/aigdg/ - AI Game Development GeneralSTILL YOUNG EDITION>"Just use AI, nigga" gamejamsNext: https://itch.io/jam/agdg-juan-jam-2Previous: https://itch.io/jam/agdg-juan-jam>Agentsopencodeclaudeantigravitycodex>How to startNewbie? Ask AI.Already a programmer? Just act as a director. Use software engineering models (Waterfall, AGILE, scrum, etc)>How to WebMhttps://obsproject.comhttps://github.com/argorar/WebMConverter (works via wine)>Linkshttps://itch.io/games/tag-aigdg>Friends & Fags>>>/vg/agdg/>>>/g/gedg/>>>/g/vcg/>Previous>>110130722
I'm currently stuck on performance for my game, it includes physics + world destruction. Apart from "Analyze, Improve and benchmark", have you found any specific things to prompt for which have the AI do better?
>>110152660Tell the clanker to move whatever it can to different threads.It'll depend on the game, and depending on how far along you are into the work, it'll be easier/harder and or create issues that you'll have to fix, but:PhysicsRendererAI callsSoundsUsually those four (and then more, just ask your clanker) can already multiply your framerate by 10x.
>>110152660if you haven't yet, help it help you.e.g. you can have it create bespoke tooling, like a lab, where you can test things quickly and determine pain points yourself.if you narrow it down to e.g. cpu-load, you can direct it to rewrite heavy components into a faster language, move things out of the main thread, and e.g. stagger calculation updates so they don't all occur on the same frame(s).
>>110152802based
>>110152639Godot doesn't need any MCP, an MCP is just going to slow you down, and I haven't felt the need to install one for unity because... I haven't had any reason to, either.Methinks MCPs were mostly needed/a good idea on earlier models, which is why everyone's blabbering about them, but I have yet to see a need for it on my end, but for all I know I'm just ignorant and an MCP would somehow help with something, I'm just unaware of what, I asked claude about it and he seemed pretty in line with my current understanding of it.Skills though: Earshot's been pretty good; https://github.com/Leonxlnx/earshotAnd I'm currently trying KILN and LOOM, but astra is slow and astra is shit no matter what so her results might turn out shit, and I'll have claude try them as well and see what he can churn out instead, we'll see.
Yeah see, sometimes it does an 'okay' job, but this did take a few tries and it's got way more tris than it should (at first it modeled the wheels directly rims directly, rather than use normal maps, so that was +4000 triangles just for the wheels lol) , but good enough, faster than I could've done it for sure.
>>110153001was this your kiln test?i linked it in the previous thread after reading about it but haven't used it for anything more intricate than rooms yet.
>>110153015No, the kiln test is still running, I'm not sure if it's slowing her down that much or if she's overthinking the design (clunky industrial low poly power armor, which should be simple but who the fuck knows with them), we'll figure it out someday. Today hopefully.
>>110153001I'm hearing mixed things about Astra from it's dogshit to it's meh to it's the greatest thing yet that can create any mesh.
>>110153099It definitely cannot create anything, anything SIMPLE maybe, yeah, but opus can do that too, and it'll do a better job at it, and cheaper. It's still moving forward, mind you, I'm just being difficult because I expect great things from it (at that price, it better fucking should), but compared to what it could do last year?I'm surprised it can even DO 3D anything in the first place. Maybe 6.2 will be better.
>>110153164Yeah, it's way better than it used to be, but it has a ways to go. I'm going to have to have AI teach me how to get good at modeling/animation, but it can at least handle implementing code ideas very well.
I had asked Claude to take an old 2D enginedeved game I had and remake it in Godot with 3D models it made. Also asked it to me some real boss fights and new enemies. Its doing better than I expected
Why can't you retards use AI for something cool like a new version of the bubonic plague that only affects Indian people
>>110153323that's the goal that caused china to make and release covid; let's not repeat it.
>>110153275You working on VR?>>110153323Israel had a virus in the 90s that would wipe out Arabs. Hopefully they release it if Iran and Palestine fuck their shit up and it also effects other darks.
Okay, this took 1h57m of astra ULTRA.Lol? I mean it's not horrible, but what the fuck are these animations, I didn't even ask it to do any animations, the textures are 'ok' but they're not at all fitting with my Art_direction.md file, it was completely ignored. I'll let claude try kiln+loom too, just in case, but this was not worth it.
Have never tried this but is it really as easy as people say to steal peoples games with AI? I see that story about the needle in haystack guy having his game stolen by AI before it even came out, and im wondering if its legit that simple or it's just 1 case that everybody is echoing.
>>110153413nah, first person game
>>110153557Yeah, it's really that easy to do it
>>110153557It's easy to clone a simple game like needle in the haystack. There's going to be a lot of meme game and random bullshit games like never before. It's difficult to clone a game that actually feels good to play. Simple premise lolsorandom games are easy to clone.
>>110153557If the game is nothing more than a memeable gimmick then ya, its probably very easy to steal. If there is any real substance then not so much, though vibing up something that in a picture or a short twitter video looks like you stole someones game would be pretty quick. Finally steam isnt helping by apparently letting multiple games have the same name
>>110153275I cant tell if the play is to learn to model my own stuff or just wait another year and have AI eclipse anything I could model anyways. Things are moving so fast now
>>110153509Lmao at this point I'm convinced it'd just be easier to use Spawn for all our model gen then steal the assets
I don't think LLMs are AI and I think the whole industry is a scamhoweverif it does end up working and lets me easily make the games that I want to play, exactly how I want them to be, then that would be great
>ai game development>not a single decent ai developed gamevolumes
>>110153557No, it really isn't.You'll see it happening to games like needle in a haystack; because those aren't real games.What the fuck do you mean a game?Is tetris a game too in the same way gta san andreas is?AI is insanely good at handling game logic and writing all the code, and doing very simple models and animating them (not greatly but passably). Needle in a haystack requires nothing that poses a threat to AI; ask it to clone GTA san andreas and you'll quickly see the limits.Well, with enough time and patience, it will, there's also nothing crazy about GTA SA that would make me think claude couldn't do it, but it'll take you a couple months to nail down all the game design in the right way and iterate, and that's considering you'd have access to all the assets in the first place (including sounds, which AI is still shit at). It can clone gimmick notarealgame games, but it can't yet clone actual video games.
>>110153713So not AAA but just indie games?
>>110153621That sounds like a retarded plan, but nothing stops you from doing it, you can also do it and hope to make banks enough to quickly hire artists to fix it before whoever you stole it from sues your ass, I mean vampire survivor basically did that and it worked out for him, so who knows. >>110153736No, not even all indie games, 'some' indie games.It could probably copy amongus pretty well. It can probably copy most indie games pretty well, but most indie games are pretty fucking shit, desu.Can it copy the tight combat of hollow knight or how good the driving/piloting is in brigador? No.Not in one prompt, maybe in 10, or 20, or 40, depending on how good you are at prompting too, but not easily, no. It couldn't copy the reloading mechanics of beyond citadel or of synthetik 2, well, it took me a few tries to get it right, for example, and that's just one part of the game. The more intricate it is, the more prompting you'll have to do, if you're thinking 'just one prompt and it's my game now' then, yeah, no, not happening, not any real games.
>>110153768Spawn.co not a game. It's a three.js vibecoding site. They won't give a fuck if you make a game then take the assets, they're all chill dudes.
>>110153736It's never been a better time to be an idea guy indie dev. The competition will be bigger than before, but if you've got a game idea you want to prototype out, it's never been more possible. I think we'll see some interesting game ideas coming out over the next few years. Think about PUBG. That game could have been made at any time, but it took that guy thinking up the concept to spark an entire genre. Now more creative people who might not have programming chops can apply theory.
>>110153736Depends on your definition of Triple A. Palworld was made by an indie studio on a moderate budget and had better graphics, performance, and gameplay than the slop Pokemon has been putting out. Yet Pokemon is considered Triple A.
Consider the Following:>Palworld was made with a handful of people on a low budget>Only one guy knew Unreal, he applied saying he could teach everyone and taught everyone Unreal as they made the game>It broke absolute records>Expedition 33 was made about 30 guys in Unreal>None of them even knew how to program. They used zero coding>Won game of the year>Tim Sweeney wants to make Unreal as Vibecode friendly as Three.js10 dudes with entry level experience gamedev experience, programming skills, and some AI could probably make a game of the year in Unreal soon. THAT 10 DUDES COULD BE US!
