Wind Waker static recompilation makes its way to PC

waker.jpg

Are you tired of waiting for Nintendo to release a definitive version of their GameCube classic on a console that's sold more than four units? Then step right up! A fork of BlueWake, a static recompilation previously targetting iOS, the creatively titled Wind Waker Recomp has just launched on Windows and Mac with a swath of new features.

Those wanting to jump in can expect to play with smooth motion, wide-screen support, high resolution rendering, and support for HD texture packs. On top of this, experimental support for 60Hz game logic is available, though not currently recommended in its current state.

For those interested, the below AI disclosure is included in the project readme file:

AI Disclosure said:
Wind Waker Recomp is developed with substantial AI assistance for code, testing,documentation and debugging. The status log records what has actually been checked, and on what.

As is common with these releases, no game data is provided in the project repo, with a legally obtained copy of the US GameCube game required to play. You can find the project's GitHub repo linked below.

:arrow: GitHub Repo
 
Didn't expect to read something this naive when I woke up today.

There's a massive difference between scripts written by hand to handle a bunch of tasks, a compiler that runs natively on your machine and translates from one language to another, and an LLM that's trained on a massive amount of stolen data and ruins the lives of many others by consuming resources that are local to them just to let people not think for themselves. Please do the basic minimum amount of research so you don't end up spouting such nonsense in the future.

As for the main post, was excited to see another recomp only for it to end up being AI carried. Will pass on this.

I am not defending the use of "Ai". I am well aware of all the current problems . I am only stating that it is just another tool, and if used correctly its a very powerful tool.

You are mixing different discussion topics.
 
I am not defending the use of "Ai". I am well aware of all the current problems . I am only stating that it is just another tool, and if used correctly its a very powerful tool.

You are mixing different discussion topics.
The problem is you're treating both tools as if they have the same issues the moment you compared using an LLM to a compiler or a script, when that clearly isn't the case. The consequences of using an LLM is vastly different than the consequences of using a script or a compiler.

Sure, an (online) LLM is a very powerful tool, but how it got there and the consequences of using said tool outweigh its benefits by a very long shot. If you have a button that gives you a million dollars every time you press it, but kills a random person, then that's a very powerful tool but the consequences of it isn't worth it.
 
You say this as if the majority of AI complains on this forum aren't coming from people with no experience in coding whatsoever, engaging in undeserved and uninformed elitism, whereas the majority of developers in 2026 use AI coding assistance in some form. The purist anti-AI stance is, regardless of the ethical and environmental arguments behind it, removed from the reality of the field.
You say all this under the assumption that at least half the people who use AI to do 60-70%+ of the work know what the hell they're doing on the coding end to begin with. The reality is that half (or more) of them don't even know the bare basics of coding (ex: they never took a single computer coding class in college, nor read a single page of "[C/C++/C#/Javascript/etc] for Dummies"), and just let AI and a couple kindhearted volunteers do basically all the work. I know of a Phantasy Star Online project that was doing that for a browser "remake", and got DMCA'd by Sega.
 
Last edited by ChronosNotashi,
The problem is you're treating both tools as if they have the same issues the moment you compared using an LLM to a compiler or a script, when that clearly isn't the case. The consequences of using an LLM is vastly different than the consequences of using a script or a compiler.

Sure, an (online) LLM is a very powerful tool, but how it got there and the consequences of using said tool outweigh its benefits by a very long shot. If you have a button that gives you a million dollars every time you press it, but kills a random person, then that's a very powerful tool but the consequences of it isn't worth it.

A well trained LLM model can, in theory, be a good tool.

I am not discussing those topics in this thread. I have already shared my opinion on those topics in other threads and i'm tired of repeating myself.
 
Rather just wait for the Windwaker decomp to be done and then polished up to see a true source port
It’s closing in on 80% complete with rapid but not eyebrow raising progress
Been checking in on it all year
Same

Been watching it stuck at like ~70% for mooonnnths
 
It doesn't matter in the slightest how capable the model is, or how good you are at making a prompt, if *you* do not understand what you're doing, *you* can not validate, verify, test, and improve the code being held in your face.

Sure, I didn't say anything to the contrary. And I didn't say anything specific about this exact project, which I haven't cecked. I'm commenting on GBAtemp's (and social media in general) way of engaging with literally anything AI, that's why I called it purist. It lacks the very distinction that you're making. Whether a project is fully vibe-coded, whether AI was used for documentation, or simple clean-up tasks, as soon as the letters AI appear all the discussion afterwards is about "AI slop".

Doesn't matter. Ports like these are also made by people with little to no coding skills. "removed from the reality of the field" actually it isn't. You just don't like that I have a different opinion than yours. One that isn't harmful to our world.

I didn't share an opinion about the project or whether I approve of AI or not. I said that the majority of developers in 2026 are using AI coding asstistance tools in one way or another. That's not an opinion, that's a simple fact established by numerous polls. And it's true regardless of how harmful it is to the environment. And the trends are continuing to go up. Avoiding all softwares, systems, codes, etc. that involved ANY use of AI assisted coding is effectively already impossible. The Linux kernel has already received multiple AI generated and/or assisted updates and the documentation explicitly allows the use of AI coding assistants. The same goes for Android and, given how heavily Microsoft promotes AI, probably Windows too. None of that is an opinion.
 
Not that quality code isn't welcome, it makes future improvements more feasible.
But a typical player during the course of active gameplay doesn't need to know how a game was coded.
You're talking as if the subject of this thread doesn't run like absolute garbage.
"A player doesn't need to know how it was coded" but they can certainly tell, now, can't they?
 
Definitely. I wish there was more quality control.
Quality control is the individual. As long as the use of AI is disclosed, everybody can make their own determination on whether that's a deal-breaker or not.

