can we get an OP that doesn't use these retarded edits

Last thread no one answered me.
So again: How good is Jetbrains Junie compared to other options, especially with Rider and Unity.
I already have the subscription and it comes with free credits so I might as well use it.
Also I would like to use it with modding and hence have it use the .net debugger by attaching to the thread.

File: snailcow.jpg (198 KB, 640x640)

198 KB JPG

>itch.BO

>>110263682
redpill me on bevy vs godot

File: file.png (54 KB, 220x222)

54 KB PNG

>used up both of my pro plans
>nothing to do all weekend

I used to enjoy being free of obligations now I just want to vibe more!!

>vibetranny's paypiggy account ran out of cucktokens and desperately needs to seek attention on 4chan with a trananime picture

>>110263682
>India Games Showcase

oh are we attempting to turn this daggy general into shitpost cesspit already?

>>110261536
Thanks. I will try it out.
Have you ever tested doing the same, but with GPT 6.1? Might not be as good, but if we are giving it 20 passes, maybe it will do something decent.

File: file.png (35 KB, 563x140)

35 KB PNG

Dude what? The original coder/cofounder of 4chan just liked my random ass tweet calling out a Luddite Mexican girl for her art profile pic. He's a DJ now, saw him playing in VR Chat at a rave another 4chan user invited me to. So random that he liked my tweet with like 30 likes. If I wasn't an ancientfag that remembered the "Bring back Snacks" campaign the name would have slipped right by me.

>>110264566
>saw him playing in VR Chat at a rave another 4chan user invited me to.
W.T Snacks must be like 50 now
thats cringe as fuck

I got too cocky
I wanted an inventory system and removable clothing on a 2D sprite and Astra just got lost, it can't do the final 5% of the work
This is a new project and I didn't even set up a repo argh so many tokens lost

File: Untitled.png (303 KB, 901x495)

303 KB PNG

Still trying to port TES II (original from DOS) to PSP with Argon (plus checks up code and summarizing of chats by Deepseek and GLM 5.3)

>>110264610
I mean the music wasn't bad, I was drinking and playing so I don't remember much of it. I'm going through his stuff out of curiosity and he's not bad, but not really my thing. I prefer classic Eurobeat where there's goofy Engrish like "DO YOU LIKE, MY CAR!". But I'd go to a club and listen to it if someone invited me.

https://x.com/MidnightSnacks/status/2082652596792099175

How do I work more efficiently?
I'm currently doing "prompt it do 1 thing, then when it's done, I check the result, then continue"
And I can't even fully use the Claude's Usage at the end of the week.
I can't help but feel like I'm not using AI to develop games correctly.
I even asks it to use Opus already.

File: 0311426004282208.png (238 KB, 428x426)

238 KB PNG

okay, is this better? fuck you guys yes it's better, it's okay, i can't take any more slopping on my selfinsert.
>>110265238
set it on xhigh. split the repo in parts. the hardest thing with opus i'd say is your own brain bandwidth to understand what you want him to build. if you want to make some unoriginal slop literally tell opus to decide and do everything for you.

Does Clover Studios still exist? The anons who made those light novels?

>>110265238
>"prompt it do 1 thing, then when it's done, I check the result, then continue"
This is unironically peak vibeloop. You have to test each thing carefully (because even Opu55y does weird unintended things) and make sure to perfect what it did before moving on. Vibeslopping everything autonomously just allows things to grow out of control.

File: mira_angry_talk_blink.gif (152 KB, 600x560)

152 KB GIF

>>110265238
I think that's mostly the correct approach, though. Everything else will most likely lead to ungodly slop, as Opus is a terrible game designer when left to its own devices. You could tell it to spin something up to scrutinize its work before having you review it, so you don't waste time on obvious garbage.

>>110265257
Just accept that your character sprites will have a slightly sloppy look to them. Most people won't even notice or care, if the rest is good.

>>110265395
What are you sloppin' in? I just use chatgpt.

>>110265425
ANIMA and QWEN2.1 for the characters. Krea2 and MOSS for enemies, backgrounds and textures. Trying to do that stuff locally.

I was wondering if you could do three.js games multiplayer completely without a server, as in static pages living on github pages or something then using webRTC tokens and joining through a friend code. Turns out you can and there's several libraries for this. This one is 3 years old but shows it works.
https://rameshvarun.github.io/netplayjs/

I probably wouldn't use it though. There's another one being maintained more recently from poki.com

https://developers.poki.com/guide/netlib


I'm thinking about just writing a vibecode starter pack of networking tools and versioning so people can make multiplayer games for the web that live on Github pages.

>>110265645
Seems someone already wrote the tools for seamless tab switching and not noticing players drops out for their game with this library.

https://poki.com/blog/developer-spotlight-no-pressure-studios

>>110263999
Time to spent all weekend writing up a detailed design doc for what to vibe next.
Or get a local clanker set up so you can vibe small stuff even when the good models are out of usage

Has anyone figured out how to make sprite sheets without it looking like shit and drifting away too much?

I had to resort to making 3d model from a prebuilt rig so that animations look consistent. But Astra sucks at 3d modeling and models look nothing like concept art it make in 2d

File: d142c1ace5de9a03ae3a8328a(...).jpg (267 KB, 1350x1615)

267 KB JPG

>>110263999
just buy max. max doesn't have the 5-hour bullshit. are extra 100$ a month actually worth constantly logging in?

>>110264622
I saw the screenshot first and immediately thought "That’s Daggerfall"

>>110265238
That's how I ended up going for multiple projects at once; but it is exhausting on the creative front.
But also; a lot of work can be done in parallel, but it depends on your game and depends on your setup, but you just create a separate branch and you try to figure out how not to impeded on the others' work.
I'm focused on a class based action tower defense game right now; and I got a billion different things I can do while something is being worked on; while one clanker focuses on rendering, how the shadows should look, the sky/skybox, I can have another one focus on improving the animations of the player character, then another one can focus on the mobs' AI, then another one can focus on creating more props for the map, yadda yadda.