>>110153969Good point anon. Buying epic stocks right now.I for one am glad Tim is embracing AI, even if it is in part to try and distinguish his store from steam. An AI friendly engine is the future, just needs to get to the point it can make decent visuals too. At least PS2 tier I think. I'd love to vibe myself a ratchet and clank like
>>110153969>that 10 dudes could be us*if you have insane artistical talent and put days and weeks and months into meticulous design, imagine thinking expedition 33 was just randomly made up slop
>>110154080all the code was pre-made blueprintsall the animations were asset storea lot of the models were assetsconcepts and posters ai generatedstory made by 1 redditorgo and make game
>>110154132>all the code was pre-made blueprints>all the animations were asset store>a lot of the models were assetsThese still cost money anon. They allegedly spent 10 million on the game, which already makes people skeptical due to how low it seems. Give it another 2-3 years though and who knows. Graphics and voice AI might just get good enough by then
>>110154080I said soon. As in if mesh gen gets solved.
>>110154174>he thinks good and consistent art direction can just be ai generated
>>110154215What makes you think we can't do that?
>>110154298the same reason AI can't write good music, you actually need sovl for it
>>110154148Just pirate it or have AI do it for you. Who will tell the difference?
>>110152603based, is this a new general? this was long overdue, yuge potential
>>110154314>Muh SovlYa'll never learn
>>110153282thats very cool
>>110153969making good games requires skill and talent
>>110154408Bold of you to assume no one who uses AI has talent.
you're getting claude to use deepseek 4.1 flash for FREE right?
>>110154444Explain.
>>110154435if you're just dipping your toes in now because of AI then you don't have talent
>>110154444nigga I got OPUS, why the fuck would I give a shit about that retarded chink shit? can it even make tetris?
>>110154470Bold of you to assume I wasn't already submerged.
>>110154474read the post i replied to
>>110154483Bold of you to assume I'm not the same person.
>>110154492>10 dudes with entry level experience
>>110154468just asked opus how to delegate the monkey code to opencode's free deepseek 4.1 flash. basically it told me to create a skill for claude to delegate coding, edited the claude.md so opus calls deepseek whenever it needs some code to be spit out, then deepseek outputs the code back to claude for orchestrating. supposed to save a shit ton of paid usage>>110154472opus orchestrates while the chink shit writes the monkey code, your usage ain't unlimited boy
>>110154494Entry level experience into coding. Yes. As in a year or so rather than senior devs doing it for 20 years.
>>110154507that means you arent talented
>>110154517>ONLY TALENTLESS PEOPLE USE AIAbsolute /gedg/ seethe. The point was no one in the Expedition 33 team had coding experience, and no one in the Palworld team had Unreal experience, yet E33 devs created a GOTY without code and Palworld devs learned it on the fly. Or in other word, Entry level devs.You seem upset.
>>110154530>The point was no one in the Expedition 33 team had coding experienceThey absolutely did, that team was full of talented people. Palworld devs were not entry level eitherTalented people use AI too, but like I said if you're beginners just climbing onboard now because of AI I've got bad news for you
>>110154552>but like I said if you're beginners just climbing onboard nowSo your entire argument is assumptions. Got it.
>>110154503it's... not?How the fuck do you people run out of tokens, nigga I'm running 3 terminals all fucking day long in xhigh, you guys are foolish
>>110154556You literally said entry level...
>>110154559i got the $20 subscription
>>110154567sorry nonny, I just felt like sub-mogging someone because I've been bashing my head on astra's wall because she's a fucking retard and it pisses me off
>>110154564Correct. Entry level. Someone with even a bit of programming is ahead of the curve over E33 devs who had none. Someone with even a bit of Unreal is ahead of Palworld devs who only had one Unreal user. Entry level guys can learn Unreal. and have AI on top of that.Why do you keep repeating yourself, are you having a stroke?
>>110154559what terminal is that?
>>110154584>E33 devs who had noneI already told you that's not true
>>110153282Oh man, I decided to record the old pre-ai version I made for context and I didn't realise how horrible it looked lmao>>110154380Thanks anon, pretty cool what AI can do these days desu
>>110154597L + Wrong + Black + Gay. Not knowing any code is entry level.
>>110154621The E33 developers did not not know how to code and the Palworld devs have released 3 games before Palworld over 6 yearsYou have dementia
>erhmm, sorry I can't make this picture because you said the word 'cannon' haha violence is bad>was literally imagegenning armored knights burning people alive yesterdayI'm so fucking tired of their shit, modern american culture needs to die>>110154598it's not horrible at all, I like the special effects and the bone/slime/blood coming off, pretty cool bro>>110154592it's just a vibecoded powershell with special effects like CRT or backgrounds or animations and w/e because I needed a way to distinguish between them when I have 6 of them open at once like now, so they all have their wallpapers and colors and effects etc
>>110154632Palworld devs were entry to Unreal. E33 were entry to coding. You're wrong + gay + black + suck dicks for the taste.
if you want to learn how games do something, you can just ask your clankerin about an hour, sol>told me how to install pokemon unite in an emulator>figured out the whole package format>extracted the vfx assets >started reconstructing them in unrealand then ran out of 5h usage because $20 plan
>>110154647Anyone can switch engines, that doesn't make you a beginnerE33 developers were not entry level programmers, that's just completely made up
>>110154669Sorry you lost homo. They used nothing but blue prints.
>>110154677blueprints are Unreal Engine's visual programming tool
>>110154578i tried astra to do my home assistant automations and it was underwhelming, never touching openanus shit againopus found a trillion bugs in the slop it designed
>>110154644You have terminal skins for each of your clankers? That is pretty neat, what was your prompt to make it happen?
>>110154644I'm glad I didn't run into any refusals from Claude yet. It happily made the amount of blood adjustable and added bleeding effects for severed limbs.Video unrelated. Just added basic abilities and made the inventory a bit more readable at a glance.
>>110154678what's next nigga you gonna make a game in scratch lmao
>>110154701is there a brain in your head
>>110154688you have no idea how fucking bad it getsthis is astra ultra, by the way, ULTRA, and it can't even tell left from fucking right, I'm so tired of this shit, I only started with codex 200 because I had banked resets, but if I had known, holy fuck>>110154692It's honestly not that much work, I had to fiddle with a few things because I wanted very specific settings but I basically just told him I wanted to have a GUI/launcher I can start with a right click contextual menu anywhere in windows, and in that interface I can chose which clanker to start a terminal with, using whichever environment I want (in case you have multiple accounts so that you don't have to constantly swap in and out of them), and then have a specific background + effect for them, and voila, that was it. >>110154698It's more than that though, because the bitch has been making tanks for me for forever, but for some reason, sometimes, at completely random, she decides 'okay no more guns for you', it's fucking retarded.But yeah, claude is the way, but claude doesn't do imagegen, and good 3D without imagegen is death.
>>110154698This is really cool btw bro, I'm looking at your stuff, I can't say I'm a big dungeon crawler guy, but I like what I'm seeing
>>110154762That is interesting, so you do use Codex when you need image gen when you are only doing code you are using the terminal? That is neat
>>110154377The soul is in the intention.
>>110154762I'm trying to go with local imagegen, but that has it's very own hurdles. The models might not refuse anything, but they're just not as good or able to have the results edited to a satisfactory degree.>>110154766Thanks. Going for that style saved me some hassle with models and textures not having to look all that good, but the UI is a pain in the ass, because the style doesn't work with small text. I just noticed that the second half of the clip has the HP numbers behing the HP bar, because I increased the stat associated it. Lots of shit like that kek
fml been working on this shit for two days now, it's finally taking shape, and yet there's still so much shit I need to fix; random ship generation with objectives that should stop you from getting access to certain parts of the ship.I'm going to have to get the clankers create thousands of props for the rooms, and I'll probably have to check them by eye too, ugh.https://files.catbox.moe/lkolqg.webm>>110154781Well I use codex in a terminal so that I can have my second codex account running at the same time as the codex app (which is using my main account), yeah. I run claude mainly through the terminals tho, I find it easier because claude's app is... messy. And the terminals are good enough. >>110154939>local imagegenYeah I hope it gets better, it's just too limiting right now, or rather, I don't have the time to look for a hundred different loras to try to get something working only for an update to come the next weeks that breaks something, fuck that. Maybe in the near future it'll be good enough, I'm assuming it will, but we're not there.
