Building Game Prototypes with LÖVE(healeycodes.com)
healeycodes.com
Building Game Prototypes with LÖVE
https://healeycodes.com/building-game-prototypes-with-love
80 comments
MOONRING too which is another really fun game, although one that hasn't yet gone as mainstream as Balatro.
There is so much "draw the rest of the Owl" between a very primitive framework like LÖVE and stylistic games like BALATRO and MOONRING.
There is so much "draw the rest of the Owl" between a very primitive framework like LÖVE and stylistic games like BALATRO and MOONRING.
Balatro recently published a mobile version of their game, but I was under the impression Love2D only supported desktop platforms. Has that changed?
Yes, https://www.love2d.org/ has links to both iOS and Android SDKs.
For Android you can also install the stand-alone LÖVE Loader app that can open .love files to play. Some games on itch.io comes with a .love-download (in addition to platform-specific executables), but I guess they can sometimes be extracted from other games too.
The .love-files, if provided or if they can be extracted from a game, are also useful on desktop, for playing the games in an operating system (like FreeBSD) that the game's developer did not provide binaries for.
The stand-alone app is also a fun way to do some gamedev on an Android phone. The .love file is just a ZIP, so it can be easily created from termux.
For Android you can also install the stand-alone LÖVE Loader app that can open .love files to play. Some games on itch.io comes with a .love-download (in addition to platform-specific executables), but I guess they can sometimes be extracted from other games too.
The .love-files, if provided or if they can be extracted from a game, are also useful on desktop, for playing the games in an operating system (like FreeBSD) that the game's developer did not provide binaries for.
The stand-alone app is also a fun way to do some gamedev on an Android phone. The .love file is just a ZIP, so it can be easily created from termux.
I downloaded a mobile version before the official release (I uad bought it on steam already). Somebody made an apk with some löve-to-apk generator, and one or two fixes for the shaders on mobile.
Love may not but there is nothing inherently unworkable with Lua and iOS. I released an iOS game which used luajit on ios and it worked great.
It is pretty easy to build for iOS, Android, and web.
Getting it built isn't too bad (at least iOS and Android, Web was giving me WebAssembly errors I couldn't figure out), but accessing native functionality seems harder and do not have any good tutorials (a couple really old discussions online that are probably out of date is all I could find).
Like if you want to release a game on iOS you're going to need some sort of Ad framework and access its game center for achievements and leaderboards at the very least, and that's been tricky for me to get working. I know you need to write some hooks and saw some old code that I'm trying to use for reference, but it was still giving me enough issues I gave up on it for a the time being.
And I've released games in native iOS before (although it's been a while), so it's not like I've never used them.
Like if you want to release a game on iOS you're going to need some sort of Ad framework and access its game center for achievements and leaderboards at the very least, and that's been tricky for me to get working. I know you need to write some hooks and saw some old code that I'm trying to use for reference, but it was still giving me enough issues I gave up on it for a the time being.
And I've released games in native iOS before (although it's been a while), so it's not like I've never used them.
Why would you need an Ad framework and achievements? Isn't balatro a prime example of a game simply being a paid app? No ads, no bullshit. Just a game and a pricetag.
Balatro doesn't have ads but it does have achievements and game center.
As much as I'd like my game to spread as far and wide as Balatro, I understand that it won't, not anywhere close. It's just not going to have the polish that Balatro has (I don't have the time or energy, I'm in my 40s with a spouse and a day job).
But hopefully it will be fun to play (it is a more fully-featured remake of an old free web game I made many, many years ago that was played millions of times).
And games that aren't already super high quality mostly don't sell on iOS anymore, at least not before people download it for free and try them. You pretty much have to make the base game free so people will download it and then provide an unlock later.
I'm not trying to be obnoxious about ads. I'm wanting to use the Zach Gage model (Knotwords, Flipflop Solitaire, Good Sudoku, SpellTower). Some minimal ads in between matches (the matches last 5-10 minutes by themselves) and a few extras like themes or detailed stats behind a one-time unlock of a few bucks. I hate the standard model of throwing ads in your face every 30 seconds and wouldn't do that for my game.
Also another native feature that Love2D doesn't support is the iOS Haptics library. Love2D only allows for a half second vibrate at full power and it's way too much. Balatro uses the iOS native haptics library as well (it's obvious, they're very light and quick vibrations). These haptics are super common in iOS games nowadays so it will seem weird if it doesn't support them properly.
As much as I'd like my game to spread as far and wide as Balatro, I understand that it won't, not anywhere close. It's just not going to have the polish that Balatro has (I don't have the time or energy, I'm in my 40s with a spouse and a day job).
But hopefully it will be fun to play (it is a more fully-featured remake of an old free web game I made many, many years ago that was played millions of times).