It used to be a bad idea to do this because models weren't smart enough to just... figure out how to merge everything and you had to do it yourself, but now honestly opus just understands and I don't even have to explain it to him. He'll cherry pick the individual commits, and he'll ask in case he's not sure I want to keep A or B.
It just works bruh.

File: 1763488718125189.webm (309 KB, 480x854)

309 KB WEBM

It's over. I need a job. Even the most simple AI workflow is too expensive for a broke neet.

I've been working on my game for 2 and 1/2 years almost entirely with vibe coding.
I have developed games in the past without AI, but no commercial releases, and only one or two finished ones.
Ask me anything.

AI exposed me. I can't actually make a game. I've tried game dev again after years of pause but now with AI and made more progress in one weekend than several months before AI. My ideas just were not fun to play.
The fun part was having an idea that was in my head and then work on it.
At least AI gets right of the 'idea parasite' fast. I could be walking around with a 'million dollar' idea for months and now I can just prototype in few days and get over it.

File: 1791146234486432.jpg (51 KB, 978x890)

51 KB JPG

>>110266119
Wasn't AI awful 2.5 years ago? Do you keep getting better models to rewrite it or how are you handling the huge quality shift over time?

>>110266129
Your not going to get good it idea guying without implementing ideas, AI lets you blitz through the prototype phase, its great. Got to kill off some overscoped ideas that have been bouncing around in my head forever since I didnt have the time to actually make them till now lol

>>110265359
>>110265395
But it feels like a waste for not using it to the limit.

>>110266037
This could be a good idea.
I should just run 2 projects at the same time.

>>110266129
This is going to become a normal thing that a lot of people will have to go through, and that's precisely why I don't see the AI sloppa wave to be a problem; most people will take a jab at it, fail, and realize it wasn't as easy as they thought it would be and will give up on it and do something else.
Balance in all things, all going according to plan.

>>110266199
>I don't see the AI sloppa wave to be a problem
Except there's still the problem where parasites use AI to quickly copy a game. (Or straight up stealing it)

File: luola.png (276 KB, 960x540)

276 KB PNG

>>110266225
>AI is le stealing games
>*looks it up*
>tumblr dev is making overpriced generic blobber clone flash game #54321
>AI technicians open sourced the original game in Rust, no not your game, the original game from the 90's that you were stealing from
You lost.

>>110266172
I do some prompts on Extra or High, mostly when planning or trying to figure something out (and return to Medium for the implementation and testing), that increases usage decently.

>>110266199
>most people will take a jab at it, fail, and realize it wasn't as easy
You overestimate a lot of peoples ability for self-criticism. Youtube is already flooded with AI generated garbage. AI is great but its singular problem is it lowers the skill required to create something below the intelligence threshold at which you can tell if your own shit stinks or not. I am convince the AI hate would be a lot lower if access the tech had been gated behind needing at least 100IQ or some other metric correlated to intellect (wealth, geography, class, etc)

>>110266143
It was. I would copy and paste stuff, and since I was obstinate about using gdscript, much of my code came back as python (models did not understand it yet). I did learn a lot about gdscript from a few years ago when I tried making a 2d game, so that helped.

>>110266257
What do those low, medium, high settings do?

>>110266225
I don't see it.
All the games I've seen quickly copied are all simple shitty gimmick game, like that needle in a haystack one or that family time one.
If your entire game revolves around a single gimmick, then how is AI the issue?
Anyone would've copied it, game cloning isn't a new thing either, the fact that it happens faster is not a problem either.

I never had a problem with the chinks cloning games before, because I'm only interested in getting the better product; if the guy stole your gimmick and then made a better game out of it... why should I give a fuck that you had the idea first?
Ideas is the easy part, making the game good is the hard and interesting part, and the better game will win out in the end.

I'm not saying you should just tell the AI 'remake gta san andreas XD' and then sell it as is, or rather, I don't have a problem with you doing that, but you're foolish if you think that will sell anything, and then you're fucking yourself in the long run because for the few that might fall for it, you'll just build up bad blood and resentment from the public and you'll be known as a filthy jeet.
And gaben's 2h refund policy will fuck you over anyways; not a problem.

>>110266270
Well that's just because the market will need to adapt.
Youtube is foolish to allow for such easy uploading of gigabytes and gigabytes of videos; but that won't last, they'll add limits, and the third worlders will get fucked over; and even then, how is that a problem?
How is that my problem?
Low quality content has always been a thing, this is the internet, 80% of this site is low quality content; how does that matter to me?
If I had a button to force it otherwise, I'd press it, but I don't live in lala land, sadly.

People will consume whatever they deem is worth consuming, and I don't think it's anyone's job to decide where the line needs to be drawn, it'll just happen naturally and organically, like all other things.

File: 1789562553334690.jpg (63 KB, 578x433)

63 KB JPG

>>110266129
maybe you're just low iq?

File: IMG_20261107_085812.jpg (486 KB, 1280x2197)

486 KB JPG

>>110266143
To answer your other question, I refined my core terrain system probably four or five times, until I finally got the main engine good enough. Today I'm hunting hitches and hiccups; it never ends.

>>110263682
>use AI

G O O D
O
O
D

>>110263925
>Bevy

First time I heard about it.

Use the one with the less retarded community.

>>110264610
You never stop being a 4nigger, pretender zoomer.

t. in 4chan since 2008.

>>110266286
Basically how hard the model works and thinks. Low will try to do it quickly, Extra will put lots of thought and research and experimentation into it, like create test programs you didn't ask for just to answer a question. Opu55y is known for being much better at going deeper and thinking of things you didn't specify when you go on Extra effort, so if you have a vague idea and need it go think about it thoroughly you want to switch to Extra.