>>110154503Give me a basic Gestalt on deepseek. I know it's not as good as Claude obviously. But what if I used something like three.js that's unmatched for vibecoding and has a fuck ton of code for the models to be trained on (and we fucking know they had to have). Do you think deepseek could narrow the gap with claude just doing something like a three.js game?
>>110155284Looking good anon!
>>110155330I'll show you something that looks good *unzips dick*
>>110154444does opencode have a headless version too? sucks that the free tier only works with opencode's shitty harness
Progress: Got the roof building set up so it can make these pyramid style roofs.
why are blatantly off topic threads allowed to thrive here now? at least /gedg/ was actually focused on programming. the world would be a better place if every single one of you vibetards blew your brains out with a fucking shotgun
>>110154762talking like a retard to your chatbot gives you more retarded outputs btw.
>>110152660you will be stuck forever because llms cannot write performant software, only a million useless "smoke tests" for edge cases that never happen
>>110155584shut the fuck up schizo, go suck a dick elsewhere>>110155592I'm fully aware and I mostly do that for venting purpose because it's already being retarded and to assert dominance
I decided a few days ago to make doom build into a single executable from the '97 source code. Since then I've been slowly adding modern features to it like mouse look and uncapped fps. I don't have an overarching creative vision, I just want to have a baseline to which I can add random shit and see what seems fun. Am I shooting myself in the foot by not starting with something like gzdoom?
>>110155634The code is so archaic and so well understood now that it doesn't matter, it'll be a prompt away from whatever you believe gzdoom could have that you don't.
>>110155645>a prompt awayThat's what I thought as well but I've spent half my sessions so far telling it to fix bugs it introduced while adding unrestricted mouse look, for example.
>>110155666What the fuck kind of model are you using that it can't even make doom without adding bugs? Deepseek 3? Astra?
>>110155695Opus 5.5. I am basically asking it to rework how major parts of the game work so I'm not too surprised by the outcome.
>>110155695*starts raping you*Shhhh it'll all be over soon
>>110152660uh oh, peddite meltie
>>110152660What physics engine?
I thought more about what >>110154215 and >>110153969 were arguing about and came up with a solution to multiple problems, and thought about the best kind of game for us to make. The answer is a tabletop like CRPG, like Baldur's Gate 3 or Divinity 2. Here's my reasoning.>Art directionDiscounting Agar.io the two most popular games 4channers ever created were Visual novels. Katawa Shoujo and Snoot game. People loved the creative writing of the games; but at the end of the day there's no actual game play. So clearly, a game that relies heavily on dialogue, choice making, and story telling would follow the strong suits and get other 4channers involved in writing, if ever we were to do a collaborative type game.>CodingPractically a solved problem with AI. You can easily get the base mechanics of a game in a day. Especially character controller stuff. A game without advanced physics would be even easier. Such as a table top type game with set walking/jumping/flying paths the way Baldur's Gate 3 has.>AssetsGen AI is garbage still so if you made a game, one thing you could do to not have tons of different character models is character creation. Have character creation for things like face, species, race, hair, gender, class, name, titles, everything as classes similar to a character sheet. Then the focus could be more on clothing assets that are re-usable rather than designing models for every character, and include that in the character sheets and add a clothing mounting system. This would cover a huge chunk of assets. Most objects otherwise can be found as free assets online anyways.>GameplayPractically a solved issue given how many table tops there are out there.>MusicPlenty of copyright free medieval type music and orchestral music>Voice actingOptional if we use dialogue boxesUnless anyone has a better idea, I think a CRPG would be the best collab project. Plus, there's a hole in the CRPG scene now that Bioware is gone and Larian lost BG rights.
if you are going to use a chatbot to do all the programming for you, there's no excuse to not put that extra energy into artwork. no matter how much you cope, everyone can tell ai art, even if they cant explicitly say it they will know something is off. dont make excuses to be an even lazier fuck than you already are.
>>110156510Only most of the work.
>>110156515dont pretend like you are reading 9001 try catch throw exception CoreSmokeTest.piss, if you are a vibetard you are uninterested in programming.
>>110156441>the two most popular games 4channers ever created were Visual novelsaren't you forgetting someone?
vibing
>>110156510>everyone can tell you used ai artworkDeboonked years ago little bitch>>110154377>>110156542>don't pretend you're No one is pretending goofus. AI is good enough to one shot most stuff for us Three.js Chads to the point we only have to clean up minor errors or do debugging to get what we want. You're the one sitting around debugging all day wasting time little luddie nigger boy while we make stuff and three.js re-tweets me. How does it feel to have accomplished nothing anyone will use in life little nigger boy? Do you feel worthless? You should. You should feel useless like the little nigger boy you are.
>>110156616Oh yeah the Chud King himself. Forgot he used to post his game there. I mostly posted on /b/ back then.
>>110156639>LITTLE BENCHOD BITCH>WE ARE OF BEAUTIFUL THREE JEETSquintessential indian post
>>110156678Show game let's see if you're above web Chads or if you need to lower your little nigger head when talking to the white man.
Been vibe coding this game, which feels like a disservice as I've been actually putting a lot of attention and thought into details.https://www.youtube.com/watch?v=RfMFSyhR7qghttps://www.youtube.com/watch?v=ubou5vCiHLIBut people seem upset that some of the art has been generated with AI. Is it worth hiring an actual artist? Where would I even find somebody to do a better job than this?
>>110156719Who cares what some people think? They're nobodies. Do what's fun. The artist community seethed because they couldn't charge furries 500 dollars for fur affinity smut art anymore and lost their "jobs" to doing that shit to midjourney in 2023 and have been screeching ever since. A bunch of liberal art faggots. They're not the majority any more than trannies were ever the majority, just because they try to scream the loudest like they are by pushing woke shit in video games, ultimately bombing their own studios when no one wants to play it.AI has literally become politicized by unhinged leftists for no reason other than "corporate greed bad" because in reality they were the greedy ones overcharging people for shitty drawings. Three.js showed their the minority by ignoring their bitching and becoming way more popular, so now Tim Sweeney is following suit. Don't listen to the haters they're retarded.All that matters is do YOU like it and are YOU having fun and is it helping YOU get more done.
>>110156778Thanks, this is good advice and I appreciate it
Why has no one posted a game that actually... looks.. funAnd no, a DOOM clone isnt fun
>>110156719Looks fun
>>110156719why doom and not 3d?
>>110152660Luddite propaganda.
>>110157053You do not have access to the same level of agents as Epic Games. They have an unlimited budget relative to you and probably ran thousands of agents. Your 200 dollar plan can't do that.
For my Opus 5.5 sessions that are running agents, reasoning level should I use?
>>110152603I don't know anything about programming or ai, how hard would it be to make a simple 3d zelda-like if I don't care about monetizing it and steal every asset just for the love of the game?
>>110157374pretty easy actually.
>>110157374I'm in the same boat except I know how to model, animate, draw, etc, I just can't for the life of me code anything worth a damn. Is there a starter guide for people like us?
>>110157487There is a 2-words tutorial in the OP for newbies.
So what is the most ai friendly game engine? I guess there's multiple important criteria>can't be heavily UI drivenThat excludes unreal and probably unity?>has to have a lot of code out there the models trained onThat excludes godot So what's left?
>>110157512I'd rather ask someone with experience and who's integrated ai programming into their game engine of choice.
>>110157606three.js
>>110157637>Draw-call bound by default>Scene-graph overhead>Weak culling>Shader compile hitches>Forward lighting>Main-thread bound>Manual GPU memoryYeah no lol
>>110148977I actually have an idea. It'a just that my work ethic has been destroyed, I'm very indecisive, and I freeze up when looking at the difficulties I must endure with making my game.
>>110154644 #>wasted over 5 minutes only to tell you niAnd I assume it wasted your tokens too? I just can't see myself paying for it if this is in fact the case.