And games that aren't already super high quality mostly don't sell on iOS anymore, at least not before people download it for free and try them. You pretty much have to make the base game free so people will download it and then provide an unlock later.
I'm not trying to be obnoxious about ads. I'm wanting to use the Zach Gage model (Knotwords, Flipflop Solitaire, Good Sudoku, SpellTower). Some minimal ads in between matches (the matches last 5-10 minutes by themselves) and a few extras like themes or detailed stats behind a one-time unlock of a few bucks. I hate the standard model of throwing ads in your face every 30 seconds and wouldn't do that for my game.
Also another native feature that Love2D doesn't support is the iOS Haptics library. Love2D only allows for a half second vibrate at full power and it's way too much. Balatro uses the iOS native haptics library as well (it's obvious, they're very light and quick vibrations). These haptics are super common in iOS games nowadays so it will seem weird if it doesn't support them properly.
Sure, making money with games is hard and I guess you have to have ads unless you strike it big. I just wish mobile gaming went a different way, but that's not the world we live in.
Best of luck to you and your project.
Best of luck to you and your project.
I'm also planning to release it on Steam, and it will be an upfront cost there.
Also after paying the 'remove ads' price, which will likely be cheaper than the price I'm planning to sell it for on Steam (thinking $2.99 for mobile right now), the game will effectively become the same as the Steam version anyway.
I agree that it sucks. I wish I could just put $2.99 as the price and it would still have a chance of selling decently. One of the iOS games I made a long time ago I did exactly that (actually I priced it at $1.99). But people were also willing to pay $1-2 for Angry Bird-style mobile games by the millions back then.
Now there are so many free games on there, they seem to just skip over to the next free game instead, and maybe pay to have ads removed (my own partner does this, bounce from free game to free game, and if she likes the game she pays to remove the ads).
Also after paying the 'remove ads' price, which will likely be cheaper than the price I'm planning to sell it for on Steam (thinking $2.99 for mobile right now), the game will effectively become the same as the Steam version anyway.
I agree that it sucks. I wish I could just put $2.99 as the price and it would still have a chance of selling decently. One of the iOS games I made a long time ago I did exactly that (actually I priced it at $1.99). But people were also willing to pay $1-2 for Angry Bird-style mobile games by the millions back then.
Now there are so many free games on there, they seem to just skip over to the next free game instead, and maybe pay to have ads removed (my own partner does this, bounce from free game to free game, and if she likes the game she pays to remove the ads).
Have you looked into distributing on Apple Arcade? I haven’t myself but it seems like it might be another potential source of income that doesn’t rely on ads/in-app purchases. I think Google Play has a similar game service?
From what I've heard, Apple Arcade won't even consider you for Apple Arcade unless you're already established and have made popular games in the past, and they're very selective even then. So there's probably no chance my game would be accepted for it.
This is one comment that corroborates that:
"It requires support for Mac OS and TV OS and committing to significant content updates every year for three years, at minimum. Revenue is based on how many people play your game compared to everything else on the platform.
Because of the contractual commitment and the need for marketing spend (If people don't play, you don't get paid), it's hard to recommend for a small developer without commercial experience, which is why Apple is rather particular about who they approve."[1]
[1]: https://www.reddit.com/r/gamedev/comments/dwwi87/experience_...
This is one comment that corroborates that:
"It requires support for Mac OS and TV OS and committing to significant content updates every year for three years, at minimum. Revenue is based on how many people play your game compared to everything else on the platform.
Because of the contractual commitment and the need for marketing spend (If people don't play, you don't get paid), it's hard to recommend for a small developer without commercial experience, which is why Apple is rather particular about who they approve."[1]
[1]: https://www.reddit.com/r/gamedev/comments/dwwi87/experience_...
My 2c having released both premium and free apps on iOS and Android: premium apps do not sell. People don’t want to fork over any amount of money just to try an app that they may or may not enjoy. You can still make _some_ money, and get some downloads, that way, but it’s just completely incomparable to making it free.
Some rough numbers from my recollection: I released Card Crusade (deck building mobile game) for $4 back in 2019 and i think in the first year we got about 2k-3k downloads total. When we made it free in early 2020 we got 25k downloads in the first _week_. The revenue was less for us as a free app, but that’s with no ads and IMO we designed the IAPs pretty poorly.
But, a while after making it free, we accrued enough positive reviews that the iOS App Store started recommending the game to more people, so we have a steady 200+ downloads per week, despite not having updated the game since 2020!
I released another premium mobile game, Barnard’s Star, back in 2022 (and am still slowly updating it). Coincidentally, both apps make roughly the same amount of money (about $500 every 6 months), but Barnard’s Star makes that money from a fraction of the downloads… but the word of mouth effect (and game community) seems like it would be a lot stronger with f2p. So I’m planning on making Barnard’s Star free eventually (with hopefully better designed IAPs this time!).