Love using Bezi with Unity

File: 1787377935587441.png (150 KB, 357x305)

150 KB PNG

>>110266295
>I don't see it.
>Actually, I see it, but those don't count because I think they deserve to get stolen.

>>110266129
Theory-crafting games in your head has always been the subject of mockery in the industry.
You can't expect yourself to develop an idea without expressing it in some form and establishing a feedback loop. Without friction, all you have is masturbatory junk.

>>110266627
>>Actually, I see it, but those don't count because I think they deserve to get stolen.
Yep, not my problem, copyright is a meme and I don't care much about your emotional appeal.
Not that your reductionism is right, because that's not what I said, but you'll require to use your gray matter to understand the nuance here; I don't see it because it's not happening yet, those two examples I gave (which are the only ones we really have) are all schizobait and blown out of proportions, if you look at the games and their receptions on steam you'd quickly realize how right I am in my assertion.
And yet, I still don't care for it even if it was actually happening at scale.

Your own pseud-take on 'nooo you only need to put in hardwork but like, in the way I want to and everything else doesn't matter' shows you will never make it, you're just using emotions as a cope, and that's also not my problem.

>I don't see it.
>Actually, I see it, but those don't count because I think they deserve to get stolen.
>Actually, fuck copyright, not my problem. Stop being so emotional.

File: .gif (983 KB, 500x364)

983 KB GIF

>ctrl + f 'deserve'
>0 results
>I'm not being emotional but that's what you meant
take your estrogen

>110266712
https://paulgraham.com/submarine.html
tldr; You're parroting a news outlet that no one falls for, which is why you're paid to parrot it. Which is why it's pointless to even bother to point out it's made up and fake.

>>110266428
>Use the one with the less retarded community.
what did he mean by this

File: raids.webm (3.85 MB, 1280x757)

3.85 MB WEBM

I'm vibeslopping a custom engine similar to the PS2 in terms of specifications, but with limitation bypass and mixed rendering to either make more demanding games, or games similar to the PS1 or Amiga, or all at once. It's still on alpha, but it advanced well enough to start making some games.

I setup an autonomous game studio workflow to make some demos (research + creative vision + technical lead + implementers + QA). I asked for a rally racing game with MotorStorm and Twisted Metal modes and mechanics, and this was the result. It has almost all known PS2 render tricks, like the half-resolution rendering to make water refractions, object projection into textures (the shadows), stencil shadows, per-vertex distortions for heat effects, and more. The .webm is not good enough, but you get the gist.

File: raids2.webm (3.82 MB, 1280x751)

3.82 MB WEBM

>>110267386
more raids: shadow legends

does anyone think opus 5.5 got slow? it didn't get dumber or doesn't gobble up more tokens, it just takes a billion years to do anything

File: Game-2.mp4 (3.67 MB, 1400x740)

3.67 MB MP4

My character creator now has a game attached to it. A poorly animated unpolished game, but I'm working on it.

File: sotc-at-home.webm (2.64 MB, 1280x785)

2.64 MB WEBM

>>110267386
last one, I asked the workflow to make a Shadow of the Colossus at home, and while the animations and models are shit because Astra and Opus can't model for good, I'm surprised it got right the rest of details, like the bloom, motion blur, scenery, camera shakes...

File: 1790341754376377.jpg (39 KB, 512x512)

39 KB JPG

Been vibeing prototype after prototype to get a feel for what AI can do and to feel out which of my ideas are most worth perusing, but I have no naming convention for the projects' working titles, what do you anons do?

>>110267593
I just come up with WIP titles, sometimes I even let them come up with the names; and then more often than not doing that leads to me thinking of something connected to it and I just go with that

this is a non issue, you see plenty of games even in the AA field that don't have working project names and are called fucking 'unreal game' or other stupid shit like that, and then they change the name where it actually matters

>>110267386
>>110267410
I love what I'm seeing but I gotta admit: I'm not sure why I'd ever go with your engine if... there's probably dozens of projects like this but backed up by big, big money and having more than one viber on it working full time, I mean whoever wins the 'AI engine' bowl is going to be rich lol

>>110267659
I'm more just annoyed at how disorganised its making the folder I store all of them in, but claude made the (in hindsight very obvious) suggestion to number them so they stay in chronological order of creation lol.
But also coming up with a naming convention for working names is a peak ideaguy time wasting task, how can I miss out on that fun?

File: expansions.png (192 KB, 1103x597)

192 KB PNG

>>110267659
I would say the engine itself is just a part of all the whacky shit I'm going to do.

>>110267593
fart1, fart2, fart3, etc

File: animviewer.mp4 (1.21 MB, 1884x996)

1.21 MB MP4

I have a bunch of animations and I've been having a hard time browsing through them so I got claude to make me an anim viewer with tags, etc.

>>110267960
I feel like people still underappreciated how nice AI is for throwing together some custom quality of life software that only need for your use case but probably wouldn't have bothered to write yourself

>>110267960
Where do those come from?
Looks so good, maybe I should invest in a few animation bundles coz I can't animate for shit and the clanker is really having a hard time without references.

>>110267994
Absolutely, you can pretty much cook up a very niche but useful tool in less than an hour. It's insane

>>110268036
anon, the name is right there

>>110267960
mine did it for sounds, it's pretty cool, didn't even ask for it but now it's my go to because claude can't evaluate sounds properly

File: 2026-11-07_21-32.png (13 KB, 376x123)

13 KB PNG

>>110264566
>retarded jeet opinion
>is another wojak profile pic aigobbler

File: file.png (68 KB, 1131x386)

68 KB PNG

claude is pretty decent at making clothes
I thought it may struggle with collars because they're sort of complicated but no

>>110268133
That is neat. would you mind sharing some of your process?

trying to vibe code a basic fps game just for my own use out of interest with new opus 5.5 model but struggling to get claude to make realistic(ish) weapon models in blender? so far its just making blocky crap i could do myself. does attaching images as example help or best to just describe in detail?