>>110157053Crazy how we even have this comparison in the first place. Steam botnet is basically the same as an electron app. A fucking fat web browser instance and a web server that serves you a literal web page that pretends to be a desktop app. The fact that e-pig botnet isn't 10 times faster by default is a disgrace. Like how the fuck did you manage to fuck up your native application so badly.
>have so many ideas for a porn game>It's illegal to develop porn game in my shitty countryI feel so defeated.
>>110158166Tell me them anon. I'll dev it for you and take all the money.
>>110158166Have you considered leaving Australia?
>>110157374>>110157487it's absolutely feasible, but you'll need to work harder because you'll have to prompt for it more (and it'll take longer), lack of experience only means you'll probably do mistakes that'll waste your time, and that's fine.There is zero blockers right now that would stop you from making games as simple as this; just keep in mind that AI is still not great at 3D and animations, so if by zelda like you think OOT, maybe wait for the next model generation. Game design and knowing where to take the game and what to do with it is the hardest part, if you can handle that, that's pretty much all you need, the rest is shit AI can deal with for you or teach you.
>>110153164This is Sol on extra high. It definitely can't 1-shot complex 3D models yet, but I think if you do it step by step, like starting with the chest plate, refine it till you're happy with it, then moving on to arms, legs or head, you'll probably get something decent.
>>110158263It's not really a million dollar idea, but it's definitely a improvement.You know how it's always a limitations on porn games for avatar?Because when you have a character, you can't really do different pose because each different base character sprite require you to have different clothing sprite, which causes the number of assets to explode.But with AI generating the image, you can now input the required information, and just generate the avatar in real time.>Small dick>Whore makeup>Red Bra>Pink Stocking>Red blush on face from shyness>Seemingly nervous>etc etcI just wish there's a game using AI to generate image like that.
>>110156778>They're not the majority any moreMost consumers have a negative view of AI
>>110158166what country? korea?
>>110158438I don't know why I even bother with you idiots, but there's a difference between stated preference and revealed preference. Stated preferences are worthless when it comes to making decisions. People will say what feels politically correct to say. Every gamer will claim to hate lootboxes, gacha, p2w. You'd think mobile games with aggressive monetization practices are a really bad idea, that the audience hates them. Diablo Immortal is a cash cow for Blizzard.AI does not by itself really change that much the number of sales. >inb4 reports claim fewer sales for AI gamesYes, because of all the shovelware slop not selling.Just make a good game and it will sell, AI or not.
>>110158438Most consumers don't care. What they hate is low-effort slop. If the game is genuinely high quality, the only people who'll object to AI being used to develop it are a tiny minority of vocal luddites.
>>110158469>>110158475No, consumers revealed preference is against AI use, it's not a vocal minority complaining about it like lootboxesIf consumers tolerated AI then game studios would be going all in on it
>>110158514lol, lmao
>>110158514You mean the screeching retards that are doing their doxing and smear harrasment campaign and review bombing while froathing at the mouth at steam at any mention of AI?
>>110158528That's a minorityThe majority is either ambivalent or finds it distastefulThere's limited use of AI assets in real games but not much, if consumers were more open to it there would be much moreOf course everyone is doing AI-assisted programming because you don't see the end result, AI art is risky
>>110158392The point of ultra is precisely to not have to do the step by step yourself, but even then, it's overly complex where it shouldn't and not spending enough time where it can, the shoulders area is a good example of this; it cannot understand the very idea of a plate of armor wrapping itself around a limb like it should, it's either making a shitty cylinder like mine or a shitty one like yours, but yeah, I know it can be tard-wrangled, but this is a lot of time and tokens being wasted on doing that, when what twitter shows you is always some miraculous 'oh wow it one shot this incredible and extremely complex piece and I definitely didn't adjust the model myself afterwards!' bullshit.Mind you, shills will shill, it's just how it is.
>>110158550And the one I showed above was already after trying to do a step by step fix, starting with the shoulders and arms, then the head, then the chest, then the legs, then the weapons and then the pose.
>>110158576And that isolation lead to thisIt'll get there, I'm sure in a year's time it'll be capable of doing this, but then it'll probably still not be good at doing a specific thing, like how modern image models still struggle with complex and unusual tasks, it'll be the same here; they just haven't fed them enough 'armored knights' models etc for it to understand how an armor should work, I guess.
>>110158550You're right, but I think the thing worth considering is that these things CAN be done with AI, even if it still requires tard wrangling and a small nations GDP in tokens. It's a genuine options for game devs that have the time and budget to fine tune it.Give it a year or two more, and we'll probably be able to 1-shot whatever model we want.
>>110158598Well see, we have the same goal, and I'm not telling people that AI shouldn't be used for 3D; every single model I use is AI generated (well except for a very small handful of them), but here's the kicker and the point I'm saying:>a small nations GDP in tokensThis is time and this is money.People going into the hobby might think 'oh okay 3D models are solved, the AI can do anything', but it's not true, that's my point; you will either waste time or tokens trying to achieve something simple; or you will have to know how to use blender, which is what I had to do to fix the pauldrons and the heads of my guards, for example.AI couldn't do it, I wasted hours trying to achieve it; it still couldn't. It still did the rest, and I'm pretty happy with it though I had to clean up the meshes here and there in blender; but it was good enough; but not everything.I still had to fix it because it was messy. So people shouldn't come in thinking 'okay I don't have to do that', no, you will have to, or, you will have to spend hours on a simple model that you could've spent 20m fixing in blender, and that will use your tokens (and if you're poor and you got a 20$ sub, good fucking luck). You shouldn't lead people to have expectations that the clanker can't be met, I'm extremely pro clanker, but this isn't a solved problem yet.
Been working on a little incremental gameJust launched the Steam page, demo coming soon.What do you think?https://store.steampowered.com/app/4468160/Oh_No_Meteor/
>>110158741The trailer doesn't really communicate what the game is like very well, and the jump cuts makes it hard to follow.The simple art for the trees and ground clash with the more detailed art for the buildings, and the meteor looks like it's skating on it tail.Gameplay wise it looks fun, but it doesn't look like it does anything particularly unique or interesting. I'm not exactly the target audience for it, but good luck with the launch. Hope it does well.
>>110158863Thanks. There are indeed some minor issues with the graphics I hope to manage to fix them by the time of release.As for the trailer, I looked at how similar games made theirs trailer.
>>110158896I'm no expert on trailer cinematography myself, just looking at it like I would if I was interested in buying it, and I found that I didn't even know what the game was about after watching it. Had to check the description to get a decent idea of it.For all I know though, that kind of trailer might be the way to go.
>>110153694>Game and Engine development>not a single mention of games or engines
>>110158741the fuck is incremental game
>>110158741I think it needs a lot more cooking, not only as Anon said, it's hard to understand what it's about, but everything just looks too basic, this literally looks like a cheap, quickly slopped on together slop.Mind you this is the case of a lot of incremental games; but that's precisely why I only buy a few of them, and this doesn't look like one yet.
>>110156719Looks good to me. If you have an artistic taste and know a thing a two about drawing, it doesn't really matter where you source your art. Unless you want to offload that work to the dedicated art director person.
>>110156719What the fuck how did I miss this?I'm buying it, this is awesome.
I'm going to publish my prototype on itch. Wish me many views and positive feedback, /aigdg/
>>110159378Good luck senpai
made an XML format for Unreal's UMG. I know that you can work on UMG directly via MCPs or PinWright but I wanted to try a proper plaintext language that clankers can write without extra tool calls
>>110153969Like 26 guys made Grand Theft Auto 3. 25 years ago without Unreal. They had to write their own physics, world chunking, and other shit in-house. If you ignore modern graphical asset bloat you haven't needed very many developers to create large games in decades.
>>110159796This isn't exactly true, because rockstar wasn't just making GTA and most of those physics and the 3D engine was already made and working on their other titles then; that's something a lot of people seem to misunderstand because things are different nowadays where your engine is just some third party trash that everyone else uses as well, but back then for a successful studio, you had to have your own engine and think of long term work on it, vertical integration was important; you didn't just make a driving system for one game and be done with it, that was a waste, you instead worked on it for six months for the first game and then developed it further for the next one.You can't and shouldn't compare what dev'ing was like back then to what it's like now, it's a completely different world tbqh, and that's without even adding AI to the equation.