Personally I don't mind it in instances where the goal is creating a 100% accurate port of a game with a few extra bells and whistles. It's when people use AI to try to make "improvements" on the original gameplay/graphics that things start to get really messy and divisive.

And hey, free is free. Anybody with enough money to pay for a subscription to one of the latest frontier models could instead be using them solely to make themselves richer. So while it is a low bar, at least providing something to the public at large is a much more interesting and worthwhile use of the technology.
 
  • Like
Reactions: Sir Tortoise
Quality control is the individual. As long as the use of AI is disclosed, everybody can make their own determination on whether that's a deal-breaker or not.

Personally I don't mind it in instances where the goal is creating a 100% accurate port of a game with a few extra bells and whistles. It's when people use AI to try to make "improvements" on the original gameplay/graphics that things start to get really messy and divisive.

And hey, free is free. Anybody with enough money to pay for a subscription to one of the latest frontier models could instead be using them solely to make themselves richer. So while it is a low bar, at least providing something to the public at large is a much more interesting and worthwhile use of the technology.

The issue is the scope.

You want to use AI to make a snake game, go for it. You want to use it to reverse engineer an entire software team's code and call it a day, you've got problems. So now it seems like you're pivoting to "at least they're spending the money towards FOSS", while the code is watermarked, unmaintainable, ruins the environment, and provides a substandard product that does functionally nothing an emulator can't do, not even counting that the majority of these projects don't even have a proper licensing format, further preventing actual dev use.

You have a car. It's made of a million other different cars. There is no manual, and the car drives itself 90% of the time. It looks like a Ferrari. It goes slower than my walking speed. Convince me that car is somehow a better replacement for my legs.
 
I thought wind waker was already decompiled , is blue wake not then, sorry for not knowing much about these things.
Post automatically merged:

Sorry in advance, I don't have much of an understanding of these kind of things, I thought wind waker was already decompiled, is blue wave not a decompilation? Just installed it today
 
I thought wind waker was already decompiled , is blue wake not then, sorry for not knowing much about these things.
Post automatically merged:

Sorry in advance, I don't have much of an understanding of these kind of things, I thought wind waker was already decompiled, is blue wave not a decompilation? Just installed it today
Bluewake was a static recomp targeting iOS. This thread is about a recomp for PC and others, but the two projects seem to be merged now.
 
I'll stick to the Wii U version unless the improved graphics, fast sail and other QoL are ported to this version. That's still the definitive version.
There is a romhack for that! It's Better Wind Waker, which is basically the randomiser hack without randomisation. Don't count on the recomp to implement it, it's a recomp after all.

Obviously the improved graphics are nowhere to be seen, but I don't think either release looks worse than the other. They're both timeless.
 
It looks like a lot of complaints come from lack of info. Using AI to fill in tedious parts of source code is a must.

This isn't generating boilerplate for a hello world script, it's reverse engineering standing software with virtually no exposed interface.

The typical tree is usually:

1. Determine how the file itself is loaded. Headers, magic, partitions, whatever. Pretty simple, and something AI could realistically speed up.

2. Converting the static, possibly packed bytecode (yes I'm aware bytecode is a different thing, ease of explanation) into a readable format for parsing into a higher level language, think

`
B8 01 00 00 00
BF 01 00 00 00
48 8D 35 10 00 00 00
BA 0E 00 00 00
0F 05
B8 3C 00 00 00
31 FF
0F 05

48 65 6C 6C 6F 2C 20 77 6F 72 6C 64 21 0A
`

To

`
mov eax, 1 ; syscall: write
mov edi, 1 ; fd: stdout
lea rsi, [rip+16] ; address of our string
mov edx, 14 ; length
syscall

mov eax, 60 ; syscall: exit
xor edi, edi ; status = 0
syscall

db "Hello, world!", 10
`

One of these is significantly easier to read than the other.

3. Disassmebling, analyzing, and reconstructing (or directly) to a higher level language like C++ (or whatever the original was written in) for reuse.

4. Then you have to sort the various syscalls that are specifically for the origjnal platform's ABI, and either convert those into a format the host toolchain can accept, or make it a generic and allow the compiler to fix it per platform. This is the hard part. Anyone with enough time can understand how assembly works. You can even write scripts to help convert signatures across the entire thing, but it still needs direct human intervention usually.

The objective of this kind of thing is a 1:1 recreation of the original source code in the absolute sense, but that's extremely difficult to achieve, so we settle for "close enough". It's not like the source is just sitting there packed away, it's already packed and compiled and ready to go. We have to abstract the *how* from the *what*, and that's not something AI does well, especially when a lot of specialized software (games) depends on specific functionality from the host hardware (GameCube).
 
So now it seems like you're pivoting to "at least they're spending the money towards FOSS", while the code is watermarked, unmaintainable, ruins the environment, and provides a substandard product that does functionally nothing an emulator can't do, not even counting that the majority of these projects don't even have a proper licensing format, further preventing actual dev use.
I didn't "pivot" to anything, this is the first time I'm voicing my thoughts on the matter. Like I said, the individual and the community can easily differentiate what is accurate/performant from what falls well short of available emulation. Not all LLMs are gonna fall short of that standard, and not all people using LLMs to aid in those projects are gonna be without coding experience of their own.

Now, in the grander scheme of things, do I think AI is gonna be a net positive for humanity? Hell no, it's a vampire draining our economy and vital resources at a time when inequality and human rights abuses were already piling up. That said, even if every hobbyist in the world stopped using it, big tech's sunk cost fallacy isn't suddenly gonna disappear.
 
I think you mean make code that when compiled makes a 1:1 recreation of the original target. the original nintendo source would be very different :)
 
Last edited by PhyChris,

Site & Scene News

New Hot Discussed User Submitted