One place where the “just a game and a price tag” model still works well is Steam. If i were trying to make a living off making games, I think I’d focus on releasing stuff there.
Some rough numbers from my recollection: I released Card Crusade (deck building mobile game) for $4 back in 2019 and i think in the first year we got about 2k-3k downloads total. When we made it free in early 2020 we got 25k downloads in the first _week_. The revenue was less for us as a free app, but that’s with no ads and IMO we designed the IAPs pretty poorly.
But, a while after making it free, we accrued enough positive reviews that the iOS App Store started recommending the game to more people, so we have a steady 200+ downloads per week, despite not having updated the game since 2020!
I released another premium mobile game, Barnard’s Star, back in 2022 (and am still slowly updating it). Coincidentally, both apps make roughly the same amount of money (about $500 every 6 months), but Barnard’s Star makes that money from a fraction of the downloads… but the word of mouth effect (and game community) seems like it would be a lot stronger with f2p. So I’m planning on making Barnard’s Star free eventually (with hopefully better designed IAPs this time!).
One place where the “just a game and a price tag” model still works well is Steam. If i were trying to make a living off making games, I think I’d focus on releasing stuff there.
Yeah, I'm planning to release the game on both Steam and iOS. With Love2D, I'm able to use mostly the same codebase outside of native iOS or Steam features I'd like to add.
Thanks for helping to confirm my suspicions. I used to release mobile games myself but it's been well over a decade. The landscape was a lot different back then.
Thanks for helping to confirm my suspicions. I used to release mobile games myself but it's been well over a decade. The landscape was a lot different back then.
That's true.
I have managed to expose some native apis from iOS, web, Windows and Steam but that is far less straightforward and the LLM is sure it knows what to do and it does NOT. Web searches are not great either.
I have managed to expose some native apis from iOS, web, Windows and Steam but that is far less straightforward and the LLM is sure it knows what to do and it does NOT. Web searches are not great either.
If you've found anything online that has helped you get that working, if you remember and don't mind sharing I'd appreciate it.
I wrote a gist for you
https://gist.github.com/kellydornhaus/45b8571807ec3ca659ea50...
https://gist.github.com/kellydornhaus/45b8571807ec3ca659ea50...
This looks excellent. Thank you for taking the time to do this. Have you released anything with Love2d yet? I'd love to support it.
No problem!
I made https://love-15.com which works on the web and as of today is on the iOS app store. I'll probably get it on Steam after that. I prefer playing the native builds personally.
It's fully free and about the smallest game design I could think of - I did it partly from a longstanding desire to make this specific game (I played a very similar thing decades ago on Palm Pilot) and partly as an experiment about just how much work can you get an LLM to do.
I made https://love-15.com which works on the web and as of today is on the iOS app store. I'll probably get it on Steam after that. I prefer playing the native builds personally.
It's fully free and about the smallest game design I could think of - I did it partly from a longstanding desire to make this specific game (I played a very similar thing decades ago on Palm Pilot) and partly as an experiment about just how much work can you get an LLM to do.
Should mention love.js isn't officially supported, and isn't 100% compatible with the LÖVE API. It will work just fine for a lot of games out of the box, though
Yeah when I got to the point of trying to get my game working in love.js, all I got were webassembly errors. I was able to get a basic Hello World game working with the same setup, so it's something within my code that's not supported, but I don't really know how to pin down what the problem is.
It took them many months to get it on Android. It was wildly successful initially, and then it took many months to reach phones and missed much of the hype cycle. So I think it took them a lot of effort to port.
> It took them many months to get it on Android.
It's surprising given that back in June? 2024 I just extracted the .love zip from the executable, patched a file or two, and merged it into the love android apk and it Just Worked. Just goes to prove the last 5% rule
It's surprising given that back in June? 2024 I just extracted the .love zip from the executable, patched a file or two, and merged it into the love android apk and it Just Worked. Just goes to prove the last 5% rule
The difficulty of porting to mobile is generally changing the UI to work well on a small touchscreen.
This. I'm making a game and I have an Android and iOS experimental branches. Even with a multiplatform framework (Monogame), compilation and basic run isn't a given. Then thing like small buttons needs biggee touch target, text input requires adaptation. All those 3 days + 1 day + a week + ... quickly adds up.
I've made games in Love2D for Android and for PC, I also as part of my converting Balatro to android, patched it a few places.
> Even with a multiplatform framework (Monogame), compilation and basic run isn't a given.
This isn't the case for Love2D. The game framework is already compiled for android, and the game is a bunch of Lua scripts. There is no issue with compiling, only very minor differences between the Lua for Android and the Lua for desktop platforms and making changes is instant except for transferring the apk to my phone and installing it to test (I didn't set up a full local android test environment for something so small). Balatro worked pretty much out of the box, except for some small changes I made — the first was to add in my profile from Steam, I didn't have to change the code for this. All the buttons were already reasonably sized. Text input was just adding a call to https://www.love2d.org/wiki/love.keyboard.setTextInput so that the on-screen keyboard appeared and disappeared when it was supposed to. I also patched a small bug that made it go blank on screen-reorientation, which was a one-line fix.
