Hello, just wanted to share my story as a reminder with all the lay offs in the gaming industry, it has been terrible outside of it too. I've been working as an automation QA for a pretty decent software company and I was about to hit like 11 years on the job. Also on the side I've been trying to get my own indie game dev studio off the ground with releasing 3 small games in the past 3 years.
After saving for like 10 years and living in small rented apartments, last summer I finally managed to secure enough money so I can get a loan to buy my first apartment. After a year of dealing with contractors (yes, they are that slow!!!) I finally moved and was hoping to be able to focus my free time on my game again. However literally first Monday we have a meeting with a head of our department and they said we are practically being replaced with A.I. So around 60% of the company is gone across all offices in 3 different continents...
I guess my dream has always been to go full time game dev, it's just I was hoping to be able to do it on my own terms and not kinda being forced into it. But with no real money coming in from my previous games I might not be able to make payments for the place I bought and spent most of my adult life saving for.
So now I guess I don't have a choice but to go all in on my current game. If you want to check it out and leave feedback on how it looks or even wishlist, you can so here STEAM
Wish me luck guys and gals!
So let me get this straight: you say its not generated with AI, yet you include the fucking Prompt of ChatGPT?
The translation:
Heres the Translation into German. I have chosen a clinical, depressing tone (spoiler: it didn't, its quite meta) to underline the tension and hard ethical content of your game.
You were not kidding lol. Also the case for Italian. Not for French and Spanish though. Not very consistent copy paste skills
So what? I hate AI but blowing a gasket over using it for translations just makes you sound like a witch hunter. Like this is supposed to ruin all the hard work this person clearly put into their game? Yeah right
Having worked in the game industry for over 20 years, let's not forget that translators and technical writers are also professions
Last time I cited my asking price to localize a whole manga series, the lemmy deleted the thread, and pretend to never have asked.
He still posts on c/crackwatch@lemmy.dbzer0.com. The lesson I learned is that when I can reafford to host mine pylova, I will have «Remote instances cannot delete posts» enabled by default. That way, I can’t be gaslight again.
Like, part of me is curious what your rate is now but I'm not actually planning a commission so is that, like, rude to ask?
I was trying to find a translation for a PSP game and couldn't find any online but I had no idea how much one might cost to commission but I guess a single game would be cheaper than a whole series...
It depends on the type of work you're asking.
If you only want a transcription, with no notes, alternative interpretations, directions, dictions, etc., it can be really cheap on the quality you're seeking. The requester had asked several additional things than just a transcription: typesetting, font work, editing, translation notes, sfx, 擬態語, 擬容語, hosting, etc.. Scanlation was more than he could afford for a singular chapter, even at a salaried rate.
For a PSP, it is correct that a single game is cheaper than a whole series, but not many people buy a transcription to a game linearly. Games have options, branches, game lingo, settings, impact, etc.. It's not just asking “does character A do this action Y” but “goal YZK is achieved by the exact complicated arrangements of GOHEDOTN, in order to progress branch Σ of mode O.” Often too, we are requested to reverse engineer parts of the game to make an appropriate working non-hazardous patch file, so the game emulates the origin language narratives as close as the target language can yield. Meaning, game design elements need to be considered as well as testing and quality assurance. Making game localization one of the most expensive media to localize, while also one of the most cut-throat professions that budgetaries eliminate first.
So depending how large and convoluted the game is, you're asking in the tens thousands and up. Maybe up to a million and some change. It reads like a lot, but recall, every part of game localization needs to be tested, retested, and attested. It's not as simple as transcribing a visual novel, even if some VNs break conventions and standards.
And unlike most cut-throat localizers, I am politically, financial, and media literate, meaning, I shan’t be asking for Iranian Rial when I know I can ask for Kuwaiti dinar. I want to be compensated fairly, and anyone I hire to work I know I will need to iron out work that I am not strong in.
