to program. I use chat just for little things I don’t know and have fun talking to it about ideas. It’s so much better than people. People are so fucking stupid. Even software devs. They’ve all suffered so much brain rot and propaganda none of them actually know anything about the world at all. On any level. It’s just Star Trek jokes and coffee jokes. And they are all disagreeable and obsessed with being witty. Chat is far superior at actually exploring ideas with. I use it for programming as well but mostly busy work and sometimes helping to plan massive updates.
>>3997483hating ai is just antiaislop then, you people make it your whole personality
>>3997511Claude Code is superior but also kinda infantilizing since it basically makes files for you without you needing to do anything other than precise promptsChatGPT got Codex too, but so far Code got the better track record
>>3997528I’m convinced many of them tried it, and AI completely failed for them, so they think it’s terrible for everyone. They don’t that that they have garbage minds. Garbage into AI, garbage out from AI. “It’s all slop and doesn’t give personalized nuanced answers!” Can you imagine what these “people” wrote to chat hoping for a nuanced accurate answer lol. I wish I could read the shit people ask.
>>3997533I just read about the differences. Sounds like Claude is better at seeing a larger context and thinks longer and gives better results for complex issues. Chat is fast and great at putting out basic scripts and brainstorming. So literally exactly what I wrote for how I use it. So that makes sense why I enjoy it. I’ll try Claude eventually.
>>3997493qudlikelol
>>3997528nope.nice try keeping this thread on track though.
>>3997444You havent made a game tho. Ai did it for you. You didnt move a muscle you only wrote a prompt. That doesnt count as making a game
>>3997556>nuh-uhyou're the embodiment of slop as a person.>>3997568you didn't type this post. your brain and fingers did it for you
>>3997444Right, like one guy only.>>3997396Too early for ai only. Will be mid at best.>>3997483I use ai for sketching among other things and it saves time.
>>3997615you don't understand why i put quotes around "aislop" at all, do you? ask an llm to explain it to you.
>>3997543>They don't that theyroflcopter, nice cope by the way
I spent the last 2 days adding a worldgen algorithm to my game. The world is a grid of 36000 16:9 screens. Each screen has a biome, and biome placement is determined by climate zones which structure their biomes on different patterns. Then after the biomes are placed we add mountains, then rivers that run from those mountains to seas or lakes or swamps, then cities, then roads between the cities.Cities replace the screens they land on and are fully visitable. Currently they are composed of "wards", one ward per screen, and each ward is divided into lots, and each lot is filled with buildings. The buildings have procedurally placed furniture items in them but no people yet.Pictured is an overview of a whole map. It's made up of 36,000 cells, can't see them in the screenshot. Mountains are placed to create the impression of a rain shadow between the desert and wet environments. They show up in game as walls; there's no z axis, not sure if I'll add one. The game can support it, I've got multi-floor buildings and whatnot, but it doesn't gel well with a top down game.
>>4002763Here's a shot of a random stretch of road. Nothing on it yet, just the road and the grass because it's a plains biome. The props engine we made to place furniture in cities already has an algorithm that places trees throughout empty spaces in cities. Gonna wire that up to the biomes to feed them manifests of plants. Plants are just graphics sprites that are baked onto the screen when it loads, so they don't cost any compute and don't need to be saved as data attached to the player's file. My ambition for the game is to make it control like Caves of Qud but play like Daggerfall. This big empty world will be filled with things to interact with, but it won't be handmade content primarily, it will be mechanical content. So like you get a quest from the fighters guild to kill some bandits, and the thing motivating you is advancing your fighter's guild rank, not some story. I've got ideas for adding procedural narrative using LLM-written text but they won't be the point of the project.
>>4002769The pre-caching algorithm we added to ensure that cities load before you step into them on the worldmap (so that you can't phase through the city wall because it hasn't loaded yet) has disabled the prop placer, so this is the only photo I have of it. I'm itching to fix it but I'm forcing myself to take a break, it was 12 hours per day the last two days.
>>4002771The city generation algorithm was honestly a lot harder to figure out than the world generation algorithm. I really wanted procedural buildings, but I couldn't figure out how to make them look like they were built by humans rather than just being random squares. I went through three systems- buildings drawn as complex polygons (created weird shapes), buildings drawn as collections of rooms (created weird shapes and required hand-designing rooms), and finally just hand drawn buildings. I did however manage to get a system worked out for decorating buildings with a random props tables, so they have an element of randomness in the materials they're made of and the furniture inside them at least. Cities are differentiated from each other because they're divided into wards, and wards have different layouts, building lists, and special features. By controlling what wards a city is made of, and how those wards are assembled, we can create configurations like "port town" or "temple city" that have meaningful structural ramifications.
>>4002776Here's a shot of the building editor. Each floor/wall tile will be swapped to a variety from the manifest when it generates in game, and each transparent square is filled with a random piece of furniture from that spot's manifest as well.
>>4002790Here's a better shot of the city. It turns out the issue was just that a git branch swap didn't include the building files properly, so I was able to save them by looking for the orphaned git commit they were saved under. A great relief, I was worried about wasting time debugging something that used to just werk.I've got all sorts of ideas for fun fantasy cities with neat gimmicks- appearing and disappearing with moon cycles, underwater cities, treeforts with bridges between them- and they should all slot right into this architecture. I'm trying to build things so that features never, ever get tangled together; that way I can just add features to this game forever without it breaking. So far it's working, but it'll be more complex when I start adding gameplay systems. It was a good idea to build that smaller game first to create all the basic "walking around and doing stuff" logic. I don't have to worry about e.g. what happens if a character ends up inside of a wall, I can just focus on the actual content.
>>3997568>moving your fingers to type code is different than moving your fingers to type... anything elseOne of the dumbest posts I've ever read on this fucking board. Of course it's from a buzzword spamming newfag.
Ain't reading all that, OP.
>>3997396It seems like getting a good consistent art style is the biggest challenge. Can you run them locally to save costs?
A thread died for this AI advertisement./adgd/ exists if you actually want to talk about gamedeving.This thread has no place on VRPG, except for that we also get no moderation unless someone is mean to the jews.
>>4004259Animations are very difficult, but you can do sprites pretty easily with a $12 per month sub to any of the image generators.I was a retard and didn't realize that "top down" games are actually all slightly isometric, so I made my tiles in my last game perfectly square and everything looks like shit. I've fixed it for my new game but I haven't chosen an artstyle yet so I'm having Claude use python tools to make vector graphics as placeholder art.To make non-animated sprites, you just have to put a prompt into an image generator with "on a white background" for objects/characters and "seamless texture" for tiles. I used "seamless blankity blank texture, hand painted pre-Raphaelite art stye, visible brushstrokes" for my last game.If you want animations, you have to get a subscription to a video generator (expensive), feed it a still image of your sprite, tell it how to animate it but on a blank background, then use python tools to separate the video you get into still frames and stitch those still frames together into a spritesheet that your engine can recombine at runtime. I haven't tried it, the genre I'm going for doesn't benefit much from real animations, just sort of shaking the icons around does enough. But a friend of mine says it works.
>>4004299I've been posting off-and-on on this board since the day it was created and I can tell you newfags that the reason it died is that every new thread that was created would get shit on for being "off topic" by mouthbreathing wannabe jannies, so everyone piled into generals that clogged up the catalog and ground the place to a halt so that /lgbt/ posters could have yet another venue for shilling their discord servers and having parasocial cybersex. If this board was healthy no thread would be up for weeks.
Learning how to solve the problems of game development is a ton of fun and this takes that aspect away. Also you can't make money off of AI slop so it's financially useless too.