The total time spent was around an hour including codebase orientation, patching, testing and doc lookup (I haven't used the framework in a while).
> Even with a multiplatform framework (Monogame), compilation and basic run isn't a given.
This isn't the case for Love2D. The game framework is already compiled for android, and the game is a bunch of Lua scripts. There is no issue with compiling, only very minor differences between the Lua for Android and the Lua for desktop platforms and making changes is instant except for transferring the apk to my phone and installing it to test (I didn't set up a full local android test environment for something so small). Balatro worked pretty much out of the box, except for some small changes I made — the first was to add in my profile from Steam, I didn't have to change the code for this. All the buttons were already reasonably sized. Text input was just adding a call to https://www.love2d.org/wiki/love.keyboard.setTextInput so that the on-screen keyboard appeared and disappeared when it was supposed to. I also patched a small bug that made it go blank on screen-reorientation, which was a one-line fix.
The total time spent was around an hour including codebase orientation, patching, testing and doc lookup (I haven't used the framework in a while).
The desktop UI actually worked pretty much fine on mobile, to be honest. I ran it for a while using the porting methods available from the desktop binaries to an APK. The mobile UI basically just changes the interactions from a small button on the card to a drag-bases UI, which is slightly better on mobile.
The funny thing is that the original Desktop UI is still in the Android version. Sometimes I'll get jokers/consumables stuck in the desktop's "selected" mode, where the tiny sell/use buttons pop out underneath. So the whole drag-n-drop thing seems to have been added on top of the original ui, and sometimes the old ui events still fire.
Example: https://i.postimg.cc/P5VrVhQ4/Screenshot-20250101-103921.png
Example: https://i.postimg.cc/P5VrVhQ4/Screenshot-20250101-103921.png
Yep, I've noticed this as well.
Considering it's the first game to dethrone Minecraft on mobile it doesn't seem that it missed out on much.
To the OP, try DragonRuby for your next game if you miss hot reloading but like the scripting, code-first nature of LÖVE. It’s also cross-platform. Not sponsored, just a big fan and it scratches many of the same itches as you described with LÖVE.
Hmmm. DragonRuby seems to be a closed source proprietary game engine (at least as far as I can tell). With open source engines like Love2D / Godot / PhaserJS / Pyxel becoming more popular - this may be a deterrent for many devs on HN.
Something else to consider is traction, if you get stuck in DragonRuby there's probably not a huge amount of docs/tutorials around it. That being said, it does seem like they have a pretty active Discord at least.
Something else to consider is traction, if you get stuck in DragonRuby there's probably not a huge amount of docs/tutorials around it. That being said, it does seem like they have a pretty active Discord at least.
> DragonRuby seems to be a closed source proprietary game engine
Yes, but check out: https://docs.dragonruby.org/#/misc/faq?id=dragonruby-is-not-...
And
https://docs.dragonruby.org/#/misc/faq?id=what-if-i-build-so...
> If you get stuck in DragonRuby there’s probably not a huge amount of docs/tutorials around it.
You’d be surprised. The docs are extensive and there are sample games galore in the distro. https://docs.dragonruby.org/#/
Also the Discord is incredibly helpful and the creators chime in regularly.
It’s a pretty special engine and community IMO
Yes, but check out: https://docs.dragonruby.org/#/misc/faq?id=dragonruby-is-not-...
And
https://docs.dragonruby.org/#/misc/faq?id=what-if-i-build-so...
> If you get stuck in DragonRuby there’s probably not a huge amount of docs/tutorials around it.
You’d be surprised. The docs are extensive and there are sample games galore in the distro. https://docs.dragonruby.org/#/
Also the Discord is incredibly helpful and the creators chime in regularly.
It’s a pretty special engine and community IMO
I totally get it - I'm not definitely not casting aspersions on their choice to go commercial with it. I just feel like with the insane Unity pricing restructuring (which they ultimately rolled back) - a lot of game devs have been pretty gun shy about investing in another proprietary engine.
That said, we can always use more game engines / frameworks!
That said, we can always use more game engines / frameworks!
Love 2d can be quickly extended with hot reloading.
See https://github.com/rxi/lurker
That’s awesome! It’s been a core feature of DragonRuby since the start, and they share similar framework values from what I can tell. It’s worth a look. I’m not the biggest Lua fan, personally, so DragonRuby scratches the right itch.