Say e.g. the game is a rhythm-music game, I already know I will need to hire a sound editor to assure the localized songs emulate more than just voice synthesized melodies, no? I know my strengths and weaknesses. I also know the importance of having another person than a client and I review our work, and mediate conflicts. Last thing I want is to get sued for not owning rights to the work that was commissioned. NDAs cost plenty more than doing open commissions.
The gasket that's gentrifying folks, polluting water, destroying habitats, and it's overwhelmingly ineffective in it's stated goal to solve, as we read on this very🧵. Why use a gasket when bombing valleys is more effective?
This is just the translation on the Steam page it has nothing to do with the development of the game. Simple as.
I don't mean to hate. The game looks interesting and I'm sure you worked hard on it.
However, you got laid off due to AI, and then went straight to ChatGPT for translations? Sadly, most online translators are highly AI nowadays (and usually produce worse translations than before AI), but there are still better options than chatbots.
It just looks lazy, especially that you completely missed ChatGPT's commentary. Such an obvious and easy mistake to avoid.
My advice would be: absolutely avoid using LLMs and Gen AI for anything related to your game, because your potential customers will blacklist you for using it. And ask yourself: do you really need to translate the game (or its Steam page) to many different languages if it's not even selling enough copies for you to withdraw the money?
A lot of indie games are English-only, precisely because translation is too expensive and unnecessary at first. It's great to make your games available to as many people as possible, but don't forget about principles and quality. First, focus on those two, then maybe try writing translations for any languages you know well.
Language translation tools have been AI-based for nearly 30+ years now (Literally, ML based translation started in the 90's). What are you on about?
Using ChatGPT is definitely egregious for language translation with other translation tools out there, but you seem to think that translation tools using AI are a bad thing. They literally only work worth a damn because they are ML models ("AI"). And have been since they started working well....
I loosely used the term 'AI'. Most people don't exactly know what artificial intelligence is or how it works, so using the terms according to their definitions sometimes creates more confusion than clarity.
Of course translation tools like DeepL have always been based on machine learning. The difference is that a few years ago DeepL worked well and its home page wasn't filled with 'AI this' and 'AI that'. They seem to have transitioned from their old ML models to some of the newer LLMs. From my experience the translation quality is much worse than it used to be.
Man, I'm a laid off dev too. You really should know by now how scrutinous people are about AI anything.
You're not even getting the 100 dollar monthly payments off steam. Forget "just use google translate like pre-AI", why not just focus on one language and look into localization later if this game catches on? Now instead all languages will be scanning your game with a fine-toothed comb.
Well, we are on the Fediverse here where there's a very loud echo chamber of anti-AI sentiment. Outside of this bubble, there's a loud minority that's anti-AI, but the vast majority of end users are more neutral on the subject.
Not in the last 9 months or so. Even if people ignored everything else, data centers have been awful PR since that is only adding to the affordability crisis on people's energy bills. AI sentiment in the West has completely turned around in the last 2 years.
As an indie dev at the scale of "not even getting 100 dollar paychecks" , your audience for the few few dozen reviews (the most important) are more likely to be these niches who are looking to try smaller stuff. And who's more sensitive to slop than the mainstream. You gotta be careful with your PR, especially in your writing and art.
Bro even the rural chuds and soccer moms shit on AI here in magaland
That’s completely incorrect when it comes to gaming, especially the indie gaming scene. Every use of AI that has been found in an indie game has had a lot of controversy to it. Look at The Alters from last year for example. There’s a large enough amount of indie gamers that REALLY hate AI and WILL boycott a game over it. Those people are also usually the most open to new and weird indie games.
Still, marketing is a part of the game, and also one of the easiest things to do without AI via deepl or something
This deepl?
Also note that Google Translate is AI too. It uses neural networks. Has for like a decade I believe.
Really what it seems to me you're saying is that nobody should use machine translations and one should always hire a translator for translating things into/from languages they don't speak?
Yes, Google Translate has used neural networks for quite a while. Recently they've also switched it to using the same Gemini LLMs that they give you chat interfaces with, so it's pretty equivalent to using ChatGPT for translations. It's just another company's version of the same thing, wrapped in a translation interface.