>>110268272
I think claude's modelling skills are too limited for realistic weapons? Though if I am wrong I'd love to know too lol

>>110268272
LLMs can't model

>>110268297
Thanks, thats fair enough. only wanted to play with it as I saw a youtube vid the other day and it looked pretty good but guessing they just used existing assets from elsewhere

>>110268310
claude can interface with blender, gadot etc. but seems to just be writing code to make the model files. its still impressive and was fun to play with but not good enough and does seems an inefficient way to do things.

File: file.png (202 KB, 1559x857)

202 KB PNG

>>110268269
using marvelous designer python
>claude drafts a pattern via a script
>I do things by hand that the python API doesn't support, there's not a lot of this but for example MD won't let you sew between internal lines via python so I need to sew the front placket myself which is just 5 seconds of work
>i simulate
>I run another script which exports the result for claude to examine
I have no idea how claude is able to draft the patterns, or how it's able to examine the results, but it can.

If you asked about the process itself, it's a fairly standard one for clothes
>take measurements (I have a tool for this >>110260710 )
>choose a block(UK)/sloper(US) to start from (this shirt is using a bodice block)
>once that's done, add easing and seam easing (not seam allowance, you don't need that)
you can see excessive seam easing on the sleeve caps here, because they bunch a little where the sleeves meet the shoulders. This is a problem in general with Claude, because it seems to follow real (physical) patternmaking concepts but they don't always translate well to sims. For example real threads seem to "eat" some of the easing so you would use more than in 3D. What this means is this shirt needs less seam easing on the caps which we'll do in a while.
>start detailing and turn it into a proper garment, add the collars, the placket, the cuffs, etc.
This is where I end the simulation step. You can take it further in MD by adding a lot more details but for games in general this is about as far as you take it, then you do retopo and ZBrush work.

>>110268373
Very cool. Thanks anon!

File: file.png (2.15 MB, 1774x887)

2.15 MB PNG

>>110268272
yeah, claude on his own is not going to give you any realistic weapons
try giving him a reference like picrel (astra just made it up, I 98% expected her to say no because it's a violent picture but sometimes she's fine with them, idk)

references are king, always, no reference makes things blurry and fuzzy, and they get very retarded very fast, the more reference you can feed them with, the better
better yet, but more expensive token-wise, you separate this picture into 4 distinct ones and give all 4, that's less information the model will have to go through and 'filter out' before getting what it wants as it develops the gun in blender

File: rifle-component-atlas.png (834 KB, 1480x1790)

834 KB PNG

>>110268776
better yet, you color code everything, but this gets very expensive because you'll probably want to create the gun in 'bits' and pieces, say focus on the muzzle and the grip first, then do the middle part, then the trigger area + shoulder, etc etc

>>110268776
>>110268831
NTA but aren't you making low-poly games? wouldn't it be easier for the clankers if you generated references that are low poly to begin with, rather than giving them realistic-looking stuff as a foundation that they need to strip down themselves for ultimately low-poly models?

File: file.png (807 KB, 2089x856)

807 KB PNG

>>110268991
oh yeah, the two pics I posted there aren't for me, I just genned them to show anon how I'd do it

though the witch game isn't exactly low poly per-se, but I'm planning on going with cel shading for it so I'm not going to go too far but definitely not something most people would call low poly (2k for a simple weapon is pretty average, I think, while the character models go up to 10-15k)
but yes, otherwise, it is easier to get low poly references directly, but not so much because it's easier for the clanker but more so because it's faster to imagegen your model and see what it would look like as low poly (aka, which part can give up on details and which part you want a focus on) rather than just produce the model and hope it turns out okay (which, more often than not it's 'fine', but the topology will need some iterating because it might've given the ears 90 triangles each, instead of 20, and used those extra triangles on the elbows/knees, etc)

File: game.webm (3.66 MB, 1372x720)

3.66 MB WEBM

Asked claude to make a game inspired off the PSP Dissidia games to see how it does and it seems to have come out better than I dared hope for desu

>>110269345
damn that's pretty cool yeah

File: SA80.png (687 KB, 1594x919)

687 KB PNG

>>110268776
>>110268831

Thanks! Looking surprisingly good!

File: file.png (1.22 MB, 1280x720)

1.22 MB PNG

I want to make a car interior (like this farcry screenshot) that's customizable (replaceable / modular part)
But I don't know what tools can I use to achieve that.
Can anybody help?

>>110269992
blender MCP and a heap of tokens

File: cockpit editor 04.mp4 (3.33 MB, 1564x722)

3.33 MB MP4

>>110269992
I haven't touched my editor in a while, but: make your own tool.
You can probably have the clanker make a few 'mockup' and 'bases', and then you create a bunch of different gizmos and different props to add to it, and your tool will be just about adding it and moving it where you want it to go and connecting it to whatever function your game needs (gas, oil, whatever you need)

This is the result of maybe at most 8 prompts over a month (iteration, addition of the new systems, etc) and not more, and it already fills any and all needs I could ever hope for on that front.

If your goal is to allow the player to place different things at different spot themselves, then work with simple 'slots', exactly the same way, except they all have minimum and maximum sizes allowed, shapes, whatever, and then you just leave those slots around the cockpit while keeping into account how much space they need to fit, and voila.

>>110270038
I won't need super detailed 3D models.
Even some simple geometry with no textures would be fine.

>>110270079
>If your goal is to allow the player to place different things at different spot themselves
I'm more of thinking changing the design every so often, so your tool idea is gonna works just fine.

