@Martin_H There is the matter of whether the editor think that an opinion piece like this, which is bound to collect a lot of opinions, should be put on a site with more traffic. Or maybe the author writes exclusively for this site.
Either way, it isn't a big deal. These Hookshot sites cross-post a lot.
@axelhander Ah, SaviorS. I followed the Twitter of the developer of that game ( https://x.com/SaviorS_OOTB ) for a while, so I can say a few things:
The dev said that he used AI as part of his toolset, but whenever people asked him at which capacity does he use AI, he never answer.
So people eventually concluded that the dev used AI for making the sprites and CG. at which point the dev came out and admitted it. He said that he wanted the AI-generated assets to only be placeholders, and he hope he would be able to replace them with man-made arts at the end of the development, but it all depended on how financially viable would it be.
The AI-generated sprite animations don't look good. They look misaligned, choppy and have low frame count. But there is a point to be make about how the type of game SaviorS is lessen the impact of bad animation. SaviorS advertises itself as a beat 'em up & RPG hybrid, but it's more like one of those side-scrolling MMO from the 2000s. In actual side-scrolling action RPG like the Tales games, the sprite artists have to take into account different things like skill cooldown, hitstun, etc. when drawing and timing the animation. But when the level of interaction between the playable character and the enemy is limited (no positive/negative frame, mostly about spamming skills, etc.) then you have a lot less to consider.
When it comes to other aspects, the dev seems genuine in his effort. When I still followed him, he was concentrating on adjusting the squad NPC's behavior.
So yeah, not the best thing ever, but still a far cry from the kind of lazy AI slop that inundate Steam right now. The dev actually treats the AI as just a tool and tries to design their game around the limitation of the tool. A rare example, I say.
@Martin_H This site posted plenty about new games with retro sensibilities or new games from old developers way before consumer AI proliferated, so talking about new games on Steam isn't out of their ballpark.
@axelhander WTF are you talking about? Even before ChatGPT was a thing, there were plenty of free and open source software, royalty-free arts & sound assets, etc. alongside a wealth of documents to teach you how to make a game with them. The barrier of entry was so low, especially for making PC games. Saying that before AI, game making was the domain of the rich is just ahistorical.
Bad games have always existed, and people call out the terrible ones made by humans too. But do you know what makes an interesting bad game versus a boring one? Either the game is bad in a singular, unthinkable way or its flaws embody the neuroses of its creator. You can learn something new from an interesting bad game. Being boringly bad like vibe-coded slop just means it fails at the fundamentals. Don't expect people to show any mercy to boringly bad games.
It's weird for AI supporters to act like they're surprised when other people show disdain for AI. Sure, AI has done great things, but when people see their social media or game storefront being inundated with lazy AI-made *****, it's normal for them to feel negatively toward AI.
But I am willing to hear your case. Just show me a game made with AI from this Steam Next Fest that you think is a gem.
@axelhander Only a fool would think an LLM model is some "collective intellect" of humanity. Especially given actual cases when the clones created by AI ended up way worse than the original:
Rip-off games have always existed, and people call out the terrible ones made by humans too. But do you know what makes an interesting bad game versus a boring one? Either the game is bad in a singular, unthinkable way or its flaws embody the neuroses of its creator. You can learn something new from an interesting bad game. Being boringly bad, like the AI clones, just means it fails at the fundamentals. Don't expect people to show any mercy to boringly bad games.
Wake me up when AGI is actually a thing and you can make the case that AI is greater than the sum of its parts.
@Thad I get you. That is what artists have been saying for years, that a lot of people who aren't knowledgeable about arts will look at the flood of AI arts and start questioning every piece of art they see on Twitter if it's AI or not. It just dampens their enthusiasm. This phenomenon has already happened.
The funny thing is that AI bros will always say "I tag my AI arts properly." to cover their asses. Then they found out that most people absolutely doesn't want to engage with AI arts. So they're now whining that tagging AI arts is discriminatory. They shows their true intention of wanting to replace hardworking artists with their lazy slop.
Back to the topic. What I was trying to get to with the original post is that there is a community component to it. Call me old, but if you were in the forums scene back in the day, you just know it. Some fan who does a proper localization isn't doing it out of vanity and their ego like the AI bros. It's hard work for no money and little fame. Sometimes the only thing powering the translator are a few friends he made on the forum.
We can talk about the low standard of people who say "I am fine with terrible vibe-coded translation and recomps." But to me they look less like people who participates in the community and more like some atomized hoarder who only looks to add more and more to their hoard.
@Thad Thanks for the heads up. And the first PSP Suikoden translation was a Google Translate mess. The proper human-made one came out like one year later.
I could be wrong here, but I remember there was a small debate at the time among PSP fans because of the first Woven Web translation and another debacle related to Sword Art Online Infinity Moment — Some guy grabbed the English localization of Sword Art Online Hollow Fragment (which is the remake of Infinity Moment), backported it to Infinity Moment, and then paywalled the whole thing.
There are some localization and decompilation projects that use AI appropriately as the tool it is, like bug testing. However, these projects are few and far between, and each of them is being drown under dozens of lazy vibe-coded "projects" by people who don't know the first thing about programming or retro hardware architecture.
I've seen the same thing happened to the translation community when machine translation became more intuitive and wide spread. Suddenly there were dozens of lazy, sloppy translation of weekly manga & anime. The BS drown out the actually good translation by groups who took their time to do it and therefore were unable to catch up to the treadmill of weekly manga & anime. To say that the slop divided the community was to put it lightly
But it doesn't have to be like that with game localization and recompilation. First of all, we have all the time in the world. Retro gamers are not some overhyped Weekly Shonen Jump fanboys who absolutely need to know what happened in the newest Naruto or Bleach chapter. The debate about speed vs quality is not applicable here.
Secondly, people who are going out of their way to play a fan localization or a game recompilation is already deep in the bush. They know what they want, and they will scrutinize every aspect of what you do. If your audience is already like that, then why even bother throwing together a buggy vibe-coded bowl of slop that you've no intention to fix?
For all the people who say "the bad ones will sort themselves out". It's easy for you to say so as a solitary consumer of the end product. But we're talking about fans producing things, which means the community aspect is unavoidable. You might think "there is enough room for everyone" and the human translators and coders will just go back to do their job properly. Love it or hate it, those fans are doing the job for more than themselves. They need the community by their side, be it for constructive feedback or just cheering. And there is a very real frustration when you see some donkey-asses MF taking your hard work, order the AI to slap some codes onto it and then tell everyone this is the "complete version".
@Beyerun That is the thing. Emulator is good enough for normal people. Only those who are deep into the bush will deliberately seek out recomps, like people who just have to play with mods.
So why do people even bother making vibe-coded recomps in the first place if they aren't going to fix the obvious bugs? The type of gamers who play recomps know what they want, and they're going to scrutinize every aspect of the game. It all seems like a waste of time to me.
@RupeeClock VideoGame Esoterica has been sketchy even before decomps and recomps became popular. I don't trust anyone who put out several YouTube videos each day to be able to do their research properly, especially in a field like retro gaming. The guy is spinning a content mill to make money.
This argument that "There is room for everybody." has happened before with manga scanlation. The moment machine translation became more intuitive and widely available in the 2000s, I started seeing more low-effort manga translation. They sucked and were a waste of time. Nevertheless, there were scanlation groups that took time to make a good translation, only to be overshadowed by some machine-made slop because in the content mill of weekly manga, so many people only care about the "plot" and not the literature itself.
It sucked back then to see some people having wrong perception of a manga because they only consumed the machine-translated slop. I'd rather it not happen to retro games. Though like I said at the beginning of this post, you've to be in the know to deliberately seek out recomps. If you play a vibe-coded recomp and have a bad time, it's kind of your fault.
@ElkinFencer10 The PSP came out in the early 2000s. Back then flash storage was still expensive and Sony wanted 3D games with PS2-ish graphics, so PSP games were much larger than DS games. UMD was the only way to get those games on a medium cheap to produce and thus could maintain the publishers' profit margin.
Not to mention that the UMD grew out of the MiniDisc that Sony created, which meant they already had the technology and experience working with it.
Ditto with the plastic screen. Again, it was cheap enough and worked well enough for the 2000s.
Anyway, Shawn Layden is right about Sony getting back to the handheld business. You need an entirely new system set up to support the thing. Just look at the how Nintendo had to focus on the 3DS to the Wii U's detriment. The Switch, a handheld-console home hybrid, had to happen because Nintendo wanted to have only one workforce and one budget to maintain their ecosystem.
If the rumor about Sony cooking up a new handheld is true, I hope that they know which portion of the market will it occupy. Many first buyers will be people who already have a PC or a Switch as their main gaming system. As such that handheld needs to deliver really good handheld experience instead of being some kind of portable PC like the Steam Deck or the Legion Go.
@Martin_H Shrug Not my fault that many people have been doing AI stuffs so weird and off-putting that they invite the categorization. Flooding hobbyists' space with ill-informed and comically bad fake news and fake documentaries. Stealing artists' artworks by running them through AI and then claiming credit. Circumventing websites' anti-scraping measures that are set up by owners who explicitly don't want to be scraped. Etc.
If these people don't want others to name them names, they should take a look at the mirror and think hard about how unpleasant their actions have been for others to put them in the "to avoid" basket.