>>110153164I tried this model in Tripo from your front, back, and side views. It got the geometry, but the muddy texture definitely needs some work.LLM modelling is good for basic box modelling but it really can't be used for anything beyond that.
>>110153164It took me a lot of iterations, the one on the left isn't the first version either. But the one on the right is decent enough for my low poly game. Once the clanker gets it right, it can easily reproduce things.
without causing a flame warWhy not just do the modeling yourself?I'm at a point where I might do that since so for claude seems to be shit and I do not want to pay another 100 just for gemini especially since Itrust google about as much as I trust a mouse in Siberia
>>110160105Okay I think this looks great, I've been pushing away my need for tripo but I guess I really should give it a go. Maybe it'll be a matter of feeding it enough different angles to get it to do it right, I wonder if they handle API requests so I could just get claude to ping them when needed. >>110160192For sure, but keep in mind that that low-poly style is going to get old really fast because, well, look at what's happening now; everyone wants to make low poly, because AI can handle it, and this specific style of low poly is not that interesting to look at.It depends on the game, mind you, this is great if you're doing some sort of management/bird eye view type game where your guys just do shit, but would I want this in an FPS game?Pic rel is what opus cooked for me and it took a few tries; and it's good enough... for a prototype, but there's no fucking way I'm keeping it in game as is, I'll either have to go with tripo, or get an actual artist to do it. Or maybe I'll feel inspired and I'll do some blender work myself but I'm pretty shit at textures so, eh.It's a matter of expectations and what you want from it and what you expect will be 'good enough' or even 'good' for the players.Or you're just doing it for yourself and this is good enough for you too, mind you, hobbying is good. >>110160208First of all, you shouldn't be using gemini no matter what, google's models are shit at pretty much everything, argon isn't too bad at modelling but its token costs are fucking insane, so why not stick to opus?Secondly; Why not do everything yourself?Why not write the engine yourself? Why not write a new renderer?I'm not sure what your question really means; do what you want, idk what to say, it's a matter of time, energy, needs, and direction. Like everything else related to game making.
>>110160208>Why not just do the modeling yourself?it requires skill
>>110159958>You can't and shouldn't compare what dev'ing was like back then to what it's like now, it's a completely different world tbqhThe only difference is the scaleGame studios still have in-house engines that they maintain
>>110160279it is indeed for an RTS, so top down.I think AI gen will get better, so I am keeping these as prototypes and when better models come out, I can hopefully just give em these and say "make them better"... if that makes sense.
>>110160208I feel like by the time I learn to model and animate the AI will eclipse my skills anyways
>>110160559Yes, but an indie studio now was a AAA studio then; 25 people working on a game in the early 2000s was huge.It's a whole different beast, and what their engines are now like the rage engine or dice's frostbite are also completely different beasts, not even comparable to what it was then. Then we get into the gamebryo engine discussions, but anyways, I made my point, it's not easy to compare what's happening now to what was happening then, because the context surrounding them and their products was completely different to what it is now. >>110160565No I get you, it's also something I have in mind, mainly because my projects aren't just slopped on together and I plan on working on them for another 2-6 months at the very least (well, probably a year for the witch one, I'll see how popular it gets).
>>110160661>25 people working on a game in the early 2000s was huge.25 people working on a game in the 80s was hugeteams were often over 100 in the 2000s
>>110160105Wow, that's pretty impressive. Didn't think it would work so well, having seen the Mega Man attempts in the last thread.
>>110160713You have a warped idea of reality bro, blizzard was one of the biggest teams then. You're thinking of maybe EA or something, but that's because they were already moving onto becoming publishers rather than a studio. The first call of duty, for example, the studio only had 21 people.I can't even think of a single studio that had 100 people in the 2000s, who?
>>110160746I worked in the game industry in the 2000s
>>110153164>>110160105So dumb question. Why can't Opus learn how to do the shit Tripo does?Do you have a Pro account or ? Tempted to try it for a month, 20 dollars is cheap.
>>110160757okay, now, answer the question; who?
Are there any good AI for sprite generation?
>>110160769Blizzard's core team was over 100 people when they worked on WoW I know that much
>>110160793okay, now, can you fairly tell me that games like warcraft 3, gta 3, call of duty, medal of honor, age of empires, civilization, free lancer.... etc all had teams ranging from 10 to about 30 people, yet you tell me that 'it was normal for a studio to have well over 100 people' when the one example you could potentially give me is wow, one of the biggest and most ambitious games that were ever made back then?I mean it doesn't matter, I don't care, pointless discussion I guess.
>>110156719Would it be a good idea to make the character portraits less similar and more recognizable at first glance?
>>110160815you said "25 was huge"25 was a normal team size around 2000through the 2000s they grew into the hundreds, now they're in the thousandsit's just so you can have a more accurate picture of the amount of work these games took
>>110160844Anon, 25 WAS huge. You think a normal indie team nowadays could afford the money they needed to work with 25 people then, bro what are you saying, the examples I gave you are all ULTRA SUCCESSFUL GAMES/studios, all of them were wildly renowned, I'm not talking about the random agdg shitposter here, come on now. Just because you can think of one outlier doesn't mean all of a sudden everything else goes out the window, you think the hiroshima nuke was small just because you can compare it to the tsar bomb?Get fucking real bro.
>>110160836Do they matter? Is there a campaign mode; do you want to keep them alive? If yes then yes.
>>110160768Tripo and Meshy converts your images to depthmaps, then to NeRFs, then to polygons through a large reconstruction model (LRM). They work very differently from LLMs. LLMs are trained off of large volumes of text, whereas LRMs are trained off of large 3d asset libraries.tripo3d.ai/blog/text-to-mesh-how-ai-generates-3d-models-from-text
>>110156719What did he mean by this?
>>110160866>Anon, 25 WAS huge.It was normal around 2000 and even earlier throughout the 90s for larger projects, it was never "huge"
>>110160983It was huge for 90ssmall for 2000sboth of you are [spoiler]technically correct depending on the game[/spoiler]
>>110160768And yeah I gave them 20 bucks, to answer your other question.
>>110160922>>110161028Thanks. I tried Meshy the other day and this was the result. That was on my first try and hadn't give it much thought.
>>110161159Yeah I don't blame you for not giving it much thought, that gen looks like ass. I've only ever tried Tripo so IDK what Meshy's capabilities are.
>>110160565>>110160661btw what kind of "art style" would you suggest for fantasy medieval RTS... I feel like the common styles are done to death, if it ends up looking like another dota2 it'll just be too boring, no?
>>110161349Just copy warcraft 3.Unironically, unapologetically, the retards at blizzard can't seem to understand what the appeal of warcraft 3 was anymore, so just use that.Well, make sure to keep it legally distinct, but that's exactly what I'd go for if I made a tower defense or rts style bird-eye view game. Alternatively, pixel arts, but yeah, models aren't nearly good enough to be reliable on that front, even for unimportant unit works like this.
>>110161380My “dream rts” as a kid was a crossover of warcraft 3 and warlords battlecry. I love the visuals of battlecry, i will try to copy that into 3d warcraft 3 style.For my prototype the current style is fine
>>110161380nice visuals, but the rotation of the planet looks harsh when changing direction. it might be worth smoothing that out a little.
>>110161380Engine and language?
Day 10 of checking Castellano's progress on his three.js renderer. Nothing new. I'm itching like an addict. I want to play with it.
>>110161778>>110161802Literally just told claude 'here's my godot 4.7 executable, recreate populous the beginning with it' and voilà, feel free to take the idea tbqh, I'd love for a new populous-like game, but I got way too much on my plate already.
>daily shitting on codexHoly fuck, what the fuck is wrong with you you nigger, I'm just asking you to draw the animations of my infantry coming out of their respective military vehicles, but the imagegen goes full retard because there are guns involved.How come we don't have a chink imagemodel that's good yet?What the fuck are they doing? Where's kimi image or GLM image, anything but an american one please?
>>110161380Reminds me of Populous
>>110162366I wonder why it would remind you of populous anon!
>>110162468Based on that title, it should remind me of any game BUT populous
man how do you get decent graphics into your games? I'm manually trying to generate sprite sheets just with gpt and they're mostly awful even if you're manually editing it.
>>110162621You don't.There's not a model right now that can create good, consistent pixel art, or everyone would be using it and lauding it (or seething about it). If you can't get what it gives and fix it yourself; don't go with pixel arts. Or wait for the next models, who knows, but it could take a while, pixel art isn't easy, let alone animated.