I prototyped a game in a few frameworks/engines, namely: Godot, GMS, Defold, Love, and DragonRuby. I ended up sticking with DR because I found the api the most intuitive, and because I’m a career ruby dev and just can’t get myself to enjoy Lua. I find DR and Love to both be fantastic for those who just want to code.
Preach! Similar experience here. Agreed on Lua too. Love that some people dig Lua; to each their own, it’s just not for me.
I never was a big fan of Lua and that was something that kept me from looking closer at Löve, but a few months ago I discovered fennel (Lisp-like that transpiles to Lua) and that you can use that to write Löve games.
This looks like an up-to-date introduction:
https://itch.io/jam/love2d-jam-2024/topic/3484009/getting-st...
The "absolutely minimal Löve 2D fennel" repo was useful to me as I found the other templates a bit bloated (but possibly those are more useful for real-world projects beyond small experiments?):
https://sr.ht/~benthor/absolutely-minimal-love2d-fennel/
I did find that startup-time was bad for Fennel om some very low-end devices, or when running even a small game using LoveDOS in DOSBox, so I set up my test-projects to pre-compile the fennel-code to Lua as part of building the .love-files and that solved that problem.
This looks like an up-to-date introduction:
https://itch.io/jam/love2d-jam-2024/topic/3484009/getting-st...
The "absolutely minimal Löve 2D fennel" repo was useful to me as I found the other templates a bit bloated (but possibly those are more useful for real-world projects beyond small experiments?):
https://sr.ht/~benthor/absolutely-minimal-love2d-fennel/
I did find that startup-time was bad for Fennel om some very low-end devices, or when running even a small game using LoveDOS in DOSBox, so I set up my test-projects to pre-compile the fennel-code to Lua as part of building the .love-files and that solved that problem.
I'm a big fan of Love and Lua, but one thing that always bothered me was that there's no real "great" solutions for a) cross platform/web/mobile distribution. I know there are 3rd party solutions but for as mature as the framework is, you'd think it could get a first party solution. The same goes for several API's that you'd think are core to game making, like sprite animations. There are fairly mature libraries, but I always hate having a bunch of patchwork libraries that should be included batteries instead of 3rd party boilerplate.
Defold is not as well-known as Löve, but it seems to fit your description: https://defold.com/
> cross platform/web/mobile distribution
Unfortunately, each and every platform has its own unique quirks[^], and I think the problem is too large for a project like LÖVE to tackle.
In fact, LÖVE's authors have already done "90% of the work in 10% of the code": you can make a ZIP file, change the extension to .love, and just distribute that. Install LÖVE, double-click YourGame.love, and you're done. You can also concatenate the game executable with the ZIP file to create a self-contained binary distribution for that target platform (super old & cute trick; it's possible because ZIP's "header" is in fact, at the end of the archive).
[^]: In 2016 I've attempted a "one game per month" challenge; but I also wanted to distribute executables for Windows, macOS, and Linux, so that all of my friends could just play them. It went something like this:
- Windows: distributing executables from your personal domain is considered a crime; the browser and the OS will yell at the user, block the download, blacklist you, etc.
- macOS: simply concatenating won't work; you need an elaborate dance to create a .app bundle, and distribute that in a .dmg - good luck building that on a Linux host, also good luck debugging without an actual Mac. Nevermind code signing, you can bypass that with shift-right-click-open. Oh wait, macOS 15 won't let you.
- Linux: are you ready for the .so hell?
- BSD: bonus! The documentation for SDL_GetPlatform doesn't even enumerate all of the possible values! You need to source-dive to get them: https://github.com/libsdl-org/SDL/blob/f45ce45603a00e92332f9... - plus any possible patches in ports; most developers won't ever care, and will just hardcode "Linux". Checking for $DISPLAY is also unreliable: macOS users may have XQuartz installed, Wayland purists may run without XWayland, etc.
Unfortunately, each and every platform has its own unique quirks[^], and I think the problem is too large for a project like LÖVE to tackle.
In fact, LÖVE's authors have already done "90% of the work in 10% of the code": you can make a ZIP file, change the extension to .love, and just distribute that. Install LÖVE, double-click YourGame.love, and you're done. You can also concatenate the game executable with the ZIP file to create a self-contained binary distribution for that target platform (super old & cute trick; it's possible because ZIP's "header" is in fact, at the end of the archive).
[^]: In 2016 I've attempted a "one game per month" challenge; but I also wanted to distribute executables for Windows, macOS, and Linux, so that all of my friends could just play them. It went something like this:
- Windows: distributing executables from your personal domain is considered a crime; the browser and the OS will yell at the user, block the download, blacklist you, etc.
- macOS: simply concatenating won't work; you need an elaborate dance to create a .app bundle, and distribute that in a .dmg - good luck building that on a Linux host, also good luck debugging without an actual Mac. Nevermind code signing, you can bypass that with shift-right-click-open. Oh wait, macOS 15 won't let you.