@Martin_H Yeah, and the AI companies are buying off every local and state officials they can. I know a few cases in Louisiana and Oregon where the locals had already voted down the motion to build data centers, only for the councilmen to overturn the decision in closed meetings. It's not even "bad city planning" anymore. Anyone can see how bad the idea is from miles away. This is pure uncut corruption, and the lion's share of the blame still falls to the AI companies. Not that the politicians are any cleaner.
You would think that with the billions they have received from investors, the corporations could have had done something like setting up a solar farm. Nope. It's all about taking advantage of the existing grid and demanding the state to keep giving out those tax breaks.
"Witch-hunt mentality". Now you are just over-exaggerating the situation. Some people here have very strong opinion about AI and like to ridicule the incompetent individuals who used it. I don't think I've seen anyone who had threaten to take any harmful action. Conversely too many AI bros like to claim victimhood's high ground just because people ridicule their bad AI creations and the way they irresponsibly used the models.
@Martin_H And just because the people making the point about theft may have participated in piracy themselves doesn't mean their argument lacks ground. Germans are still being fined for pirating and sharing movies through torrents. Theft is still illegal in the eye of the law, just like secretly collecting data from your customers and feed it to a LLM. The reason cops don't stick it to individual pirates is because they'd rather spend their limited resources to go after the big fishes like illegal streaming sites, etc.
Until the authors of these vibe-coded translations and recompilations can prove that they only used locally-run models trained solely on data that have been allowed by their owners, then people have the right to be suspicious and critical of these projects from the environmental and intellectual rights angle.
@Martin_H If you're talking about the tests to determine the capability of the the AI models, then they have been and are still being done by academic institutions and AI companies. You can just search for them and read them either for free or pay for a research hub paywall.
However, there is a discrepancy between what the corporations admit to themselves and what they're telling consumers and lawmakers. Way before the news reported the first case of suicide encouraged by AI, Facebook's internal report had already shown that their AI could induce psychosis. Engineers and safety activists have been talking for years about putting up guardrails, yet corporations like Facebook were so scared of their models bootlicking their users one degree less that none of the safety measures was implemented.
Instead of talking about actual problems like this, what are the AI corporations' biggest talking point nowadays? It's Skynet, again. It's an attempt to distract from the actual consumers' rights. And don't get me started on Silicon Valley trying to use the legislation as a cudgel to choke out the "dangerous" open-source models from China, which is a blatant attempt of unfair competition.
And I've already talked about locally hosting AI server in my previous post. Again, I've to repeat that many people using AI to do the low-effort fan translations and recompilations are not using local models. They talked about how many tokens they had burnt, and in the same breadth talked about how to evade the rate limit by setting up several accounts. These ain't your Ollama users with their MacMini. This is just how low-rent these people's operations are. And if you talk to them about building their own rack, they're just going to whine that "but I don't have the money."
Your point about city planning: The Chinese have been building several power plants in the Western half of their country to supply the data centers nearby. The thing is they built these complexes in sparsely-populated regions, and the companies still have to adhere several regulations. America ain't building new power plant, data centers are plugging into the existing power and water grid, and many are operating while consuming way more energy and water than the agreed amount. All the while enjoying enormous tax breaks paid by tax payers. I know the Chinese aren't subsidizing their AI companies this much.
With that in mind, shouldn't the burden of responsibility falls onto Silicon Valley instead of whichever local official you blame for "bad city planning"?
@Martin_H Did you actually read my posts? Nowhere did I say that there should be some kind of test for people who want to use AI. What I said is AI naturally attract a lot of numbnuts who don't know the first thing about what they're doing the first place.
But I can't effectively stop someone from using AI if they really want to. This is the crux of my argument about consumer protection. The industry has an obligation to drill into heads of the dumbest people about what AI can and can't do, what is the impact to user personally or to the environment, etc. If the industry want to withhold the information then the state will step in.
And @81632bit made a great point that adds to mine. The consumers have the right to know if the model they're using have been taught on ethically-sourced data, is the data centers in use not endangering the local community, etc. They need to know these things before wasting valuable resources on whatever BS they want to use AI for. As I see it, the way your average Joe is using AI produces little thing of use while incurring proportionally high cost on the resources currently available. And they aren't even remotely aware of it.
If these people are hosting their own local AI server and footing the electric and water bill out of their wallets, then I wouldn't have cared. But they aren't doing it. The cost of operating the AI business is so heavily subsidized by the state and financial institutions that it creates the perfect circumstance for enormous waste to happen right now. At a time when everyone are talking about a climate collapse and an economic recession. Normal people only see how much their tokens cost and think AI cost proportionally to run. It isn't.
@Prizm Of course the guy know about the reality of Genesis piracy and bootleg cartridges. He is doing a Genesis port in the year 2026. Anyone doing such a thing is surely relying on emulation and the Everdrive because, surprise, they ain't gonna have an official Sega Genesis dev kit laying around.
Monster Bath most likely just want to make some bucks. If he think he can do it in a legal-ish way, then I say let him try. I'm not very optimistic that he'll be able to convince Bandai-Namco, and even if he can, the ROM is going to be dumped and the carts are going to show up on AliExpress within weeks anyway. By that reality, might as well try to do a legal release and get the money from the few people who, for some reason, absolutely need Bandai-Namco's seal of approval.
@SlangWon I need some elaboration about your idea of "for the fans" here? Monster Bath Games seems to be pretty passionate about this project. Are you perhaps thinking that for it truly "for the fans", the complete ROM should be released for free for everyone?
(If we're talking people who really, really want to play the game but can't afford the arcade board, then why are they even bother with this port? The superior arcade original is easily emulated. Or play the Namco Museum/Arcade Archive version if they don't want to deal with emulation.)
Like you said, the ROM is going to be dumped and the bootleg carts are going to show up on AliExpress within weeks either way. And people who buy bootleg carts or put the ROM on their Everdrive isn't going to buy the official release anyway. Do a legal release and the guy can probably make a few bucks from people who absolutely need Bandai-Namco's seal of approval.
And, like, it's a known practice for porting houses to take the assets from an existing game without permission to make a demo, then take it to the rights holder in hope of getting a contract to port the game. The only problem I see here is Monster Bath made the demo public instead of taking it to Namco-Bandai first.
@ElkinFencer10 As for how likely Namco-Bandai is going to agree to this, it all comes down to who know who and whether think company think they can sign a deal with whoever the actual publisher is for a cut of the profit.
This is Namco-Bandai we're talking about here. Games like Xenosaga and Soul Calibur have ten times the number of fans of something like Splatterhouse, yet Namco-Bandai have given them less than crumbs despite these fanbases yelling at the top of their lungs for years. "Fan pressure" isn't going to budge them.
@Martin_H Do you really think that the air force officers who signed up to be astronauts or the congressmen who argued for NASA's budget didn't read the mountain of scientific papers to convince them that reaching the moon was both viable and desirable, and the whole thing was an worthy ordeal?
Belief doesn't come out of a vacuum you know. And you do need to convince other people. We're talking about collective efforts that require the marshaling of so much manpower, money, land, materials, etc. You know, like building rockets to the moon or developing AI? You can't just go all Voluntarism here.
Like, haven't you seen how rural Americans are reacting to new data centers? They don't believe the data centers won't suck up all the electricity and groundwater, leaving them high and dry. They don't believe the data centers won't cause constant headache because of their noise. They don't believe the data centers will generate more jobs for the locals or put more money into the local economy. They don't believe that their lives are going to get materially better if AI gets better.
They believe that their lives and their communities can only get worse when a data center is built on their area. So they're trying to block any data center project they can.
@Martin_H Sure, he is free to do everything you said. People are also free to judge this person based on the creation he put out as well as the tool he used to do it. And it's not a good look.
It's like if you care about AI and how AI is being received by the public, you'd care about how companies are just slapping the "AI-powered" label onto every product. It just cheapen the word and distracts people from what AI is actually useful for.
I've to correct you about the moon exploration analogy though. Scientists had theorized about reaching the moon since the 18th century. Jules Verne wrote From Earth to the Moon in 1865 based on actual research papers. Scientists did rigorous scientific experiments half a century before the Apollo program was even a thing. So, there is a very long history of theorizing, arguing and producing evidences to convince people about the viability of landing on the moon. It's like the anecdote about Columbus proving to the queen of Spain that Earth is round. The Ancient Greeks had already experimented to prove the Earth is round AND calculate its circumference.
@Martin_H Like I said, AI is being sold to people who have no clear idea what are the capability, limit, pays and gains. That include people who wrongly think that AI on itself can carry them to a good recomp of Knuckles' Chaotix while they're so blind to the very visible flaws in the creation they put out.
Seriously, if the bugs are that visible and severe then why would someone who, let's us assume is a competent and self-respecting programmer, even bother to publicize a demo build? Just focus on fixing the bugs and maybe put out a YouTube video if you really want people to know what you're doing. It makes people question just how much does this individual know what they're doing with AI and how much they care about it.
@leogames If it suits you to assume that I'm so close-minded and doesn't care enough about your opinion to change mine — while I've shown no sign of such — then sure, be my guest. Just get off your high horse the next time you argue with someone else.
@Martin_H AI is being sold to and used by people who have no clear idea what is its capability, limitation, pays and gains. As I see it, it's a matter of consumer protection, and the AI industry is muddying the water with all their talks about "Skynet" instead of talking about the actual problems happening right now, from the spyware embedded in the models to AI pushing the psychosis of vulnerable people to the point they committed self-harm. And yes, it includes people who wrongly think that AI can carry them to a good localization or a good recompilation while they have none of the skill & knowledge to detect the bugs in the first place.