>tell the clanker 'I need 20 different drone concepts for my guys to do X and Y and Z'>it comes up with some cool ass designs and now I want to add more of them in the game and figure out new mechanics for themfuck me, I need to stop this shit
>>110162621I saw a spritesheet creator a year ago that was pretty consistent. I'll see if I can find it.
>added big ass mountain for touge
>>110162621You got any suggestions for what you want made? I can try a few different ones for you. I can't find the one where you entered a single sprite and it made a set for you unfortunately.
>>110162621I couldn't find the sprite generator my friend sent me last year. However the creator of this game wrote their own Sprite generator for their game and it's on Github.https://store.steampowered.com/app/5045950/Dungeon_Sookuma/https://github.com/aldegad/sprite-gen/You could probably go through their steam page and see if it's consistent enough for your liking. They have a 28 hour playthrough live recording on the page and a demo you can see.
>>110162730ya thats what it seems like>>110162752those are pretty awesome>>110163304>>110163335I wanted to do like a contra / metal slug small like newgrounds flash game. at first I was hoping for metal slug level stuff. then I gave up and hoped for snes level 16 bit graphics. but now I'm like damn I could just draw 8bit graphics that look better animated than anything llms can do currently. the single shot character concept art and even "fake screenshots" look great though. but ya it seems like you have to be good enough to fix and animate that stuff manually
>>110163424oh thats not bad at all thanks!
>>110162752I'm conciously trying to stop myself from adding more content until the basic systems are all working how I want them. I've got enough content to test them as is.Still snuck chunking and loot containers in there, so I can have pointlessly large maps with a slow dopamine dripfeed. I gotta get started on the metagame and progression, though.Level was so large that the wandering hunter enemy was yapping at me nonstop without ever getting close enough to properly chase me kek
>>110163629Ugh, I really need/want to get to work on the combat for the FPS one but fuck me there's still so much shit to do with the ship gen and the infestation systems, and then I'll have to do a few rounds of framerate fixes too. And then as soon as I get started on something I start thinking of something else I need to take note of for one of the other two games and I keep switching between them to the point where there's always 2-3 claude sessions going for each three games and then there's 2-3 godot sessions starting every 10 seconds and I can't even fucking play the games and test them correctly because I feel like I need to keep working.And now I want to start a fourth fucking project as well but there's no way I'd have time to squeeze that in zzzzzzz
>>110163818Can't imagine working on multiple games at once. One game keeps me fairly occupied as is.How did you get such a relatively neat 3D model? Tripo or was it actually made by Astra?
>>110163884ADHD has its perks and its issues I guess, there's just no way I could keep busy on a single game tbqh. Maybe if it was huge and had so many systems that I could work on 4-5 of them all at once without risking fucking up the work of the other agents? idk. And no, that was astra ultra, took it an hour and a half tho, tripo probably would've done better...Fine, I'll try tripo.
>>110163818Looks neat fren.
>>110163884>Can't imagine working on multiple games at once.Not him, but I'm always looking for something to do while the AI works away updating my project. Why not do a second project in parallel?
>>110163905I'd probably be completely stressed out, if I had to deal with multiple games. I'm even sometimes straining to use up my weekly limit of Claude Pro.Really impressed with how well Astra did on the model, especially with just those references. Tried a little more complex model with Claude once and it did abysmally.
>>110163997Well I've worked in production and I've been handling multiple projects at once for years so, I guess I'm just used to it by now.>Really impressed with how well AstraWait, the girl isn't part of it, if that's what you're wondering, the girl was already made and isn't made by astra kekAstra only did the robot, and yeah, it's not horrible at doing robots/mechanical units. I'm still not happy with it at all, most proportions are wrong, the handlebars are completely fucked, the EMP emitters (those copper square things) are neither in the right place, nor even in the right layout... Well it's fine I'm seeing if she and fable can fix it, else I'll just go tripo because, damn.Tripo is pretty good it seems lol>>110163917Thanks fren, happy natalia to you too
>>110164042>I've been handling multiple projects at once for yearsExactly what I have to do in my job as well, which is why that isn't really something I wanna do in my spare time. Just the robot was ten times better than what Claude made me back then.
Learning how to make music for my game is hard :/
I'm looking at this three.js cloth library. I'm writing a list of tech I want to use for my game and feed through the AI. I'm looking at customizable humans, then a clothing system to mount to human models. Does it look fine for use?https://github.com/bandinopla/three-simpleclothhttps://bandinopla.github.io/three-simplecloth/?demo=skirt
>>110164300>Blizzard removes jiggle physics from WoW>WoW Beta closes>Three.js Vibe Chads step up immediatelyTruly, the greatest timeline.https://www.spawn.co/@izkimar/escape-from-stranglethorn/play
Huh, maybe kiln has its uses? I'll look into the results, see what really comes out, but that's pretty good if true (7800 was astra's work, this is claude using kiln looking into how to improve it and reduce the triangle counts)>>110164553I'm very much looking forward to a vibecoded MMO, but I don't expect the first one to be good.Nor the second one, because it'll probably be gooks making it and it'll have trash gameplay and a cashshop.The fourth one tho, that one will probably be made by omega autists that got pissed at blizzard for adding another elf race and the wow token to forever.
>>110164632It seems more like he's just making a PVP combat simulator for WoW than a full MMO
>>110164114Yeah I still have zero clue how to do this either, I'm really just hoping AI figures it out by the time I need it, else I'll have to find some tranny musician that'll have a 50% chance of ghosting me after providing a single (but great) track. Such is life.
>>110164632https://dreadmyst.com/gummy made this like a year ago lolheres another one but abandoned alreadyhttps://store.steampowered.com/app/2506440/Trolddom/
>>110164683>>110164632Honestly I feel like no one has time for MMORPGs anymore and multiplayer CRPGs with a few friends is the way to go.
>>110163629looking really good anon
>>110164300idk but i spent a while making my own custom cloth sim for my rock climbing game. You can get pretty far by just implementing the most modern (non neural) realtime cloth sim papers and iterating.
>>110165072How would it be for a bunch of NPCs at once?
>>110165193idk, but im trying to anonymize it so i can upload it here for you. Give me a bit, here..
https://files.catbox.moe/rd75rc.7zShitty hentai game with extreme claudejank. In which you play as a landlord tentacle monster who wants to get back at his NEET ghost tennant, getting stronger by capturing the ghost hunter girls storming the mansion every night.The sfx, moans, and general voice effects are bad and I'd like some tips on fixing it (also general bug reports if people care, though I think the method by which the game works is fundamentally kinda buggy). Even still I think it's feature complete, and am *pretty* sure it's playable end to end. God the SFX sucks though, I wanted to give girls sets of voices they could get, but I think Claude has some issue parsing that and I couldn't find any proper sets of individual sound effect moans it could use as soundbytes, just like huge audio tracks. If anyone knows where I can nab (or generate?) soundbytes to fill it out that'd be great, or just mod it yourself, software security is dead, go wild.
>>110165242>>110164300Here's the source code for my cloth solver:https://files.catbox.moe/vj24rq.zipand here's the built one that you can quickly run in a browser to see it:https://files.catbox.moe/rwveyl.zip
>>110155645>>110155666hahahaha SWE chads stay winning.i am making a game and i use AI from time to time, but most of the code and the architecture is my doing, because I'm not a retarded non technical.
>>110156441oh nonononono....HAHAHAHAHAHAHAHAlike, i know AI kiddies are retarded and all, but how will you even vibe code it as "4chan"???you will have merge conflicts from every retard running agents that change half the codebase every day??HAHAHAHAHAHAHAHAhow would you even split any potential revenue?you guys are so room temp iq that it is painful.