I'm just wondering how can I just prompt it to create the gizmos & props instead of doing it myself (on blender)

>>110270187
prompt it saying you want a modular design and simple 3D models

>>110270194
Claude can do it all by itself?
I don't need other AI or Blender MCP?

>>110270197
i mean if you want it to do 3D models for your game yeah you should let it use Blender. the rest it can do all by itself, implementing the dashboard system in the game too.

>>110270197
>>110270187
from my own testing: blender MCP is a meme, it was an important tool for the older generations; opus doesn't need that shit anymore, I haven't noticed any improvements at all to the models it can create using an MCP vs not using one, but I guess ymmw, it's only a few minutes to install it anyways so who cares

and all the models I have in my cockpits are all completely made by claude or codex, I did zero models for these and didn't even touched them or opened them in blender: because it's a godot game anyways and all the models you see there aren't actual models, just code
the way I did it was pretty simple: 'here are the rules, I want components that can connect to X and Y functions of the game and show diegetically the different states of those variables (speed goes up -> needle moves), create 12 different props that would fit within the cockpit of X type of vehicle' and voila, and then I did it again, and again, and I ended up with 212 different gizmos for the different cockpits for the different manufacturers, which was overkill mind you, but a lot of them were pretty meh quality anyways

>>110270278
>opus doesn't need that shit anymore, I haven't noticed any improvements at all to the models it can create using an MCP vs not using one
Wait, you mean I can just ask Opus to give me a 3D model?
And it actually generates one?
I gotta try it first then.

>>110270294
There's honestly not a lot of things claude (or codex) cannot do at this point; mind you, the claude app is a pain in the ass with security/permission shit and you need to install plugins and whatever, I mostly use claude in CLI (and codex in app), and the golden rule is:
Ask it.
Just ask it if he thinks he can do X or Y, and he'll tell you if there's something he can't do, but using blender or something?
That's easy as fuck.

I even had codex spend an hour downloading and activating a bunch of individual bundles from humble bundle (and then connecting to the various storefronts from which the bundles came), because I was too fucking lazy to go and spend 10 minutes doing that.

The amount of shit this generation can do is fucking insane.

File: Screenshot From 2026-11-0(...).png (105 KB, 830x677)

105 KB PNG

Both claude and codex have cucked to fucking disney, when did this fucking happen?
I just tried deepseek and even that chink fagget wont do it holy shit.

>>110270361
been a while now, same for anything pokemon or nintendo in general (even trying to get 'nintendo 64 graphics' flags the request)
this is why local is the future

>>110270361
now you know who owns you anon

>>110270361
it's really gay and useless, i couldn't care less about disney or nintendo ips but the other day I wanted to generate some concepts for a monster and it refused for being too gruesome, when all I wanted to do was do a fairly basic bitch ghoul

>>110270386
how long until we have an opus 5.5 tier model in less than 500B params and have cheap enough hardware to actually run it

>>110270405
3 to 6 months.

File: 1771228188640628.jpg (117 KB, 640x1015)

117 KB JPG

>>110270386
>this is why local is the future
I trust in nintendo to find a way

>>110270425
prayge

>>110270386
>local
not even local, just ask chinese models

File: file.png (338 KB, 1696x1343)

338 KB PNG

>>110270340
What... The.... Fuck....
It took 15 minutes for it to do something that will take me probably a whole month to do...
And at this rate, I can probably make a game out of it in a week or so...

File: animlib.mp4 (2.08 MB, 1912x1010)

2.08 MB MP4

I am thinking of building a large library of animations from a lot of different games, for reference mostly

File: file.png (201 KB, 1917x959)

201 KB PNG

Working on a vibecode starterpack. The two goals right now are p2p multiplayer, and OPFS. I'm working on getting Poki's netlib working so the AI can write one source of truth for game logic, and then the player can open their game up to other players and send them an invite code for serverless p2p multiplayer. The second goal is version managing and game download so the user can download the game files and assets locally instead of in cache, and if the game dev updates it only updates the files that need updating.

I want to get it done then vibecode a CRPG like Baldur's Gate or Divinity Original Sin that can live on the web.

>>110270476
why are you driving on the wrong side of the road

>>110270576
Oh, see that customize option for driver side?
I just changed it to the right to test it out.
The amazing thing is that I didn't even prompt it to do all these.
I was just questioning if it can do a low poly car with modular dashboard parts.
And then it decided to "prove" that it can do it by putting this all together.
I didn't even request a single feature in that except for "low poly car"

>>110270576
Oh if you mean on the road, I don't know, it's a one way street or something

File: .png (436 KB, 750x1000)

436 KB PNG

>>110270476
welcome to the future old man

>>110270612
>I was just testing an option mate haha
shut the fuck up aussie, we know who you are now

>>110270492
Just use seed-based pseudoRNG. That way you have deterministic MP that shouldn't desync.
Only works with friendslop though coz of no good way to fight cheating.

File: Screenshot 2026-11-07 at (...).png (71 KB, 687x996)

71 KB PNG

do you guys think i should go through with making assets in my style or should i just buy a pixel art pack? these are character portraits

File: 1770595659001497.jpg (48 KB, 512x413)

48 KB JPG

>Claude made my game too hard
lmao, I'm such a shitter. How do I know if I just need to get good, or nerf the AI?

File: A-Market-For-Lemons-6-320.jpg (14 KB, 320x240)

14 KB JPG

>I'm a prompt engineer

Is reverse engineering still possible? Really late to this vibe coded shit and all I want to do is reverse engineer the Pokemon colosseum and XD to make an original game with original assets and animation

>>110270787
lol

>>110270736
pretty much depends on the game and how important the style is

>>110270751
I like making them generally too easy for me, and I just plan on bumping up the difficulty later, because this allows me to test shit out more freely (animation, timings, etc)

>>110270787
it doesn't matter, because chances are, you're not going to want to use the same code as a fucking 25 year old console game, anon
you can still tell it 'I want to make a game like X' and it'll do just fine, and that's the smartest way to go about it anyways

>>110267419
I loved Mega Man Battle Network when it came out.