There is also the matter how much humans are using AI as a form of plausible deniability. You see CEO saying that their products are ***** or someone got wrongly laid-off because "AI did it." You see generals saying that their airplanes bombed some school or hospital because "AI did it." Psychotic influencers used to say "The algorithm pushed me to do it." but nowadays you've got people with actual power putting up the same BS with AI. Both the AI corporations and their customers are saying that they can't be forced to take responsibility because all the decision are made inside the black box that is AI and somehow we all just have to put up with the BS.
And don't get me start about the actual evil stuffs like the IDF's Lavender.
We've tools to deal with consumer protection and negligence leading to harmful consequences. The laws, the education system, the civic society. Ethics in AI construction and usage has been an academic field since the last century. Society and the state can definitely bring this matter to the table right now and work out a legal, moral, economic and cultural consensus about what are the acceptable use cases for AI, what are the penalties and punishment for misusing AI, etc.
@leogames You presented your argument that AI-coded software shouldn't be slop when it's well tested and fixed. I presented my argument on why AI-coded software hasn't been well tested and fixed in practice.
To counter my argument, you simply said "things will get better" without actually saying how would it get better. Which forced me to go back to your original post, which I quote here:
"Whether we like it or not, it is actually pretty incredible that AI can write what it has here. Sure there are bugs - but it's very clear that the majority of the game is there and that AI has come a long way in just a very, very short space of time. Those bugs, alongside more thorough testing, won't be there for long."
Am I wrong to read this passage and assume that your position is that things will resolved themselves as AI gets better? It's the "...alongside more thorough testing..." that I question (if it will happen) and you haven't responded to.
@leogames And it's precisely this attitude that things won't get better. AI on itself isn't bad, but the way people is using it is terrible — from creating slop to generating kill lists (like what Israel is doing).
And if the fault is with us humans, then it can be changed, it can be fixed. But like I said, it stand in the way of the profit that goes into Silicon Valley. They have an incentive to keep things as bad as they are now.
You people are naive to think that just because the tool will get better mean everything will be solved when the faults with the user and the usage are still there. That is even presuming that the tool will get better soon enough when LLM has been around for a long time, scientist think that they're a dead end when it comes to AGI, and recent financial reports show that AI companies are increasingly relied on circular finance and doggy bookkeeping to hide the fact that the gain their new products made are marginal.
@leogames @Martin_H You just pointed out the Achilles' heel of vibe-coded recompilation. Emulator is good enough in the eyes of a casual gamer. One has to be pretty deep into the bush to have the kind of requirements you listed. But that kind of person is also really strict with what they want and will dig everything up to find faults.
As such, the maker of the recompilation has to be just as strict to fully satisfy their intended audience. Do you think that someone who used AI as a shortcut for their laziness and lack of skill have the will, knowledge and diligence to do such a thing in the first place? Especially when they're doing it for free and without the pressure of the boss, customer and contract to make sure the product is good?
Since this site started talking more about AI, I've been making this point again and again: There are more INCOMPETENT and LAZY people using GenAI than COMPETENT and DILIGENT people.
It isn't just game localization and recompilation. Just talk to people who are into Heart of Iron, Skyrim, etc. modding and they'll happily tell you about how their communities are being flooded with low-quality vibe-coded mods made by individuals who don't know the first thing about coding. Back in the old days these people would've had been forced to either learn on the job or just quit and don't make it. Nowadays AI allows them to just throw ***** out and stink the place.
It's an entirely predictable outcome given the way the AI industry is marketing AI to people. They're selling AI as a crutch for people with no skill and no motivation to work and learn. They're selling AI as the magic wand to turn people from being a prole into a petite bourgeois. It isn't that different from how crypto, blockchain and DeFi were sold to us years ago.
(I can go into how the AI boom is the direct consequence of Silicon Valley's failed gamble with crypto and Web3/Meta. They desperately need another growth engine to justify the unimaginable amount of investment they're receiving. But it'd be too long for this site.)
People had used AI in a scientific and business capacity long before ChatGPT was even a thing. The differences were that they used AI in what it was actually good at - data processing, pattern-recognizing and prediction — and there were always human overseers in the loop. The AI industry's hype machine desperately doesn't want your average Joe to know it. They stands to benefit the less you know about their products and the more you depend on whatever crumb they gives you.
All these talks about "AI is getting better" won't matter if we don't impose strict limits and educate everyone about when, where and how they should use AI. As long as we let the AI charlatans and AI snake-oil salesmen run the show, there will only be more disasters created by stupid people who think AI can compensate for their stupidity.
As for the recompilation matter, I'll add this point: There is a reason why porting games is hard and studios like AM2 and Iron Galaxy make a business out of it. Like localization, you just have to account for so many different things, from performance to audio & visual, when you bring a game from one platform to another. It's even more so when you're recompiling — essentially natively porting — old games that were made to be so intertwined and dependent on their original hardware.
@Aloe Ah, you talked about the "average programmer" and "great programmer". I ain't talking about them. My point since the beginning of this thread is that when it comes to fan translation, how much is AI being used by competent people who want to speed things up or widen the scope of what they can do, and how much is it being used by lazy, incompetent people with no skill and is unwilling to learn?
Scientists had used artificial intelligence in fields like cryptography for years before ChatGPT was even a thing. No argument from me about that part. But we are talking about the Dunning-Kruger effect here. The way the AI industry is selling AI to normal people, it's creating so many cases of delusion. So many lazy, low-skill, less-informed people offloads their thinking to AI, and the sycophantic way the current models are programmed to respond to their users makes it even less likely for them to see their errors and learn anything. The rise of "vibe coding" LARPers is one example.
Many commenters in this thread seem to be under the impression that there is a substantial number of "competent" and "motivated" people who just need a little wind under their feet to be finally able to vault over the fence and localize the games they love. They forget that retro gaming is still a niche hobby, and there aren't many people who have a solid foundation of how old games and old consoles work — or the diligence to do the job. There a reason why the Sega Saturn port of Symphony of the Night, a game well-known in the retro community with little to translate outside of what the PS1 version had already covered, still waited years before receiving a translation patch.
You would think that anyone who want to dabble in translating retro games is required to learn, like, how retro consoles manage their memories or the length limit of the text box. But some people (at least the ones I observed) seems to think they can skip all that "unsexy" learning and go straight into the job thank to AI. And when you nudge them about it, they act like they are being insulted to suffer the "indignity" of wasting their time learning an unmarketable skill. The same people back in the day would've had been forced to learn on the job or just quit before they become an annoyance.
@Aloe I've seen more than a few translation efforts that broke down because of unsolvable technical hurdles. The thing is that the people in those cases had enough of a standard to not put out a patch and, if possible, made themselves available to others who could and wanted to do it properly. They didn't use AI to cover their asses and put out a patch that they didn't even understand what was wrong with it in the first place.
And the example with the Air Management game is even more insulting when you consider that the guy who made the AI "addendum" didn't consult the original creator in the first place. People who've made addendums for others' fan translations did that. It's just common sense. The fact that the AI guy couldn't even bother to do it tells me that he didn't really put much thought into what he did.
You can go to Hacker News and the people there will tell you about how AI is only good when you know what the solution looks like in the first place. If you're in the middle of building the pipeline and are unsure about what comes out of the process, using AI to solve everything is just going to make you and others even more unsure about the final product's integrity. So if you actually care about what you do, AI forces you to spend more time trying to understand it and fix what it did wrong. It cost more time than just do things by yourself bit by bit and fix the errors along the way.
I don't want to use the word "gatekeep", but at some point you just have to think that the technical difficulties do have an effect on separating the people who actually have what it takes to localize a game to those who don't. At least have the people skill to find someone who can help you. If you jump right into the job thinking "I'm going to take this and that shortcut." then I'm unsure if you have the discipline to not take the ultimate shortcut and just let AI handles everything.
@Aloe The only time a machine-translated translation patch "motivated' anyone to make a proper translation was the PSP Suikoden game, and it wasn't because of any technical roadblock being cleared. It was because the first patch that used Google Translate & DeepL was so buggy and badly translated that someone felt they just had to do it themselves.
I suspect that the the majority of retro Japanese games aren't any more technically complex than the Western games on the same consoles, but the hardest ones to translate also have the weirdest technical roadblocks. The PSP Gundam Ghiren's Greed games are prime examples. It isn't just the great wall of text, but also that instead of using actual letters, the text boxes in these games display pictures — because the text contain so many different kanji characters that it was impossible to just extract the text from the Sega Saturn/PS1 versions and put them into the PSP ports.
AI works best on something that it has been trained with a really large dataset. But we're talking about obscure coding sleight of hand that one or two Japanese developers used to overcome the console's limit. I don't think the AI will work well in overcoming that type of problem, and even if it can, there is still the issue of the volume and difficulty of the actual translation. People who can't fix the technical problem or find someone to do it properly is also less likely to possess the skill and discipline to tackle the text by themselves.
There was a recent Front Mission 2 translation patch that used machine translation. But the guy who made that said the machine-translated parts were only placeholders while he worked on the technical side by himself. He also promised that he would find a real translator and work with them. So this is how you do it right instead of letting AI does everything for you and hope that someone will come along and fix your mess. Any good fan localization requires the translator and the techie to work in concert.
(And for anyone curious, the Front Mission 2 patch is on hold because the creator's laptop was stolen.)