>>110165571>>110165193Catbox truncated my files for some reason, so those links don't work, my bad.Here is the actual source:https://files.catbox.moe/k4l1h0.zipAnd the built version for running in the browser:https://files.catbox.moe/z116aw.zip
>>110156695>>110156639lol legit jeet posting hahahahayou think anyone will play your little game that is sloppier than your mom? hahahahahayou realize people can immediately notice when something is half assed, which by judging by your comments, you half ass everything in your life you talentless braindead lazy jeet.how do pea brains like you think there is any point to vibe coding a game anyway, if it takes you like a day, everyone will just make their own slop and no one will touch your diseased AIDS slop.literally race to the bottom jeet tier retard. you make a good case for postpartum abortions.
uh oh, luddie meltie
snailcats are so mindbroken that it's hard to look at. someone needs to tell them that everything will be OK (it won't be but someone please tell them)
>>110158741looks shit. you arent getting your 100 goyerinos back. good.
>>110165710it's going to be okay, anon.it's going to be okay.
>>110156510>there's no excuse to not put that extra energy into artworkv5 precise DOKO
>>110162104>no refunds ranjeet
>>110165663uh oh, brainlet who cant do anything if his kike overlords cut his LLM off...>claude, how do i take a shower without chugging draino
you'll run out of stamina before I run out of tokens, luddie-kun
>>110165692a bunch of polygons made to look like houses make a good game? good luck lol.ai kiddies get so enamored by stuff that isnt that impressive because they are so low skilled they could never imagine a Sisyphus-esque struggle to hand build it in a few days. no, it must take exactly 10 mins for their fried zoomie ADHD brains.complex systems will not survive in 50 years, gg. you zoomies better get fit, you will need it.
>>110162104seedance 2.5 is pretty good at doing blank-backdrop animations if you give it reference images from your image generator of choice. It's a couple bucks per video, though.>>110165815just breathe, anon. just take it one step at a time and breathe.
>>110165819Well the entire point of getting images out of it is to then use said images to translate into models, I guess a video might work too but that also might just make it more confusing for the model.Or not, I don't know, I guess I could look into that.
>>110165806lol ill keep an eye out when the kikes at anthropic and openGAI uber your prices, making all your subs 100x more expensive while rate limiting you even more.remember, no refunds.brain fried adhd neet zoomies wont be able to afford anymore :(
>>110165819i can breef without my ai gfno samuel altmanweitz, i will let you rail me just pls gib ai plsbut seriously, you are right, you guys should keep publishing more games to steam, you guys will escape the le permanent underclass.
>>110165793>how do i take a shower without chugging drainoI've been doing it wrong all this time??
>>110165633Based giving it a look. Sorry for the late response was playing Divinity 2 with friends.>BrowserAnother browser Chad I see. I'm looking through and I don't think I'll be able to use the Javascript personally, seems to be using webGL code and I'm using webGPU. But I can reference it at least. What'd you write it in initially?
>>110165830I think you are a spazz but I do think you are right on that regard. Right now Im using AI because its cheap and fun to mess around with, but I can fully see the future where I am priced out. Just gotta tokenmax until then.
>>110165633Checking the build out now. I think I'll have to try it on my Linux boot tomorrow. I'm still on Windows from gaming, my vite and all that is on Linux.
>>110165633I'm going to bed now, thanks. I noticed you're using Ubisoft's dataset. If you want more data, Mr. Doob tweeted this out recently to someone. It's Bandai Namco's research dataset.https://github.com/BandaiNamcoResearchInc/Bandai-Namco-Research-Motiondataset/tree/master/dataset/Bandai-Namco-Research-Motiondataset-1/data
>>110165829Sorry, I assumed you were talking about drawing animations for spritesheets or something. There was a similar conversation elsewhere itt>>110165848The drawing is webGL, but the cloth physics are actually webGPU. The cloth solver also doens't depend on three.js- you should be able to use it with whatever webGPU renderer you have.It's typescript but 100% vibed. I don't look at the code anymore.>>110165887Oh shit yeah don't put the LAFAN1 animations into your project if you're gonna publish it- those are copyrighted! My bad, I forgot about them.And thanks for that animation pack- I'll definitely scrounge around in there and see if i can use them for my motion matching engine. Sick
>>110165907Neat. I'll see if I can get it working with Meshlets tomorrow if I have the time.https://github.com/mrdoob/three.js/pull/33605
>>110165907impressive polygon cities, but useless for any game design that doesnt involve you playing a giant that just the whole city. make it it might actually sell well.
>>110166162>>110165907whoops*giant that destroys the whole city.
>>110165907Sim City Gorrillion?
>>110165862>on that regardthanks for playing sanjeet.you should make a jamboy simulator
Luddie is upset because his thread is so dead from no one actually making things there that they had to save it from page 10 several times since yesterday. It's so dead we don't even bother trolling it to give it pity activity anymore. It's dead because only skill-less neet hobbyist coders post there, talented coders have jobs to afford AI and actually achieve stuff.Notice how much we've all shared and helped eachother out this thread? People going out of their way zipping their projects for eachother, testing out different AI programs and quality testing and and answering questions? There's none of that in gedg. Just seethe and autism. Because the luddies are all hobbyists with no real skills.
wtf is a luddie
>>110166249And raging /pol/tard reveales his fat retarded face
>>110166239HAHAHAHA HOLY COPEtell me real production systems you worked on and technologies involved in the afromentioneddid you by any chance work for eMudra, you madarchod?i rejoice at your seething, you good goy.how will the neo-goy fight the kike now that the neo-goy is lobotomized beyond belief, barely functioning without some linear algebra telling him right from wrong.
>>110166269Poltards are beneath even indians low castes, all you've achieved in your life is participation trophies.
>>110166269>afromentionedKnew this guy was a nigger lol
>>110166263normally im not racist it just describes your behaviour so accurately
>>110165741SHIVER ME FUCKING TIMBERS!*starts raping her*
>>110166283Sure sure, go make the roman salute to the US flag. That should solve things.
Try not to feed the retard too much attention, that's why he's left agdg, because he's not getting any over there.
>>110166269Notice the ESL words wordsalad but zero things he created posted.
>>110166292He's from gedg. Three.js anon trolled him too hard.
>>110166302He's talking about you, the three.js anon
>>110166292Usually generals get their resident schizo around the tenth or so thread. It's kind of impressive that this one latched on so quickly.
>>110166316the schizo is the reason this thread exists
>>110166332Uncut dicks are so gross
Good news guys! My rape victim killed herself! No consequences for me!
Any feedback on this? Maybe something is obviously AI, or you just hate how it looks, bad colors?
i am vibe coding No Man's Sky like countless other people right now
>>110166416I can't say much because it's not my style, it's too vibrant for me, but I know it's something that existed and so it might be a niche you want to hit, I don't knowFor example see how the color of the bricks of that house on the left literally merges with the road because the color is the same?I think it's generally fine tho>>110166419Gotta add a game to it tho, graphics is the easy part
>>110166299>>110166302>>110166316>>110166321/g/ is gay and i cant id posters unlike pol.but in any case, if you dont have a technical background you wont make any money from your shit slop.I know.... UGH I know ..... I'm sorry!!!!!!!!!!!!!!!!!!!!It's just that you're not gonna make any money is allHAHAHAHAHAHAHAAAHAHHAHAHAHAHA
it's crazy how in 2026 you can make a game like Starship EVO that essentially has very low quality graphics and get a massively positive response. most things in Starship EVO are made out of primitive shapes, cubes, spheres, pyramids etc. it's just hidden well
>>110166416its like morrowind so it has a certain charm but its way too abstract.the obvious slop is the menus, you guys really need to figure out that the first layer of a product is UI UX.
>>110166477>but Starship EVO was made in 2020yes but its only blowing up right now for some reason
>>110166477wait until you find out that literally all game graphics are made of triangles
>>110166478I mean the menu is basically totally placeholder at the moment. Too abstract how? Or do you mean it's too saturated, too limited in color banding
>>110166499i mean there are silhouettes of npcs with zero detail.you could lean into some early might and magic rpg stuff with the aesthetic.
>>110166416the lighting could be worked on- the shadowed areas look very flat since the lighting is binary- just "in shade" or "in sun". Maybe you should get some global illumination implemented.The scene could use some dithering to break up those blobby shapes in the ground/sky, but that would probably have to be done in screenspace to avoid excessive moire patterns like you can see in the rooftops.This is a pretty cool project involving dithering in complex 3d scenes with a limited pallet like yours:https://www.youtube.com/watch?v=EzjWBmhO_1EAlso, that grass is the exact type of grass that AI tends to make by default. You could give it another prompt or two to make it more unique/varied.