File: cinnamon-2026-11-07T22340(...).webm (2.95 MB, 1920x1080)

2.95 MB WEBM

>>110270724
It's for a CRPG. It works though, just got done testing it.

File: ocr.jpg (46 KB, 1024x610)

46 KB JPG

>>110267410
that actually looks really good. but gameplay is unappealing. the game does look sweet tho

File: evnr1ualwnpo1.jpg (18 KB, 249x275)

18 KB JPG

>>110270917
just use spacetimedb what are you even doing slopping netcode.

>>110270976
Wait did you just guess I was the spacetimedb guy by my webm or was that coincidence?

File: 1794103613117909.jpg (31 KB, 720x586)

31 KB JPG

Time for (You) to show everyone how it's done in 2026

File: 1779639504522304.png (121 KB, 527x622)

121 KB PNG

>>110270982
i have no idea who you are dude, spacetimedb is just always the answer unless your engine is completely different from literally any normal engine, as is the case in my case, and i still basically recreated spacetimedb in my shit.

>>110271000
I want something that can live in a static webpage with no backend though lmao. I just want it so I can vibecode games like a CRPG or something, but the player can send an invite code to their friends. I'm not really slopping though I'm using netlib. I was going to slop but found out what I wanted already existed. I don't need a backend for just netplay with friends in a turnbased CRPG. If people want to cheat at a fucking RPG or friendslop they're ruining their own experience and their friend's experience, not other people. The friend can just stop inviting them lmao.

https://github.com/poki/netlib

>>110271038
>I want something that can live in a static webpage
And how is this possible? You still need a server to sync the actual bytes.

File: xvj1q4vuhcd81.jpg (107 KB, 750x735)

107 KB JPG

>>110271038
>The Poki Networking Library is a peer-to-peer networking library for web games, leveraging WebRTC datachannels to enable direct UDP connections between players
oh i see so like a torrent/matrix, interdesting. can that support stable gameplay tho? i def see the appeal of the idea.

>>110271049
webRTC is p2p. I think you need a server to connect them at first then it's p2p. I don't know exactly what happens completely under the hood, I've always used SimplePeer which included it's built in stun server so I assume you need a server to make the initial handshake. But there's tons of sites online with stun servers for free and Poki let's you use whatever server you want for free, or by default their own servers.

>>110271065
>can that support stable gameplay tho?
I mean I've had big chatrooms with video data being transferred to one another through it with multiple people's cameras up so I don't see why some game data shared between a few friends would cause any problems.

Vibecoding contest for the web. Winner gets $100k. Anyone want to collab enter it and split the profits if we win?

https://x.com/0xRishi/status/2119178512665845854

>>110271429
Why share with you when I can have it all for myself?

>>110271450
Multiple heads are better than one. None of us alone are beating those guys who spend all day on X making three.js vibecode games :^)

Alright fellas got p2p synchronized physics in web. How should I handle host leaves? Pass to another player? Or just end the lobby?

I could do either with Poki but for something like a crpg where host is saving the game data, obviously closing lobby is ideal. But with a multi lobby game like a shooter or vrchat the lead should get passed around. What should I do first?

What do I use to generate unique skill and ability icons in pixel art?

File is too big to post directly I guess.
Biggest obstacle right now is adding more assets and animations without wasting too much money.
https://files.catbox.moe/9q99ri.mp4

Can't seem to get Claude to cooperate with the request of :
>make anime styled models with swappable, randomized body parts that can be changed later on in the game

>>110270736
your style seems to have its own charm and seems drawn well enough so It depends on what your goal is and how you want it to look

File: orc_female.png (238 KB, 1600x512)

238 KB PNG

so this is the power of ai...

File: 1778375771260421.webm (3.78 MB, 1280x674)

3.78 MB WEBM

I made an animation editor so I can make tweaks to the outputs of my procedural animation system. These will give positive examples for Claude to hill-climb towards (as opposed to just fitness rules like don't break limbs etc.)

Said differently, I can now tell Claude "how" motions are supposed to look with much more precision by editing the existing issues out of motions to give him a positive example.

Before, I was just using video references and premade libraries, but those don't exactly exist for people using halberds to pole-vault around like a maniac.

This animation was recorded before burning tokens on the fitness hill-climb so it looks pretty bad as-is.

>>110272216
>make anime styled models with swappable, randomized body parts that can be changed later on in the game
Making decent anime-style models, even in a game with a simple style like yours, isn't something that you're going to have much success with. Many, many anons have been trying to do exactly that in these threads, and none have succeeded in my opinion.

If you give Clause a good base model to work with and a small amount of clothing, then it should be able to make variations on that starting clothes set.

>>110270736
Those look really good, I like them. They have personality. Add one or two colors for shading the skin and they'd be perfect.

I'm creating this incredibly ambitious (and I think very cool) comfy space exploration game and I'm growing anxious thinking nobody besides me might ever see it.
Yeah, I know I can just post it on itch or something (I'd rather make a site for itself), but what if I never get it to a point where it's good enough for others to see?
I'm such a pussy god damn.

>>110267453
there are incppmplete decomps floating around and I beleive a build with debug symbols so yeah no shit.

>>110263999
I used to fret over spending 40 bucks on a game I might not play, but after getting claude max and codex plus (for image gen mostly) I've realized making games effortlessly is the best game I could buy. It's worth every cent, seriously.

>>110272621
Pigheaded confidence is the most valuable attribute to instill in oneself when vibe-coding. If it isn't good, then just make it good? It's not like you won't be able to tell when you're having fun playing your game or not.

If you really aren't sure, you can post it here and I'm sure some kind anons will shit on it if it's sub-par.