- Linux: are you ready for the .so hell?
- BSD: bonus! The documentation for SDL_GetPlatform doesn't even enumerate all of the possible values! You need to source-dive to get them: https://github.com/libsdl-org/SDL/blob/f45ce45603a00e92332f9... - plus any possible patches in ports; most developers won't ever care, and will just hardcode "Linux". Checking for $DISPLAY is also unreliable: macOS users may have XQuartz installed, Wayland purists may run without XWayland, etc.
DragonRuby did it: https://dragonruby.org/
And many other frameworks did it too. LÖVE decided it's out of scope, and that's a 100% valid choice, considering the breadth and complexity of the problem. They still went quite far in ensuring it's no more difficult than it actually needs to be. Similar with e.g. cross-compiling Go programs (unfortunately, no cgo-less, cross-platform framework for interacting with windowing systems/video hardware exists).
Yes, but most other engines are much larger scope and fundamentally different purpose than LÖVE. DragonRuby, however, is in the same camp: code-first, scripting language (Ruby vs Lua), beginner-to-expert friendly. So my point was there are (at least one) project(s) very similar in purpose and size/scope to LÖVE that did solve for cross platform.
Balatro is available on Android though?
The article gives a sketch of three games the author built.
It sounds like the velocity was provided by consulting LLMs.
It’s fun.
I wonder if the creative feeling is an illusion, though. Does the fact that the LLM doesn’t feel like it is directly copying work it has ingested make the experience feel different than just ripping off a project you find on GitHub?
Anyway, I would be very interested in a write up that gives more details: how long did the author feel like they were “on the path” that LLM knows from existing code, vs how does it handle being “off the path”.
It sounds like the velocity was provided by consulting LLMs.
It’s fun.
I wonder if the creative feeling is an illusion, though. Does the fact that the LLM doesn’t feel like it is directly copying work it has ingested make the experience feel different than just ripping off a project you find on GitHub?
Anyway, I would be very interested in a write up that gives more details: how long did the author feel like they were “on the path” that LLM knows from existing code, vs how does it handle being “off the path”.
Not the author but you can see my other comment in this thread about similar experiences I had using an LLM + Love2D to build a full tiny game.
I love coding and didn't get any of that creative feeling with my code. But it sure was a magnified experience in terms of the creative feeling about the thing I was actually building. It was amazing to just think about features and that's it. Until it didn't work, I mean. Then it sucked.
I love coding and didn't get any of that creative feeling with my code. But it sure was a magnified experience in terms of the creative feeling about the thing I was actually building. It was amazing to just think about features and that's it. Until it didn't work, I mean. Then it sucked.
The current capability of LLMs feels perfect for bouncing high-level, "What are ways people commonly do X" or "How do I handle this architectural problem I'm having" off of.
I'm building a game in Monogame / C# right now, and LLMs have been great to help point me in the right direction or suggest architectural patterns I'm less familiar with to investigate.
e.g.
I'm building a game in Monogame / C# right now, and LLMs have been great to help point me in the right direction or suggest architectural patterns I'm less familiar with to investigate.
e.g.
If I have a GameStateManager, an InputManager, and a UIManager, and I represent input events as an enum of GameCommand's, what is a good way to register callback functions from these objects with the InputManager to trigger actions on particular game commands?
Is it common when using an Entity Component System framework to manage the UI using ECS? Or is the UI often managed separately using a more familiar object-oriented approach?
Have you ever had a creative feeling programming in a higher level language? Did you realize you were harnessing the power of the abstraction, “ripping it off”, without writing a line of assembly yourself?
That said, the LLMs are weak right now. They are great for getting some initial boilerplate (how do I draw a circle with HTML Canvas API?), but as the project grows you’ll be fixing them more and writing your own code. Give it a try, especially with Cursor. It’s awesome.
That said, the LLMs are weak right now. They are great for getting some initial boilerplate (how do I draw a circle with HTML Canvas API?), but as the project grows you’ll be fixing them more and writing your own code. Give it a try, especially with Cursor. It’s awesome.
I'm always amazed how people like author "learn" and "build' so quickly. I recently started to dabble with unity on basics and no way on earth I can build, say something like Heartstone clone or the interaction. My background is web back end mostly.
Good stuff OP.
Good stuff OP.
Unity is deceptive in that it is easy to scratch the surface, but gets more complex than it needs to be for simple game features quite quickly. Unity has an immense depth and breadth to its feature set, it is frankly, too much game engine to fuss with if you just want to try making games. In my opinion. There are just so many new, engine specific concepts at play that you are not learning games, you are learning Unity.
If you are a programmer and want to dable then a code first framework like LÖVE might be less cognitive load. If you prefer a more featured framework then Godot is also a really great lightweight engine, but you will spend more time in the GUI than in the code.