I maybe wrong here, but it seems like this is just the Dreamcast port of Psychic Force 2012 being ported instead of the PlayStation 2 Psychic Force Complete which includes the original Psychic Force, Psychic Force 2012 and a revision of 2012.
@kevinfishburne Much of your arguments seems to rely on the idea that AI will get "better".
Let's get this out of the way first: Claude, Cursor, Copilot, ChatGPT, etc. They are large language models (LLM). They've existed for a long time, and AI scientists think that they're a dead end if you want to reach the level of artificial general intelligence (AGI) a la the sapient ones.
I don't buy the argument that "AI is the most disruptive tech we've invented since fire, the wheel, guns, electricity, and the computer." The first recognizable AI was created when we still used vacuum tube computers. And industries had used AI in statistical and predictive roles (arguably what AI is best at) for years before ChatGPT was even a thing. It has been around forever, it has made an impact, but world changing impact? Nah.
The corporations behind the aforementioned models, they're scammers. They take an inferior product (LLM), exaggerate its capabilities to the extreme, and suck in billions in hope that they can scam it till they make it. But after hovering up uncountable amount of money, land and electricity, it doesn't seem like they are any closer to making an AGI. I can go further into how this AI boom is Silicon Valley's desperate attempt at creating another growth period after their meta/Web3 gamble failed. But I will keep this short for now.
If AGI exists, then maybe I will entrust a game's localization to it. But it doesn't. Wake me up when it actually happens.
Since you like to talk about tools and tool-users so much, I probably should use homebuilding as an analogy. You know the worst thing we got when all those convenience in construction?
We got McMansions.
You can build a beautiful and livable house from prefabricated panels, engineered marble and Styrofoam molding. But when the owners and the builders understand neither the principles of architecture nor the nature of the very tools and materials they use, they can only build houses that are ugly to look at and unpleasant to live in. They can't even ape good architecture half-competently cause they don't know what is good and bad in the first place.
Of course, it's all goes down to your standard of what is acceptable. If you're fine with living in a McMansion, fine with reading edited machine-translated literature, then you'll probably fine with AI-localized games. Me? I just can't muster enough trust when it comes to the current AI models and the people using them to localize games.
Unless I know the person behind the localization has a good record and is only using AI to fastrack one or two minor things, I'll treat any AI-localized game with a mountain of salt.
To use a more contemporary example: We're seeing many Chinese publishers using AI to localize their Steam games into French, Portuguese, German, etc. These same games have English translation made by men. And the gap between the quality of the man-made localization, even if amateurish, versus the machine translation is definitively noticeable. If people use AI that carelessly on games they sell on a commercial platform, how do you think they will treat games from decades ago that they translate for free?
What I will add to the above essays is the context of how AI is being used in fan translation. These LLM aren't being used by competent people who want to speed things up. They're being mostly used by people with no expertise to fill the gap. They don't know how to code around the game? They use AI. They don't know Japanese? They use AI.
Some of these "fans" have the decency to say that they checked "everything" the AI put out. But like, if you don't have the knowledge, skill and diligence to do the deed by yourself in the first place, how can I trust you 100% that you've correctly pointed out and removed the faults? Especially when the translation of a game can't be detached from its technical aspect and instead both work in close relationship. Even bad lip syncing can make your impression of a game worse.
Then you add the fact how context-dependent Japanese literature is and how LLM doesn't understand context, then everything become a mess. Putting a person at the end of such a process isn't QC. You can only polish a turd so much.
And your point about terrible localization. Yes, humans can put out garbage too. But when it happened, you get to know that this person or this company is bad at localization. You avoid their products from then on. But so many LLM cannibalize each other's codebase and dataset. Even if you're well aware of this particular version of this particular AI is bad, can you be so sure that other versions of it or other AI haven't been infected with its flaws?
It's the same difference between bad artisanal products and bad mass-produced stuffs. The former makes you think the artisan is bad. The latter makes you question the very tools they and others used.
And on the specific issue of the Shadowrun fan translation. The creator said that he didn't know what GenAI was. This article quoted him right at the beginning. I just don't buy this guy given that he said in the same breath that he and his colleagues used LLM at work, and they knew enough about AI to use the specific term "LLM". If you know what an LLM is, you're more likely than not to also know what a GenAI is.
@metaphysician @BellaGoth @Sketcz @McGuildicuff Hey guys, I don't think that Sega developed this "Rent a Hero Z". From the press release (https://www.linecorp.com/en/pr/news/global/20260618/), it seems like Sega simply sold the license to LINE NEXT who handled everything.
It makes sense. Even if Sega want to gamble on NFT, why would they choose Rent a Hero, which hasn't been a thing for decades? Sega know how much their marketing relies on nostalgia. They would've had gone big right out of the gate with something like Sonic.
This is just what Japanese corporations do to IP they don't really care about - just give it to some third-party to make a mobile game.
Even many collectors aren't stringent enough to always demand the best possible and most accurately-displayed sprite arts on their monitor. They're perfectly fine with a small level of blurriness and inaccuracy. What I am afraid is that this company is going to be cheap all the way out and create something that not even a retro gamer with the smallest amount of information would want.
@Andee It's all about the economy of scale. When it comes to consumer electronics, you always aim to use the components that are most plentiful. One reason why Apple and other companies stopped making small smartphones is that they were the only customers of these high-definition but small-size screens. Everyone else uses the big slabs. The economy of scale just isn't there anymore.
Whether this project succeed relies so much on which native resolution will the panel have? Are they going to stick with the most-supported resolution like 640x480 or 1024x768? Or are they going to shoot for 1600x1200, 2400x1800, 2560x1920 (maximum resolution a CRT can display) or 3200x2400? Are they pushing for the resolution that most pre-Windows Vista games support AND works reasonably fine with emulators? Or do they want the best possible resolution to make the thing a premium product?
You'd presume that for a small sub-20-inch display, 1024x768 would be a good choice. You can still get these panels made in China for security monitors and such, though I notice that new 4:3 PC monitors from China have been getting rarer and rarer. And even when they were plenty, their display quality wasn't the best. Bellow what your Samsung or Dell office monitors were able to do back in the day.
This is presume that they are actually using 4:3 panels. Remember the Retroid Pocket Mini screen debacle? There is always a chance that they cheap out and use whatever inexpensive 5:4 or 1:3 panels they have around, then do some finagling to trick the customer.
REMEMBER: LCD screens, especially the low-cost one, will show stretching distortion when set to display any resolution other than its native one, as the image needs to be interpolated to fit in the fixed grid display. And if it isn't a true 4:3, then the pixels will be rectangular rather than square, and circles will be elliptical.
And don't get me started on the loudspeaker or the connectors. You'd think that putting an LCD with fewer and smaller components into a big CRT body means the manufacturer has the opportunity to add good loudspeakers and all the legacy connectors (components, SCART, VGA, DVI, etc.) to the TV. But then again, there is a chance that they cheap out.
(There were so many old CRT that looked like they were stereo but only had one speaker. The other speaker hole was fake.)
This go without saying, don't get your hope up, and wait until the reviews come in before buying anything.
Comments 36
Re: Thin End of the Wedge: Steam Next Fest Under Fire as Demo Festival Drowns in a Flood of AI Slop
@Martin_H There is the matter of whether the editor think that an opinion piece like this, which is bound to collect a lot of opinions, should be put on a site with more traffic. Or maybe the author writes exclusively for this site.
Either way, it isn't a big deal. These Hookshot sites cross-post a lot.
Re: Thin End of the Wedge: Steam Next Fest Under Fire as Demo Festival Drowns in a Flood of AI Slop
@axelhander Ah, SaviorS. I followed the Twitter of the developer of that game ( https://x.com/SaviorS_OOTB ) for a while, so I can say a few things:
So yeah, not the best thing ever, but still a far cry from the kind of lazy AI slop that inundate Steam right now. The dev actually treats the AI as just a tool and tries to design their game around the limitation of the tool. A rare example, I say.
@Martin_H This site posted plenty about new games with retro sensibilities or new games from old developers way before consumer AI proliferated, so talking about new games on Steam isn't out of their ballpark.
Re: Thin End of the Wedge: Steam Next Fest Under Fire as Demo Festival Drowns in a Flood of AI Slop
@axelhander WTF are you talking about? Even before ChatGPT was a thing, there were plenty of free and open source software, royalty-free arts & sound assets, etc. alongside a wealth of documents to teach you how to make a game with them. The barrier of entry was so low, especially for making PC games. Saying that before AI, game making was the domain of the rich is just ahistorical.
Bad games have always existed, and people call out the terrible ones made by humans too. But do you know what makes an interesting bad game versus a boring one? Either the game is bad in a singular, unthinkable way or its flaws embody the neuroses of its creator. You can learn something new from an interesting bad game. Being boringly bad like vibe-coded slop just means it fails at the fundamentals. Don't expect people to show any mercy to boringly bad games.
It's weird for AI supporters to act like they're surprised when other people show disdain for AI. Sure, AI has done great things, but when people see their social media or game storefront being inundated with lazy AI-made *****, it's normal for them to feel negatively toward AI.
But I am willing to hear your case. Just show me a game made with AI from this Steam Next Fest that you think is a gem.