>>110272653
I just shit out an undigested dorito

>>110272410
Where can I follow you? You have the most interest project on the daggy generals currently

>>110265238
I do mostly two things at once on a single project (but I can work on multiple projects at the same time). If the project allows it because there's multiple areas that can be worked on concurrently, tell claude to set up a workflow with a separate worktree, and when both tasks are done and commited, tell one of the sessions to merge the worktree back into main.

>>110266129
Something similar happened to me, but while contesting the science behind the game's mechanics, its limitations made us change the idea to something that was better without realizing.
Talk it out with your clanker.

>>110272701
it's too much of a yesman to talk it out.

>>110272685
Wow thanks so much!

I am just posting my stuff here for now. I might start uploading some videos to YouTube when I get to a point where the project is more presentable. It might also be easier to convey what I'm actually doing in a longer format than a 4chan post.

I will definitely shill any YT channel or blog that I start here when that time comes.

>>110266129
No million dollar game idea is actually good the first time around. You make a first version that isn't fun, tweak it, change it, add and remove features, and actually craft a fun system over time. That is how all real game development works. You cannot just one-shot fun.

You have the wrong mindset. You need to treat it like an iterative process.

>>110270294
It can feed python code to the blender CLI and it generates the models that way. It can generate renders to check its work, or screenshot them in-game.

>>110272397
Would.

>>110272703
Tell it not to be. Establish criteria.

>>110272410
oh god you're mogging my try at it kek
good shit bro, keep going, you're doing god's work, I really need that shit for my projects tbqh

think you could take a look at adding a more 'precise' version (that would work in the same way) for FPS animations, like reloading and shit?
because I honestly have no idea how I'm going to get around to deal with that with my fantasy guns

I'm trying to figure out how to do clothing on a sprite in a sidescroller, it's actually way more difficult than I thought and really exposes the limitations of LLM's - they do a pretty good job but for the final 5% and it ends up looking like shit

My uncle's friend work at the deepseek and he taught me that by placing a particular character at a particular position in my prompt it uses the deepseek-5-dev model FOR FREE and holy fuck, lads. This is even better than Opus 5.5

>>110272621
there's no right answer for me to give you, because it depends on who you are and then there's going to be the luck element of 'who' sees it and when, the issue with AI projects like ours is that there's ideological blockers going on and a lot of retards will be disingenuous about your project, well, this happened before AI too, but yeah if you're scared of receiving bad critics, there's just no easy way around it

at some point you just gotta swallow the pill and stop caring, that's how I do it at least, just do whatever you want because you can do whatever you want, take the criticism as you believe it should be taken, keep your feet on the ground and just make game
it's not like iterating or changing things is hard now with the clankers, like I keep changing shit all the time, even fundamental shit that would've taken a team weeks and weeks of refactoring and engine changes to get it right; but now clanker-san will do that in 30m and you can iterate and shift focus on a dime

80% of gamedeving is motivation, 10% is idea'ing, 10% is understanding what you're making and why you're making it, so whatever gives you the motivation, wherever it is you get it from, get it

>>110272809
it's deeply rooted in its programming.

>>110272793
Yeah, but does it even looks good?
Anybody have some examples of models generated only with Claude (and Blender?)

File: 1776161847821426.mp4 (3.64 MB, 790x808)

3.64 MB MP4

>>110272848
I haven't forgotten! I'm going to upload a version of the project with just the procedural animation system and character controller here as soon as it's ready.

And yeah I have thought about doing a 1st person version of the system. It would be much simpler honestly- the rig would only need the hand positions in screen coordinates, distance from the camera, orientations, and the medial axis of rotation to determine where the elbow is pointing.

For the hand poses themselves, I did make an FK/IK grip-solver to generate rock climbing holds a while ago that look okayish at a glance, picrelated. I can quickly anonymize that and share it if you want to use it yourself. You might be able to get claude to adapt it to holding onto guns and whatnot. Just lmk

File: SCAV CRAWLING IN MY SKIN.webm (3.94 MB, 2048x836)

3.94 MB WEBM

>>110272903
There are plenty around; and the answer is no, they don't look good.
see >>110272397 for example (tho it might've been gpt making that one, but honestly they're about on the same level for 3D)

It CAN look 'okay' after a lot of iterations, but it's really not efficient and it'll cost you a lot of tokens for a subpar result.
Pic rel is a very old webm of one of my projects, and at this stage it was 100% all made with claude, including all the animations (and the texture was gpt I think).

File: file.png (1.1 MB, 1592x895)

1.1 MB PNG

>>110272903
I've been modelling machines from IRL pictures and generated concept art, and they are decent (mind you they're untextured; I'm trying to decide if I'll go with that look or not). Their animations are also solid. I think as long as you don't do organic stuff, you're good.
see >>110269679

File: file.png (820 KB, 1592x892)

820 KB PNG

>>110272903
>>110272930
I guess it's a matter of knowing how far you can ask it to go before it starts falling apart.

>>110271637
I can't begin to imagine how quickly those guys can type "make game make no mistake". Scary stuff.

>>110272930
>>110272940
This looks cool. Some kinda horror game? A submarine simulator?

File: file.png (1.88 MB, 1581x884)

1.88 MB PNG

>>110272903

>>110272930
>>110272940
>>110272946
gives me Stationeers vibes

>>110272943
Thanks. It's a space exploration game trying to be a mix of Noctis IV and 0x10c. With some Orbiter. You alone in a universe that has aged past you (due to all the hibernation you do while traveling or something, I haven't decided), in a hard sci-fi ship with as many systems modeled realistically as I can manage. I want to get a nice cozy spooky vibe and do things that make the player feel really small. But it's still very early on.
>>110272950
Yes, but Objects in Space is more of an inspiration. Sound is very important for the ambiance.