>>110166416Also, you could look into using normal maps to make the lighting pop more on those flat walls:https://www.youtube.com/watch?v=oJ0vrDTA0sA
i dont think most things in Starship EVO are even textured. i think its just solid vertex colors or whatever you call it
>>110166541yeah looks like there might be a noise filter on them too (unless its just image compression here) but looks to be mostly just solid color and letting shaders do the work.
>>110166533>>110166513Thanks for the tips, anon. I at first wanted to go for the style on this game https://store.steampowered.com/app/4144090/Shores_of_Vaelithar/ but AI couldn't do it and I don't think I have the artistic skill for that at the moment, and I think it requires different movement and scale than what my game is, so it wouldn't really work anyways. Then someone linked webm related, so I pivoted more towards that.
>>110166541Whenever I see one of those big featureless grassy lawns that are a signature of sandbox/survival games, I instantly lose any sense of interest or wonder.Mars First Logistics is another one of those very simple physics sandbox games with flat shading and basic mechanics, but I had quite a bit of fun killing some time on it. A project like that could definitely be vibecoded and sold on Steam successfully, given some love and care put towards the important bits.>>110166556Yeah covering the ground with tons of grass and bits of this and that is a pretty good way to guarantee that your game will be nice to look at. Seems like that webm uses sprites for most of them. I'd like to see how it turns out when you've worked on the environment a bit more if you don't mind posting again- i've always loved the charming look of 3d pixel art.If you want to churn out a ton of simple pixel art stuff and you're using Claude, you can ask it to break out into haiku sub-agents and model a ton of individual small assets in parallel. Doing that uses basically no tokens compared to Opus and you can get some decent results for scattering around an environment, picrelated.
>car simulator>walking simulator>town simulator>base simulator>planet simulator>cloth simulator>limb simulator
>>110166595you are in a brain simulator thread. what did you expect?
>>110166590>decent results>picrelatedAHAHAHAHA
>>110166609that horse saw some shit back in the day, refuses to eat since.
>>110158030You could never
>>110166590man obviously a few of those (the horse, the bird) suck, but otherwise those are pretty good depending on the level of fidelity you are going for. Probably took claude 20 minutes max and would take me weeks to do all that with the UV and texturing
ai driven development just accelerates the time that it takes for me to get to the same point i always get to where i question my idea. it sucks, i feel so driven and motivated and then like clockwork dread creeps in and kills it for me
>Claude weekly limit reachedI'm gonna miss that autistic clanker. It's honestly surprising how much it got done in just a week.
>>110166861is this just ottd in 3d or whats the twist
>>110162621GPT can be decent for 1 use sprites (space station, asteroid, etc) but trying to generate any kind of coherent sheet with it is an exercise in futility.I've been using it to make assets for my rpgmaker game (and make custom scripts because FUCK learning javascript) and it flat out cant do certain sprites with any consistency no matter how I prompt it.
>>110166890Currently, pretty much. Intended to be something like a midpoint between factorio and TTD eventually
>>110166477Because people don't want abstract "good" graphics that end up being a blurry mess with ghosting artifacts and shitty textures, which makes everything blend together, so you can't see shit. People want READABLE graphics. For the same reason we used to disable textures in games like quake 3, to make something like simplified reality. Most of modern devs don't understand that they should strive to make graphics visually readable instead of slapping hd 8k phototextures and calling it a day. If you manage to do that while maintaining a consistent visual style, you are golden. See superhot as an edge case example.
>>110166419Is that yours? Looks pretty cool. And I assume you mean Space Engine -like game, because NMS is literally just a space flavored minecraft with no proper planets and a fake space travel with a hidden loading screen.
>>110166902neat, good luck>>110166932>NMS is literally just a space flavored minecraftthrowing some pretty harsh shade on minecraft there
why do you persist? how many dead projects do you have? what makes you keep coming back? you say its not about the money, but you know it is, theres no such thing as a passion product. passion needs fuel and you dont have infinite time. where does the drive to continue come from?
>>110166990Making things is fun. If it's no longer fun, I make a new thing.
>>110167052This for me but i also blame ADHD. Once i get close to 90% of finishing a project I start losing interest, even though that last 10% is crucial.
>>110166990Serious answer if you're struggling though, if you want to finish your projects, you don't need drive but a daily rhythm. Drive and enthusiasm is fleeting and unpredictable, so it should never be relied upon for more than just getting through the "blank canvas" phase of a project. Getting to the finish line requires you to make working on the project an unconscious habit, rather than an active choice.
Is there an IRC or Shitcord server where us folks can hang and shoot the shit outside of the threads?
>>110167121yeah here >>>/vg/585597102
>>110167109I'm sort of driven by wanting use up my weekly limit. Maybe not the best approach, but it means that I'm seeing steady progress without burning myself out, since it's just a $20 plan.
>>110167189If it works for you, no need to change it. Just be careful using up you limit doesn't turn into a burden instead of a drive.
>>110166556eugh, we've discovered the 3d equivalent of mixels
>>110166920>For the same reason we used to disable textures in games like quake 3the reason you did it was that people did it to get 125 fps on their garbage cards, and then you cargo culted their cfg like everyone else. that and vq3 character models didn't contrast too well. q3 looks squeaky clean even without picmip
>>110165907how many days to do this and how fast is the generator itself?
>>110153969you'll just get raped like tetrachroma dev lmaonow that nobody has moat the only moat is marketing and having a publisher behind you
>>110153969If 10 autists here could band together and decide on a direction for a game, it would surprise me.
>>110167520all the bocchi xisters could unite
>>110167395It took about 120 prompts over the course of two months. Many of those prompts were just telling it to resolve issues that I had marked up, and the whole thing got rebuilt once at the early stages, so I probably could have been more efficient with the number of prompts to get there.A town of 80,000 people takes about 26 seconds, and a city of 1.25 million people takes about 120 seconds. Most of that is spent on iteratively re-generating the roads so they don't have layout issues.
>>110166659Yeah I just did it as a test of Haiku. It definitely isn't good enough for the hero props that dominate a scene, but for generating a ton of little things like grass, rocks, flowers, bottles, etc. it's pretty great. You can populate a whole world with little pieces of unique detail pretty much for free, now.
>>110165380I'll give you a (you), I only played it for a few minutes because, yeah, it's janky. I don't hate the idea, it's pretty super-deepthroat adjacent, which is good, I like the gimmick of the father/daughter thing too, but this is too much, there's too much text bullshit from the get go, it's a porn game bro, chances are people are already horny because they start it, so you need to allow them to get into the action faster, then you can figure out a way to add the lore and whatever, but you can't just drown the player in lore from the get go, shit's boringThe manor management and trapping is pretty basic and really just an excuse to get to the good stuff, you need to figure out a way to make that part interesting too (maybe you get more points if you get to trap a girl as she's showering, or maybe you can get more by getting multiple at once and/or when they see each other get captured, maybe some sort of mindbreak bullshit), and then the tentacle part is pretty cool, it's more intuitive than I first thought, but it's still very janky.It's a fun idea, but it needs a lot of work.
>>110166841But that's a good thing?Now you know quickly that you're wasting your time, how is that a problem?
>>110167409The moat is doing something actually new. Now that iteration and basic visual and technical quality are free, it's the ability to hone the gameplay and make a unique experience which fulfills a vision that will make indies stand out.
>>110168391no because if i have a million dollars and you don't, i can spend 1% of that to slop exactly what you did and the other 99% on outmarketing you. AI only took bargaining chips from you. if you're noncommercial then whatever but you'll still get pissed if another notch makes billions off your infinimineryou need to understand that people are attached to the brand and personality, not the actual content. you can take a shit in a box and put it on a gamestop shelf as long as you have marketing. people paid for monkey jpegs because someone told them to, why didn't you make a monkey jpeg generator back then? think real hard about this question and you'll understand the situation you're in.
>>110168581what the fuck are you talking about, this would happen no matter what; yes, money is power, welcome to the adult world you retardyou're only insecure about AI because you know you're a low quality, uncreative fuck and you're scared that the clanker takes the only advantage away you've had until now; speaking computerget bent
>>>110168881>>>110168881>>>110168881>>>110168881
>>110166590The objects look fine, the living things look like shit.
>>110168886Thanks for the new thread didn't realize we were page 10.