Re: 'Too Big a Risk' - Indie Dev Digital Sun Delays ReVamp Public Playtest over AI Cloning Fears
@axelhander Only a fool would think an LLM model is some "collective intellect" of humanity. Especially given actual cases when the clones created by AI ended up way worse than the original:
https://automaton-media.com/en/news/the-developers-of-a-viral-fishing-game-faced-countless-ai-slop-mobile-clones-but-the-rip-offs-were-so-bad-they-ended-up-promoting-the-real-game/
Rip-off games have always existed, and people call out the terrible ones made by humans too. But do you know what makes an interesting bad game versus a boring one? Either the game is bad in a singular, unthinkable way or its flaws embody the neuroses of its creator. You can learn something new from an interesting bad game. Being boringly bad, like the AI clones, just means it fails at the fundamentals. Don't expect people to show any mercy to boringly bad games.
Wake me up when AGI is actually a thing and you can make the case that AI is greater than the sum of its parts.
Re: 'Too Big a Risk' - Indie Dev Digital Sun Delays ReVamp Public Playtest over AI Cloning Fears
Similar stories are happening in Japan:
https://automaton-media.com/en/news/ai-generated-rip-off-of-unfinished-japanese-indie-game-beats-original-to-release-sparking-concerns-about-safety-of-sharing-dev-progress/
It has gotten so bad that when a Japanese dev's passion project has, ahead of its Steam launch (the game originally released on DLSite in March), been not just pirated but translated and modded. Mind you, this is a very text-heavy simulation title, so it's very likely that the hackers used AI to translate the game too. The dev got fed up with both the piracy and how AI is inundating the world that he just quit making games.
https://automaton-media.com/en/news/japanese-indie-developer-devastated-after-adult-fantasy-slg-he-spent-9-years-making-is-pirated-ahead-of-steam-launch-announces-withdrawal-from-industry/
There is a rare positive case, though. Some indie game got cloned, but the AI clones were so bad that they gave the real thing a glow-up:
https://automaton-media.com/en/news/the-developers-of-a-viral-fishing-game-faced-countless-ai-slop-mobile-clones-but-the-rip-offs-were-so-bad-they-ended-up-promoting-the-real-game/
Re: 'The Community Aspect of Decompilation Has Been Broken' - decomp.dev Closes in the Wake of AI-Based 'Bad Actors'
@Lanmanna Someone is maintaining a list of decomps and recomps made by actual humans: https://docs.google.com/document/d/17BgdKswrijiXCysCcE6Jk8fLoV8JuIzjLRxV2-MFXeo/edit?tab=t.0#heading=h.7nyp0xwjfvrw
Re: 'The Community Aspect of Decompilation Has Been Broken' - decomp.dev Closes in the Wake of AI-Based 'Bad Actors'
@Thad I get you. That is what artists have been saying for years, that a lot of people who aren't knowledgeable about arts will look at the flood of AI arts and start questioning every piece of art they see on Twitter if it's AI or not. It just dampens their enthusiasm. This phenomenon has already happened.
The funny thing is that AI bros will always say "I tag my AI arts properly." to cover their asses. Then they found out that most people absolutely doesn't want to engage with AI arts. So they're now whining that tagging AI arts is discriminatory. They shows their true intention of wanting to replace hardworking artists with their lazy slop.
Back to the topic. What I was trying to get to with the original post is that there is a community component to it. Call me old, but if you were in the forums scene back in the day, you just know it. Some fan who does a proper localization isn't doing it out of vanity and their ego like the AI bros. It's hard work for no money and little fame. Sometimes the only thing powering the translator are a few friends he made on the forum.
We can talk about the low standard of people who say "I am fine with terrible vibe-coded translation and recomps." But to me they look less like people who participates in the community and more like some atomized hoarder who only looks to add more and more to their hoard.
Re: 'The Community Aspect of Decompilation Has Been Broken' - decomp.dev Closes in the Wake of AI-Based 'Bad Actors'
@Thad Thanks for the heads up. And the first PSP Suikoden translation was a Google Translate mess. The proper human-made one came out like one year later.
I could be wrong here, but I remember there was a small debate at the time among PSP fans because of the first Woven Web translation and another debacle related to Sword Art Online Infinity Moment — Some guy grabbed the English localization of Sword Art Online Hollow Fragment (which is the remake of Infinity Moment), backported it to Infinity Moment, and then paywalled the whole thing.
Re: 'The Community Aspect of Decompilation Has Been Broken' - decomp.dev Closes in the Wake of AI-Based 'Bad Actors'
There are some localization and decompilation projects that use AI appropriately as the tool it is, like bug testing. However, these projects are few and far between, and each of them is being drown under dozens of lazy vibe-coded "projects" by people who don't know the first thing about programming or retro hardware architecture.
I've seen the same thing happened to the translation community when machine translation became more intuitive and wide spread. Suddenly there were dozens of lazy, sloppy translation of weekly manga & anime. The BS drown out the actually good translation by groups who took their time to do it and therefore were unable to catch up to the treadmill of weekly manga & anime. To say that the slop divided the community was to put it lightly
But it doesn't have to be like that with game localization and recompilation. First of all, we have all the time in the world. Retro gamers are not some overhyped Weekly Shonen Jump fanboys who absolutely need to know what happened in the newest Naruto or Bleach chapter. The debate about speed vs quality is not applicable here.
Secondly, people who are going out of their way to play a fan localization or a game recompilation is already deep in the bush. They know what they want, and they will scrutinize every aspect of what you do. If your audience is already like that, then why even bother throwing together a buggy vibe-coded bowl of slop that you've no intention to fix?
For all the people who say "the bad ones will sort themselves out". It's easy for you to say so as a solitary consumer of the end product. But we're talking about fans producing things, which means the community aspect is unavoidable. You might think "there is enough room for everyone" and the human translators and coders will just go back to do their job properly. Love it or hate it, those fans are doing the job for more than themselves. They need the community by their side, be it for constructive feedback or just cheering. And there is a very real frustration when you see some donkey-asses MF taking your hard work, order the AI to slap some codes onto it and then tell everyone this is the "complete version".
And beside, how many times that people took matters into their own hands after seeing AI slop? All I can name were the PSP Suikoden translation and the PS1 Air Management translation https://www.timeextension.com/news/2026/09/new-air-management-96-ps1-fan-translation-released-to-make-insulting-genai-effort-obsolete. These are the exceptions, not the rule.
Finally, for anyone interested, someone is maintaining a list of decomps and recomps made by actual humans: https://docs.google.com/document/d/17BgdKswrijiXCysCcE6Jk8fLoV8JuIzjLRxV2-MFXeo/edit?tab=t.0#heading=h.7nyp0xwjfvrw
Re: Feature: "It's Insulting & Hurtful" - Beetle Adventure Racing Decomp Dev Kicks Back Against The Rise Of AI Recomps
@Beyerun That is the thing. Emulator is good enough for normal people. Only those who are deep into the bush will deliberately seek out recomps, like people who just have to play with mods.
So why do people even bother making vibe-coded recomps in the first place if they aren't going to fix the obvious bugs? The type of gamers who play recomps know what they want, and they're going to scrutinize every aspect of the game. It all seems like a waste of time to me.
@RupeeClock VideoGame Esoterica has been sketchy even before decomps and recomps became popular. I don't trust anyone who put out several YouTube videos each day to be able to do their research properly, especially in a field like retro gaming. The guy is spinning a content mill to make money.
This argument that "There is room for everybody." has happened before with manga scanlation. The moment machine translation became more intuitive and widely available in the 2000s, I started seeing more low-effort manga translation. They sucked and were a waste of time. Nevertheless, there were scanlation groups that took time to make a good translation, only to be overshadowed by some machine-made slop because in the content mill of weekly manga, so many people only care about the "plot" and not the literature itself.
It sucked back then to see some people having wrong perception of a manga because they only consumed the machine-translated slop. I'd rather it not happen to retro games. Though like I said at the beginning of this post, you've to be in the know to deliberately seek out recomps. If you play a vibe-coded recomp and have a bad time, it's kind of your fault.
Re: "You Bastards" - Ex-PlayStation Exec Shawn Layden Reveals What A Nintendo Counterpart Thought About PSP
@ElkinFencer10 The PSP came out in the early 2000s. Back then flash storage was still expensive and Sony wanted 3D games with PS2-ish graphics, so PSP games were much larger than DS games. UMD was the only way to get those games on a medium cheap to produce and thus could maintain the publishers' profit margin.
Not to mention that the UMD grew out of the MiniDisc that Sony created, which meant they already had the technology and experience working with it.
Ditto with the plastic screen. Again, it was cheap enough and worked well enough for the 2000s.
Anyway, Shawn Layden is right about Sony getting back to the handheld business. You need an entirely new system set up to support the thing. Just look at the how Nintendo had to focus on the 3DS to the Wii U's detriment. The Switch, a handheld-console home hybrid, had to happen because Nintendo wanted to have only one workforce and one budget to maintain their ecosystem.
If the rumor about Sony cooking up a new handheld is true, I hope that they know which portion of the market will it occupy. Many first buyers will be people who already have a PC or a Switch as their main gaming system. As such that handheld needs to deliver really good handheld experience instead of being some kind of portable PC like the Steam Deck or the Legion Go.
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@Martin_H Shrug Not my fault that many people have been doing AI stuffs so weird and off-putting that they invite the categorization. Flooding hobbyists' space with ill-informed and comically bad fake news and fake documentaries. Stealing artists' artworks by running them through AI and then claiming credit. Circumventing websites' anti-scraping measures that are set up by owners who explicitly don't want to be scraped. Etc.
If these people don't want others to name them names, they should take a look at the mirror and think hard about how unpleasant their actions have been for others to put them in the "to avoid" basket.