It's not an LLM IIRC (the deepl one at least, Google now uses gemini)
Edit: apparently Deepl has been converted to an LLM since some time
what in blazes do you think is a «large language model»?
All LLMs are AIs but not all AIs are LLMs (and LLMs are what y'all are hating)
Pray tell what AI isn't trained on ❝large contested data❞
🧵, blatantly. But asset diffusers from stolen assets are also highly discouraged, as they are illegal, and it demonstrated the planet the two classes than ever before: corporate power can lobby copyright infringement, while everyone else cannot. Many hate it for quality reasons. I hate it because it's fascism in its purest form: appropriating assets without consent.
Search Engines (no, i don't mean that bullshit google did, i mean any search engine before 2020), reccomendation systems(many other algorithms too), language translation (again, the ones before 2020 not the LLM based ones that google and deepl use right now), OCR, etc etc
I can only agree, i will be a "fan" of LLMs only when someone decide to train them with data collected legally, which no one does apparently, i hope it gets easy and economic to train LLMs, even stupid ones on your own data, so i could train one on my git repos or some documentation
… how do you think these aggregating sorters are trained?
The OCR argument is contradictory.
Fascists are already colluding to make their breaches legal.
Well, i mean, legal like before the fascist scum tried to make the illegal legal
sure.
But be aware crawlers (search engines), OCR, and even suggestions sorters are trained on data, especially adversarially. Some infringing IP, some self-generated.
OP got smacked for using a LLM, when e could have hired localizers, even 🧵.
Can you tell me what 🧵 does mean?
An abbreviation of the word “Thread,” in this particular case “in this thread[ed conversation].”
Emoji for A spool of thread
Oh i see, thx
yw😊
Deepl was around way before LLMs. It was still made with neutral networks trained on research papers and the like but it couldn't be communicated with or generate stuff for you. So u can see why axolotl thought that it didn't use LLMs. However it seems to nowadays use LLMs
…
What makes one think NNs are not LLMs?
LLMs are neutral networks but not all neural networks are LLMs
LLMs are NNs for whoever is wondering btw
Edit: i wanna add that NNs used specifically for translations are way less impactful on the enviroment, i don't know how deepl got the data but google would crowdsource them because it's just wayy easier to do with something so specific and gives higher quality than stolen stuff
For professional work - yes everyone should hire translators.
Yes, people who got laid off from their day jobs surely can afford paying other people.
What about indie development is 'professional'? This is not Blizzard or EA games you're talking about.
Then don't translate your game, or crowdsource the translation like NixOS and other FOSS does.
English is a lingua franca for a reason, most people speak it or at least understand it. Now sure: there are people who don't, but if your game has netted you less than $100 that's not something you really need to worry about.
He’s selling it, so it’s at least somewhat profesional. And no one expects the game to be translated to every language ever. Most indie games only release in English, especially at first.
AI translations just cheapen your art, and as someone who speaks multiple languages I can tell you first hand - AI translations are NOT good when it comes to stories. They’re barely good enough for translating a conversation coherently. And on top of it all is it not incredibly ironic for someone who was laid off because of AI to use AI?
The demo for the game at least is English-only. It's literally only the Steam description page like he said.
At worst this is slightly misleading because someone might expect it to have translations in the game as well, but quite literally, such a non-issue to nitpick an indie dev over.
I was talking more generally, but in this case I agree it’s a small issue. I’d personally prefer if the dev removed the translated descriptions because it’s like you said, misleading and I’d even argue borderline insulting. That being said I do wish the dev good luck and I ought check the demo out at some point soon.
Will somebody tell him?
I know he means "without the use of LLMs.". That's the big thing being forced down People's throats that they don't like.
I doubt people are rebelling against reinforcement learning for physically based animations, or prediction engines to forecast the weather
For any AI Translations, it's good practice to open up a different chat with the model, and have it translate the output back to your original language so that you can double check what is being said.
Tangential: There used to be a website that used Google translate to translate a phrase back and forth between a bunch of different languages until it reached ‘equilibrium’ and stopped changing with each translate. You could get some really crazy stuff out of it.