Don't sell yourself short either, you have an advantage by knowing some backend concepts! A lot of people start from zero programming experience and jump into games.
If you are a programmer and want to dable then a code first framework like LÖVE might be less cognitive load. If you prefer a more featured framework then Godot is also a really great lightweight engine, but you will spend more time in the GUI than in the code.
Don't sell yourself short either, you have an advantage by knowing some backend concepts! A lot of people start from zero programming experience and jump into games.
Yeah my experience with Unity was that, if you are doing something the Unity way, its very easy. If you want to do something that the Unity devs never considered, you are at best developing from scratch, at worst existing Unity systems are going to stand in your way.
I built a prototype Metal Fatigue tech demo (No real game logic as such, just the framework) and I found that heightmap based deforming terrain for the underground drill rig to dig through was so outside the design intent of unity that it would be the sole thing I spent hundreds of hours working on to get it past the tech demo stage. Like I got it working, but its so tenuous I could never ship it as is. Not to mention the default modules were writing my changes back to the heightmap object, saving them when that was unintended. I think the code I was using was intended for tooling, not for game logic.
And the Unity Store is great if you need a plugin to clone some mobile game system, but once you went to any depth it was just a hook to get 50 bucks out of you for no gain.
Its an interesting ecosystem.
I built a prototype Metal Fatigue tech demo (No real game logic as such, just the framework) and I found that heightmap based deforming terrain for the underground drill rig to dig through was so outside the design intent of unity that it would be the sole thing I spent hundreds of hours working on to get it past the tech demo stage. Like I got it working, but its so tenuous I could never ship it as is. Not to mention the default modules were writing my changes back to the heightmap object, saving them when that was unintended. I think the code I was using was intended for tooling, not for game logic.
And the Unity Store is great if you need a plugin to clone some mobile game system, but once you went to any depth it was just a hook to get 50 bucks out of you for no gain.
Its an interesting ecosystem.
I’ve found using frameworks like Love, or Pixijs more approachable when coming from a programming background. It’s more like learning the library rather than an entirely new approach to building software.
Coincidentally, I’ve been building a chess-inspired game prototype with LÖVE for a few days recently, but switched to Godot this week. Love and lua are lovely, but Godot makes so many things trivial, it’s hard to argue against the pace of development I was able to achieve.
I've made games in both. They both have their pros and cons but Godot is basically a full IDE while Loves2D is a framework.
I gave this LOVE framework a try a couple years ago, but was dissuaded by the amount of spritework needed for my game. But looking at your prototype games here, they look pretty good and playable as is.
I love LÖVE for prototyping, and am about to ship an extremely small game with it. Love.js works great to make a web build too; I host mine on the GitHub pages site for the repo. This is excellent for getting playtest help too even if you don't wanna ship it that way.
Over Thanksgiving break I decided to see how far I could get an LLM to do all the work (using cursor) and I picked LÖVE as a minimalist all-code framework. I put the game design and descriptions of UI and every screen in my .cursorrules file, so they are included in every prompt, and for a long while mostly told cursor in composer mode to do whatever the next thing it suggested was. I chose the smallest possible full game design I could and managed to just barely get there, although there were some brick walls where I had to help the LLM, but I got fairly far without writing or reading any code.
Reviewing what went wrong, the biggest problem is the LLM code started pretty good but just got worse and worse over time, especially as soon as it couldn't fit everything in its context.
ANYWAY, I was intending to suggest choosing the smallest possible game design you can if you wanna ship something.
I'm ready to ship my tiny project and it's amazing just how much work it is to ship a thing even after you've got all the functionality done.
Over Thanksgiving break I decided to see how far I could get an LLM to do all the work (using cursor) and I picked LÖVE as a minimalist all-code framework. I put the game design and descriptions of UI and every screen in my .cursorrules file, so they are included in every prompt, and for a long while mostly told cursor in composer mode to do whatever the next thing it suggested was. I chose the smallest possible full game design I could and managed to just barely get there, although there were some brick walls where I had to help the LLM, but I got fairly far without writing or reading any code.
Reviewing what went wrong, the biggest problem is the LLM code started pretty good but just got worse and worse over time, especially as soon as it couldn't fit everything in its context.
ANYWAY, I was intending to suggest choosing the smallest possible game design you can if you wanna ship something.
I'm ready to ship my tiny project and it's amazing just how much work it is to ship a thing even after you've got all the functionality done.
What do people think of Phaser? Currently playing around with it. API seems pretty large
I built Twins of Caduceus (a game where you have to control two snakes simultaneously) entirely using Phaser 3. I've had mostly very good experiences with it as a batteries included framework (renderer, asset management, physics, input, etc.)