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@Martin_H Yeah, and the AI companies are buying off every local and state officials they can. I know a few cases in Louisiana and Oregon where the locals had already voted down the motion to build data centers, only for the councilmen to overturn the decision in closed meetings. It's not even "bad city planning" anymore. Anyone can see how bad the idea is from miles away. This is pure uncut corruption, and the lion's share of the blame still falls to the AI companies. Not that the politicians are any cleaner.
You would think that with the billions they have received from investors, the corporations could have had done something like setting up a solar farm. Nope. It's all about taking advantage of the existing grid and demanding the state to keep giving out those tax breaks.
"Witch-hunt mentality". Now you are just over-exaggerating the situation. Some people here have very strong opinion about AI and like to ridicule the incompetent individuals who used it. I don't think I've seen anyone who had threaten to take any harmful action. Conversely too many AI bros like to claim victimhood's high ground just because people ridicule their bad AI creations and the way they irresponsibly used the models.
Re: "Proudly Made With Human Hands" - MediEvil Recomp Is (Un)Alive, Bringing The Spooky PS1 Adventure To PC
Someone is doing a list of decomps and recomps that are actually done by humans. Here is it: https://x.com/Funky90sLion/status/2105807739398357501
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@Martin_H And just because the people making the point about theft may have participated in piracy themselves doesn't mean their argument lacks ground. Germans are still being fined for pirating and sharing movies through torrents. Theft is still illegal in the eye of the law, just like secretly collecting data from your customers and feed it to a LLM. The reason cops don't stick it to individual pirates is because they'd rather spend their limited resources to go after the big fishes like illegal streaming sites, etc.
There is also the argument about how AI companies are destroying the very thing they used to trained their models, like this one about Amazon destroying rare books: https://www.404media.co/we-tracked-a-shipment-of-rare-books-it-ended-at-an-amazon-ai-training-facility/
Until the authors of these vibe-coded translations and recompilations can prove that they only used locally-run models trained solely on data that have been allowed by their owners, then people have the right to be suspicious and critical of these projects from the environmental and intellectual rights angle.
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@Martin_H If you're talking about the tests to determine the capability of the the AI models, then they have been and are still being done by academic institutions and AI companies. You can just search for them and read them either for free or pay for a research hub paywall.
However, there is a discrepancy between what the corporations admit to themselves and what they're telling consumers and lawmakers. Way before the news reported the first case of suicide encouraged by AI, Facebook's internal report had already shown that their AI could induce psychosis. Engineers and safety activists have been talking for years about putting up guardrails, yet corporations like Facebook were so scared of their models bootlicking their users one degree less that none of the safety measures was implemented.
Instead of talking about actual problems like this, what are the AI corporations' biggest talking point nowadays? It's Skynet, again. It's an attempt to distract from the actual consumers' rights. And don't get me started on Silicon Valley trying to use the legislation as a cudgel to choke out the "dangerous" open-source models from China, which is a blatant attempt of unfair competition.
And I've already talked about locally hosting AI server in my previous post. Again, I've to repeat that many people using AI to do the low-effort fan translations and recompilations are not using local models. They talked about how many tokens they had burnt, and in the same breadth talked about how to evade the rate limit by setting up several accounts. These ain't your Ollama users with their MacMini. This is just how low-rent these people's operations are. And if you talk to them about building their own rack, they're just going to whine that "but I don't have the money."
Your point about city planning: The Chinese have been building several power plants in the Western half of their country to supply the data centers nearby. The thing is they built these complexes in sparsely-populated regions, and the companies still have to adhere several regulations. America ain't building new power plant, data centers are plugging into the existing power and water grid, and many are operating while consuming way more energy and water than the agreed amount. All the while enjoying enormous tax breaks paid by tax payers. I know the Chinese aren't subsidizing their AI companies this much.
With that in mind, shouldn't the burden of responsibility falls onto Silicon Valley instead of whichever local official you blame for "bad city planning"?
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@Martin_H Did you actually read my posts? Nowhere did I say that there should be some kind of test for people who want to use AI. What I said is AI naturally attract a lot of numbnuts who don't know the first thing about what they're doing the first place.
But I can't effectively stop someone from using AI if they really want to. This is the crux of my argument about consumer protection. The industry has an obligation to drill into heads of the dumbest people about what AI can and can't do, what is the impact to user personally or to the environment, etc. If the industry want to withhold the information then the state will step in.
And @81632bit made a great point that adds to mine. The consumers have the right to know if the model they're using have been taught on ethically-sourced data, is the data centers in use not endangering the local community, etc. They need to know these things before wasting valuable resources on whatever BS they want to use AI for. As I see it, the way your average Joe is using AI produces little thing of use while incurring proportionally high cost on the resources currently available. And they aren't even remotely aware of it.
If these people are hosting their own local AI server and footing the electric and water bill out of their wallets, then I wouldn't have cared. But they aren't doing it. The cost of operating the AI business is so heavily subsidized by the state and financial institutions that it creates the perfect circumstance for enormous waste to happen right now. At a time when everyone are talking about a climate collapse and an economic recession. Normal people only see how much their tokens cost and think AI cost proportionally to run. It isn't.
Re: "I'm Very Serious About Doing This The Legal Way" - The Creator Of Mega Drive / Genesis Fan Port 'Mega Splatterhouse' Wants To Make Things Official
@Prizm Of course the guy know about the reality of Genesis piracy and bootleg cartridges. He is doing a Genesis port in the year 2026. Anyone doing such a thing is surely relying on emulation and the Everdrive because, surprise, they ain't gonna have an official Sega Genesis dev kit laying around.
Monster Bath most likely just want to make some bucks. If he think he can do it in a legal-ish way, then I say let him try. I'm not very optimistic that he'll be able to convince Bandai-Namco, and even if he can, the ROM is going to be dumped and the carts are going to show up on AliExpress within weeks anyway. By that reality, might as well try to do a legal release and get the money from the few people who, for some reason, absolutely need Bandai-Namco's seal of approval.
Re: "I'm Very Serious About Doing This The Legal Way" - The Creator Of Mega Drive / Genesis Fan Port 'Mega Splatterhouse' Wants To Make Things Official
@SlangWon I need some elaboration about your idea of "for the fans" here? Monster Bath Games seems to be pretty passionate about this project. Are you perhaps thinking that for it truly "for the fans", the complete ROM should be released for free for everyone?
(If we're talking people who really, really want to play the game but can't afford the arcade board, then why are they even bother with this port? The superior arcade original is easily emulated. Or play the Namco Museum/Arcade Archive version if they don't want to deal with emulation.)
Like you said, the ROM is going to be dumped and the bootleg carts are going to show up on AliExpress within weeks either way. And people who buy bootleg carts or put the ROM on their Everdrive isn't going to buy the official release anyway. Do a legal release and the guy can probably make a few bucks from people who absolutely need Bandai-Namco's seal of approval.
And, like, it's a known practice for porting houses to take the assets from an existing game without permission to make a demo, then take it to the rights holder in hope of getting a contract to port the game. The only problem I see here is Monster Bath made the demo public instead of taking it to Namco-Bandai first.
@ElkinFencer10 As for how likely Namco-Bandai is going to agree to this, it all comes down to who know who and whether think company think they can sign a deal with whoever the actual publisher is for a cut of the profit.
This is Namco-Bandai we're talking about here. Games like Xenosaga and Soul Calibur have ten times the number of fans of something like Splatterhouse, yet Namco-Bandai have given them less than crumbs despite these fanbases yelling at the top of their lungs for years. "Fan pressure" isn't going to budge them.
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@Martin_H Yeah, but I don't see how it plays into our debate.
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@Martin_H Do you really think that the air force officers who signed up to be astronauts or the congressmen who argued for NASA's budget didn't read the mountain of scientific papers to convince them that reaching the moon was both viable and desirable, and the whole thing was an worthy ordeal?
Belief doesn't come out of a vacuum you know. And you do need to convince other people. We're talking about collective efforts that require the marshaling of so much manpower, money, land, materials, etc. You know, like building rockets to the moon or developing AI? You can't just go all Voluntarism here.
Like, haven't you seen how rural Americans are reacting to new data centers? They don't believe the data centers won't suck up all the electricity and groundwater, leaving them high and dry. They don't believe the data centers won't cause constant headache because of their noise. They don't believe the data centers will generate more jobs for the locals or put more money into the local economy. They don't believe that their lives are going to get materially better if AI gets better.
They believe that their lives and their communities can only get worse when a data center is built on their area. So they're trying to block any data center project they can.
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@Martin_H Sure, he is free to do everything you said. People are also free to judge this person based on the creation he put out as well as the tool he used to do it. And it's not a good look.
It's like if you care about AI and how AI is being received by the public, you'd care about how companies are just slapping the "AI-powered" label onto every product. It just cheapen the word and distracts people from what AI is actually useful for.
I've to correct you about the moon exploration analogy though. Scientists had theorized about reaching the moon since the 18th century. Jules Verne wrote From Earth to the Moon in 1865 based on actual research papers. Scientists did rigorous scientific experiments half a century before the Apollo program was even a thing. So, there is a very long history of theorizing, arguing and producing evidences to convince people about the viability of landing on the moon. It's like the anecdote about Columbus proving to the queen of Spain that Earth is round. The Ancient Greeks had already experimented to prove the Earth is round AND calculate its circumference.