File: file.png (602 KB, 1591x894)

602 KB PNG

It's crazy how you can just decide to do something stupidly high-maintenance, like keep an in-game eink reader documenting all the systems, and Claude will just do it and keep it up to date as it does other things.
We truly live in the future.

File: 1783780675145933.png (1.31 MB, 808x960)

1.31 MB PNG

>>110272964
Very cool. Will you be on a rotating section for artificial gravity? Or will you have zero-G navigation around the craft while it isn't under acceleration?

The IRL spaceship interior aesthetic is super underutilized, in my opinion.

>>110272993
Only realistic gravity. Depending on how much acceleration the landing and main thrusters do, it might be interesting to see if it's even possible to move around while they're in operation.
I also want to have a rotating section at least on a wreck or with some other excuse, since the minimum dimensions for that to work realistically are too big for the player's rust bucket.
There's so much to do.

>>110273012
Having the ship reorient around you as it fires the RCS thrusters and whatnot could be a pretty cool experience.

>>110272984
yeah, keeping documentation up to date used to be that insane chore that nobody ever wanted to do and now it just...
never is a problem ever again

I might actually become the bottleneck m for my Claude $20 subscription
I'm in my second week and I have already used half my weekly limit, but this time I'm getting slowed down user testing the in game map maker it's building me, and I'm still behind on importing more basic assets to start building, and I haven't even quite started genning assets to gen my main player models and items.
Eventually I'll run out of issues and tickets and have to leave my quota to lapse while I catch up at this rate

>>110273111
Always plan ahead and try to prepare for longer, token heavy tasks to be done in these cases; refactors, heavy system changes, maybe even simply a different way something is handled; 'what if we used dithering for the shadows', 'what if we used shaders for this specific things instead of relying on sprites and textures', etc etc etc

My launchers (I always create launchers for any projects) are filled with test scenarios like this that allow me to test specific things or changes, rendering, how we handle shadows, being able to easily switch something on or off.

Optimizations are also a very easy token black hole to invest into; let's benchmark the game currently as check where the most likely bottlenecks are, let's study the frame timers and see what easy wins we could get and then see if there are heavier refactors we could plan for for later' and boom, all of a sudden you get million token tasks.

>>110273111
Find a way let Claude test it by itself.

>>110272915
NTA but 2nding request for anything that'd made FPS animations easier to implement.

Oh man game UIs are gonna get cool as fuck on the web

https://x.com/egorvert/status/2118372794568450289

File: merged.png (419 KB, 543x701)

419 KB PNG

>>110272624
you say that, but Astra liked to ignore all of that and do its own thing, because turns out telling Astra to do something "like XYZ" doesn't translate on strong references the AI can adhere to, all while Opus doesn't have this problem (and probably the reason why Opus cost more). That is the base of my current skill/workflow, to make the AI do its own research and generate actual references and stick to them. See pic related.

>>110273216
Yeah, after this current huge long chain "continue" main prompt with side quests I still have a simplification and efficiency thing to do, but it may not get much because I'm already rebuilding the basic engine in a modular way and am good at writing what I want the first time.
Once I'm done with the new core I have more TF2 map loading issues still too fix too, but after that it's onto enemy AI and unit designs so I need to get some non placeholder models soon.
>>110273229
Claude already wrote all its own user tests on its own for its cloud sandbox that it copies the program to every new context transfer, I had to rein it in from writing and running too many unrelated ones, but some things are just better user tested directly. It tests basically everything it can and just has me confirm it works when I build it and as I meant.
I don't code at all yet I merely describe in Cowork.

File: 17867039596880160300.jpg (122 KB, 849x695)

122 KB JPG

>>110272806
>>110272927
claude oneshot that after i described what i want. i posted it for lolz, and i know a pathway to make claude model something which will get dicks hard throughout this thread.

File: file.png (19 KB, 761x81)

19 KB PNG

>ask gemini to explain a game design concept
>food analogy

>>110273614
is it wrong tho
I mean it is, it's a pretty retarded way to explain this, but also, I don't think you should ask the clankers such questions in the first place, ofc it'll pull shit out of their ass

'gunplay' is 95% feedback, and 5% gameplay
shooting the k98 in brothers in arms felt fucking fantastic (well, for the time), even if it's a shitty bolt action, but the sound, the animations, the recoil, all of that feedback was 10/10 (don't look it up now, it's fucking garbage, but that's how I felt when I was a wee lad playing it)
you need good sound, and good effects (including recoil), and that's already most of the work done, then you add feedback from the enemy (shoot at the foot -> falls/stumbles, shoot at the head -> crunch sound effect or helmet flying off etc) and you're golden, no matter the gun, no matter the technology

>>110273614
The less universal an analogy is, the worse it serves its purpose. A highbrow analogy is facetious. Everyone eats so that makes food analogies innately promising no matter how much Europeans cry about it.

File: file.png (400 KB, 1898x989)

400 KB PNG

building a massive animation library for references
I've got 2.3k anims already and I'm barely getting started, my goal is to scrape animations from lots of games into a browsable repository

the viewer is pretty much done, has tags, categories, groups, etc.
proven to work with WoW, Black Desert and Lost Ark but essentially any game that you can extract the animations from you can bring them in, and then it's a simple export to Blender via glTF.
And it's LLM friendly

>>110274162
I'm also interested in doing something like noclip.website (fantastic site for anyone who doesn't know about it) where I try to extract whole maps to explore in first/third person, but I'm less confident about it. Still would be great for studying level design.

File: 1792189662986000.png (156 KB, 237x527)

156 KB PNG

>>110273614
gemini is dogshit. why would you use anything but claude xhigh? with ai, you just need the best, and that's claude.

>>110274276
you look like THAT?

>>110274162
So mostly ripping animations from other games? Pretty awesome. Now you just need to use it as training data for an animation gen LLM to make create legally usable animations