It's a shame that the Phaser map/level editor is a perpetual subscription fee with no fallback license. That means you're relegated to either building your own map editor or using something like Tiled.
https://specularrealms.itch.io/the-twins-of-caduceus
https://www.mapeditor.org/
It's a shame that the Phaser map/level editor is a perpetual subscription fee with no fallback license. That means you're relegated to either building your own map editor or using something like Tiled.
https://specularrealms.itch.io/the-twins-of-caduceus
https://www.mapeditor.org/
I got most of the way through making a toy game with it. I kind of liked it, but I'm using Love2D for the game I'm working on now that I'm planning to release on Steam and hopefully mobile.
Although I did like that I could use Typescript for Phaser, and that it was a lot easier to show off the game to other people through the web.
Although I did like that I could use Typescript for Phaser, and that it was a lot easier to show off the game to other people through the web.
It's a full-blown game engine and you can do A LOT with it. I've created several js games with it and it never let me down. I've never tried using it with ts, but it should be even better.
edit: Just wanted to add that Phaser has one of the best docs out there including a library of tons of examples that can really get one going.
edit: Just wanted to add that Phaser has one of the best docs out there including a library of tons of examples that can really get one going.
It's definitely very much closer to a complete game engine than a library like LOVE and heavily leans on OOP.
Personally I prefer PIXI.js (which only focuses on sprites/graphics) as I can plug it more easily into my own data oriented architecture
Personally I prefer PIXI.js (which only focuses on sprites/graphics) as I can plug it more easily into my own data oriented architecture
Its mature and battle tested at this point. At least for 2D web games. The original/early version of Vampire Survivors is in Phaser if I'm not mistaken. Remade in Unity to make console port easier.
(Off topic) Which open-source engine/framework is best in terms of cross-platform? It seems that LÖVE is desktop-only.
Edit: by cross-platform I mean desktop + console + mobile.
Edit: by cross-platform I mean desktop + console + mobile.
Love2D can be cross-platform. Balatro was made in Love2D and was released on PS4, PS5, Xbox One, Xbox Series X, Switch, Desktop, iOS, and Android.
That doesn't mean it's well-supported or easy to fully support those other platforms. I'm struggling with getting iOS native functions working with Love2D in iOS, for example, although Balatro proved that it's definitely possible. The basic game (the game as it works on my PC) is working on my iPhone just fine, though, and wasn't too difficult to get up and running. I also have an APK working on a cheap Android phone I got.
That doesn't mean it's well-supported or easy to fully support those other platforms. I'm struggling with getting iOS native functions working with Love2D in iOS, for example, although Balatro proved that it's definitely possible. The basic game (the game as it works on my PC) is working on my iPhone just fine, though, and wasn't too difficult to get up and running. I also have an APK working on a cheap Android phone I got.
Like everything the answer is... it depends on what you're looking for.
What does cross-platform mean to you? Just MacOS/Windows/Linux? Web? Consoles? Android? iOS? Blackberry?
Godot comes pretty close to a batteries included cross-platform engine that handles 2D/3D.
JS engines can likely be ported (with varying degrees of effort as PWAs, electron/tauri, etc) to most systems as well - look into Phaser though its intended use is 2D.
What does cross-platform mean to you? Just MacOS/Windows/Linux? Web? Consoles? Android? iOS? Blackberry?
Godot comes pretty close to a batteries included cross-platform engine that handles 2D/3D.
JS engines can likely be ported (with varying degrees of effort as PWAs, electron/tauri, etc) to most systems as well - look into Phaser though its intended use is 2D.
Probably Monogame, a C# framework that supports Windows, MacOS, Linux, Android, iOS, iPadOS, Playstation 4, Playstation 5, Xbox One and Nintendo Switch
Don't some of them have proprietary APIs under NDA? how can an open source project support them? I ask because this is one of the criticisms towards Godot when compared to Unity.
Defold: https://defold.com/
If you are looking to do VR/3d games, LÖVR (https://lovr.org/) is derived from LÖVE.
Are there any 3d frameworks worth learning?
what an interesting article! outside of programming in my day I've written a few articles about guides on my blog https://playstationcouch.com/home.php
I have been using my engine https://github.com/willtobyte/carimbo
Demo https://carimbo.run/play/1.0.30/willtobyte/megarick/1.0.30/7...
You can use Lua as scripting language.
Demo https://carimbo.run/play/1.0.30/willtobyte/megarick/1.0.30/7...
You can use Lua as scripting language.
FYI the link at the top of your GitHub to play doesn't work. Seems maybe you changed the domain name or maybe the version is outdated compared to the link you pasted here.
play.carimbo.cloud vs carimbo.run
play.carimbo.cloud vs carimbo.run
https://www.reddit.com/r/PixelArt/comments/143ybxa/pixelated...
The OP mentions hot reloading and there are a couple of repos that may help with that:
https://github.com/clofresh/love-module
https://github.com/usysrc/LICK
Their card game prototype is really impressive for three hours' work!