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@Martin_H Like I said, AI is being sold to people who have no clear idea what are the capability, limit, pays and gains. That include people who wrongly think that AI on itself can carry them to a good recomp of Knuckles' Chaotix while they're so blind to the very visible flaws in the creation they put out.
Seriously, if the bugs are that visible and severe then why would someone who, let's us assume is a competent and self-respecting programmer, even bother to publicize a demo build? Just focus on fixing the bugs and maybe put out a YouTube video if you really want people to know what you're doing. It makes people question just how much does this individual know what they're doing with AI and how much they care about it.
@leogames If it suits you to assume that I'm so close-minded and doesn't care enough about your opinion to change mine — while I've shown no sign of such — then sure, be my guest. Just get off your high horse the next time you argue with someone else.
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@Martin_H AI is being sold to and used by people who have no clear idea what is its capability, limitation, pays and gains. As I see it, it's a matter of consumer protection, and the AI industry is muddying the water with all their talks about "Skynet" instead of talking about the actual problems happening right now, from the spyware embedded in the models to AI pushing the psychosis of vulnerable people to the point they committed self-harm. And yes, it includes people who wrongly think that AI can carry them to a good localization or a good recompilation while they have none of the skill & knowledge to detect the bugs in the first place.
There is also the matter how much humans are using AI as a form of plausible deniability. You see CEO saying that their products are ***** or someone got wrongly laid-off because "AI did it." You see generals saying that their airplanes bombed some school or hospital because "AI did it." Psychotic influencers used to say "The algorithm pushed me to do it." but nowadays you've got people with actual power putting up the same BS with AI. Both the AI corporations and their customers are saying that they can't be forced to take responsibility because all the decision are made inside the black box that is AI and somehow we all just have to put up with the BS.
And don't get me start about the actual evil stuffs like the IDF's Lavender.
We've tools to deal with consumer protection and negligence leading to harmful consequences. The laws, the education system, the civic society. Ethics in AI construction and usage has been an academic field since the last century. Society and the state can definitely bring this matter to the table right now and work out a legal, moral, economic and cultural consensus about what are the acceptable use cases for AI, what are the penalties and punishment for misusing AI, etc.
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@leogames You presented your argument that AI-coded software shouldn't be slop when it's well tested and fixed. I presented my argument on why AI-coded software hasn't been well tested and fixed in practice.
To counter my argument, you simply said "things will get better" without actually saying how would it get better. Which forced me to go back to your original post, which I quote here:
"Whether we like it or not, it is actually pretty incredible that AI can write what it has here. Sure there are bugs - but it's very clear that the majority of the game is there and that AI has come a long way in just a very, very short space of time. Those bugs, alongside more thorough testing, won't be there for long."
Am I wrong to read this passage and assume that your position is that things will resolved themselves as AI gets better? It's the "...alongside more thorough testing..." that I question (if it will happen) and you haven't responded to.
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@leogames And it's precisely this attitude that things won't get better. AI on itself isn't bad, but the way people is using it is terrible — from creating slop to generating kill lists (like what Israel is doing).
And if the fault is with us humans, then it can be changed, it can be fixed. But like I said, it stand in the way of the profit that goes into Silicon Valley. They have an incentive to keep things as bad as they are now.
You people are naive to think that just because the tool will get better mean everything will be solved when the faults with the user and the usage are still there. That is even presuming that the tool will get better soon enough when LLM has been around for a long time, scientist think that they're a dead end when it comes to AGI, and recent financial reports show that AI companies are increasingly relied on circular finance and doggy bookkeeping to hide the fact that the gain their new products made are marginal.
Re: "This Is What Happens When You Rely On AI Slop" - Fans Aren't Impressed With Vibe-Coded Knuckles' Chaotix Recomp
@leogames @Martin_H You just pointed out the Achilles' heel of vibe-coded recompilation. Emulator is good enough in the eyes of a casual gamer. One has to be pretty deep into the bush to have the kind of requirements you listed. But that kind of person is also really strict with what they want and will dig everything up to find faults.
As such, the maker of the recompilation has to be just as strict to fully satisfy their intended audience. Do you think that someone who used AI as a shortcut for their laziness and lack of skill have the will, knowledge and diligence to do such a thing in the first place? Especially when they're doing it for free and without the pressure of the boss, customer and contract to make sure the product is good?
Since this site started talking more about AI, I've been making this point again and again: There are more INCOMPETENT and LAZY people using GenAI than COMPETENT and DILIGENT people.
It isn't just game localization and recompilation. Just talk to people who are into Heart of Iron, Skyrim, etc. modding and they'll happily tell you about how their communities are being flooded with low-quality vibe-coded mods made by individuals who don't know the first thing about coding. Back in the old days these people would've had been forced to either learn on the job or just quit and don't make it. Nowadays AI allows them to just throw ***** out and stink the place.
It's an entirely predictable outcome given the way the AI industry is marketing AI to people. They're selling AI as a crutch for people with no skill and no motivation to work and learn. They're selling AI as the magic wand to turn people from being a prole into a petite bourgeois. It isn't that different from how crypto, blockchain and DeFi were sold to us years ago.
(I can go into how the AI boom is the direct consequence of Silicon Valley's failed gamble with crypto and Web3/Meta. They desperately need another growth engine to justify the unimaginable amount of investment they're receiving. But it'd be too long for this site.)
People had used AI in a scientific and business capacity long before ChatGPT was even a thing. The differences were that they used AI in what it was actually good at - data processing, pattern-recognizing and prediction — and there were always human overseers in the loop. The AI industry's hype machine desperately doesn't want your average Joe to know it. They stands to benefit the less you know about their products and the more you depend on whatever crumb they gives you.
All these talks about "AI is getting better" won't matter if we don't impose strict limits and educate everyone about when, where and how they should use AI. As long as we let the AI charlatans and AI snake-oil salesmen run the show, there will only be more disasters created by stupid people who think AI can compensate for their stupidity.
As for the recompilation matter, I'll add this point: There is a reason why porting games is hard and studios like AM2 and Iron Galaxy make a business out of it. Like localization, you just have to account for so many different things, from performance to audio & visual, when you bring a game from one platform to another. It's even more so when you're recompiling — essentially natively porting — old games that were made to be so intertwined and dependent on their original hardware.
Re: A Second English Translation For Sega CD Shadowrun Appears, And It's Also AI-Assisted
@Aloe Ah, you talked about the "average programmer" and "great programmer". I ain't talking about them. My point since the beginning of this thread is that when it comes to fan translation, how much is AI being used by competent people who want to speed things up or widen the scope of what they can do, and how much is it being used by lazy, incompetent people with no skill and is unwilling to learn?
Scientists had used artificial intelligence in fields like cryptography for years before ChatGPT was even a thing. No argument from me about that part. But we are talking about the Dunning-Kruger effect here. The way the AI industry is selling AI to normal people, it's creating so many cases of delusion. So many lazy, low-skill, less-informed people offloads their thinking to AI, and the sycophantic way the current models are programmed to respond to their users makes it even less likely for them to see their errors and learn anything. The rise of "vibe coding" LARPers is one example.
Many commenters in this thread seem to be under the impression that there is a substantial number of "competent" and "motivated" people who just need a little wind under their feet to be finally able to vault over the fence and localize the games they love. They forget that retro gaming is still a niche hobby, and there aren't many people who have a solid foundation of how old games and old consoles work — or the diligence to do the job. There a reason why the Sega Saturn port of Symphony of the Night, a game well-known in the retro community with little to translate outside of what the PS1 version had already covered, still waited years before receiving a translation patch.
You would think that anyone who want to dabble in translating retro games is required to learn, like, how retro consoles manage their memories or the length limit of the text box. But some people (at least the ones I observed) seems to think they can skip all that "unsexy" learning and go straight into the job thank to AI. And when you nudge them about it, they act like they are being insulted to suffer the "indignity" of wasting their time learning an unmarketable skill. The same people back in the day would've had been forced to learn on the job or just quit before they become an annoyance.
Re: A Second English Translation For Sega CD Shadowrun Appears, And It's Also AI-Assisted
@Aloe I've seen more than a few translation efforts that broke down because of unsolvable technical hurdles. The thing is that the people in those cases had enough of a standard to not put out a patch and, if possible, made themselves available to others who could and wanted to do it properly. They didn't use AI to cover their asses and put out a patch that they didn't even understand what was wrong with it in the first place.
And the example with the Air Management game is even more insulting when you consider that the guy who made the AI "addendum" didn't consult the original creator in the first place. People who've made addendums for others' fan translations did that. It's just common sense. The fact that the AI guy couldn't even bother to do it tells me that he didn't really put much thought into what he did.
You can go to Hacker News and the people there will tell you about how AI is only good when you know what the solution looks like in the first place. If you're in the middle of building the pipeline and are unsure about what comes out of the process, using AI to solve everything is just going to make you and others even more unsure about the final product's integrity. So if you actually care about what you do, AI forces you to spend more time trying to understand it and fix what it did wrong. It cost more time than just do things by yourself bit by bit and fix the errors along the way.
I don't want to use the word "gatekeep", but at some point you just have to think that the technical difficulties do have an effect on separating the people who actually have what it takes to localize a game to those who don't. At least have the people skill to find someone who can help you. If you jump right into the job thinking "I'm going to take this and that shortcut." then I'm unsure if you have the discipline to not take the ultimate shortcut and just let AI handles everything.
Re: A Second English Translation For Sega CD Shadowrun Appears, And It's Also AI-Assisted
@Aloe The only time a machine-translated translation patch "motivated' anyone to make a proper translation was the PSP Suikoden game, and it wasn't because of any technical roadblock being cleared. It was because the first patch that used Google Translate & DeepL was so buggy and badly translated that someone felt they just had to do it themselves.
I suspect that the the majority of retro Japanese games aren't any more technically complex than the Western games on the same consoles, but the hardest ones to translate also have the weirdest technical roadblocks. The PSP Gundam Ghiren's Greed games are prime examples. It isn't just the great wall of text, but also that instead of using actual letters, the text boxes in these games display pictures — because the text contain so many different kanji characters that it was impossible to just extract the text from the Sega Saturn/PS1 versions and put them into the PSP ports.
AI works best on something that it has been trained with a really large dataset. But we're talking about obscure coding sleight of hand that one or two Japanese developers used to overcome the console's limit. I don't think the AI will work well in overcoming that type of problem, and even if it can, there is still the issue of the volume and difficulty of the actual translation. People who can't fix the technical problem or find someone to do it properly is also less likely to possess the skill and discipline to tackle the text by themselves.
There was a recent Front Mission 2 translation patch that used machine translation. But the guy who made that said the machine-translated parts were only placeholders while he worked on the technical side by himself. He also promised that he would find a real translator and work with them. So this is how you do it right instead of letting AI does everything for you and hope that someone will come along and fix your mess. Any good fan localization requires the translator and the techie to work in concert.
(And for anyone curious, the Front Mission 2 patch is on hold because the creator's laptop was stolen.)
Re: Dreamcast Deep Cut Fighter Psychic Force 2012 Coming To Modern Platforms in 2027
I maybe wrong here, but it seems like this is just the Dreamcast port of Psychic Force 2012 being ported instead of the PlayStation 2 Psychic Force Complete which includes the original Psychic Force, Psychic Force 2012 and a revision of 2012.
Re: Sega CD Shadowrun's English Translation Was Developed Using AI
@kevinfishburne Much of your arguments seems to rely on the idea that AI will get "better".
Let's get this out of the way first: Claude, Cursor, Copilot, ChatGPT, etc. They are large language models (LLM). They've existed for a long time, and AI scientists think that they're a dead end if you want to reach the level of artificial general intelligence (AGI) a la the sapient ones.
I don't buy the argument that "AI is the most disruptive tech we've invented since fire, the wheel, guns, electricity, and the computer." The first recognizable AI was created when we still used vacuum tube computers. And industries had used AI in statistical and predictive roles (arguably what AI is best at) for years before ChatGPT was even a thing. It has been around forever, it has made an impact, but world changing impact? Nah.
The corporations behind the aforementioned models, they're scammers. They take an inferior product (LLM), exaggerate its capabilities to the extreme, and suck in billions in hope that they can scam it till they make it. But after hovering up uncountable amount of money, land and electricity, it doesn't seem like they are any closer to making an AGI. I can go further into how this AI boom is Silicon Valley's desperate attempt at creating another growth period after their meta/Web3 gamble failed. But I will keep this short for now.
If AGI exists, then maybe I will entrust a game's localization to it. But it doesn't. Wake me up when it actually happens.
Since you like to talk about tools and tool-users so much, I probably should use homebuilding as an analogy. You know the worst thing we got when all those convenience in construction?
We got McMansions.
You can build a beautiful and livable house from prefabricated panels, engineered marble and Styrofoam molding. But when the owners and the builders understand neither the principles of architecture nor the nature of the very tools and materials they use, they can only build houses that are ugly to look at and unpleasant to live in. They can't even ape good architecture half-competently cause they don't know what is good and bad in the first place.
Of course, it's all goes down to your standard of what is acceptable. If you're fine with living in a McMansion, fine with reading edited machine-translated literature, then you'll probably fine with AI-localized games. Me? I just can't muster enough trust when it comes to the current AI models and the people using them to localize games.
Unless I know the person behind the localization has a good record and is only using AI to fastrack one or two minor things, I'll treat any AI-localized game with a mountain of salt.
To use a more contemporary example: We're seeing many Chinese publishers using AI to localize their Steam games into French, Portuguese, German, etc. These same games have English translation made by men. And the gap between the quality of the man-made localization, even if amateurish, versus the machine translation is definitively noticeable. If people use AI that carelessly on games they sell on a commercial platform, how do you think they will treat games from decades ago that they translate for free?
Re: Sega CD Shadowrun's English Translation Was Developed Using AI
@kevinfishburne These two essays give succinct reasons for people to put major doubt on AI translation and its effects:
https://dateemups.com/fan-translation-llm/
https://medium.com/@hilltopmailer/the-deadly-serious-part-time-unofficial-hobby-of-videogame-fan-translation-05850fb24e45
What I will add to the above essays is the context of how AI is being used in fan translation. These LLM aren't being used by competent people who want to speed things up. They're being mostly used by people with no expertise to fill the gap. They don't know how to code around the game? They use AI. They don't know Japanese? They use AI.
Some of these "fans" have the decency to say that they checked "everything" the AI put out. But like, if you don't have the knowledge, skill and diligence to do the deed by yourself in the first place, how can I trust you 100% that you've correctly pointed out and removed the faults? Especially when the translation of a game can't be detached from its technical aspect and instead both work in close relationship. Even bad lip syncing can make your impression of a game worse.
Then you add the fact how context-dependent Japanese literature is and how LLM doesn't understand context, then everything become a mess. Putting a person at the end of such a process isn't QC. You can only polish a turd so much.
And your point about terrible localization. Yes, humans can put out garbage too. But when it happened, you get to know that this person or this company is bad at localization. You avoid their products from then on. But so many LLM cannibalize each other's codebase and dataset. Even if you're well aware of this particular version of this particular AI is bad, can you be so sure that other versions of it or other AI haven't been infected with its flaws?
It's the same difference between bad artisanal products and bad mass-produced stuffs. The former makes you think the artisan is bad. The latter makes you question the very tools they and others used.
And on the specific issue of the Shadowrun fan translation. The creator said that he didn't know what GenAI was. This article quoted him right at the beginning. I just don't buy this guy given that he said in the same breath that he and his colleagues used LLM at work, and they knew enough about AI to use the specific term "LLM". If you know what an LLM is, you're more likely than not to also know what a GenAI is.
Re: "We Are Truly Sorry That Things Must End This Way" - Sega's Rent-A-Hero Revival Is Ending Mere Months After Launch
@metaphysician @BellaGoth @Sketcz @McGuildicuff Hey guys, I don't think that Sega developed this "Rent a Hero Z". From the press release (https://www.linecorp.com/en/pr/news/global/20260618/), it seems like Sega simply sold the license to LINE NEXT who handled everything.
It makes sense. Even if Sega want to gamble on NFT, why would they choose Rent a Hero, which hasn't been a thing for decades? Sega know how much their marketing relies on nostalgia. They would've had gone big right out of the gate with something like Sonic.
This is just what Japanese corporations do to IP they don't really care about - just give it to some third-party to make a mobile game.
Re: "Relive The Glory Days Of Console Gaming" - New 4:3 Monitor Takes Aim At Retro Players
Even many collectors aren't stringent enough to always demand the best possible and most accurately-displayed sprite arts on their monitor. They're perfectly fine with a small level of blurriness and inaccuracy. What I am afraid is that this company is going to be cheap all the way out and create something that not even a retro gamer with the smallest amount of information would want.
@Andee It's all about the economy of scale. When it comes to consumer electronics, you always aim to use the components that are most plentiful. One reason why Apple and other companies stopped making small smartphones is that they were the only customers of these high-definition but small-size screens. Everyone else uses the big slabs. The economy of scale just isn't there anymore.
Re: "Relive The Glory Days Of Console Gaming" - New 4:3 Monitor Takes Aim At Retro Players
Whether this project succeed relies so much on which native resolution will the panel have? Are they going to stick with the most-supported resolution like 640x480 or 1024x768? Or are they going to shoot for 1600x1200, 2400x1800, 2560x1920 (maximum resolution a CRT can display) or 3200x2400? Are they pushing for the resolution that most pre-Windows Vista games support AND works reasonably fine with emulators? Or do they want the best possible resolution to make the thing a premium product?
You'd presume that for a small sub-20-inch display, 1024x768 would be a good choice. You can still get these panels made in China for security monitors and such, though I notice that new 4:3 PC monitors from China have been getting rarer and rarer. And even when they were plenty, their display quality wasn't the best. Bellow what your Samsung or Dell office monitors were able to do back in the day.
This is presume that they are actually using 4:3 panels. Remember the Retroid Pocket Mini screen debacle? There is always a chance that they cheap out and use whatever inexpensive 5:4 or 1:3 panels they have around, then do some finagling to trick the customer.
REMEMBER: LCD screens, especially the low-cost one, will show stretching distortion when set to display any resolution other than its native one, as the image needs to be interpolated to fit in the fixed grid display. And if it isn't a true 4:3, then the pixels will be rectangular rather than square, and circles will be elliptical.
And don't get me started on the loudspeaker or the connectors. You'd think that putting an LCD with fewer and smaller components into a big CRT body means the manufacturer has the opportunity to add good loudspeakers and all the legacy connectors (components, SCART, VGA, DVI, etc.) to the TV. But then again, there is a chance that they cheap out.
(There were so many old CRT that looked like they were stereo but only had one speaker. The other speaker hole was fake.)
This go without saying, don't get your hope up, and wait until the reviews come in before buying anything.