Hello everyone, I’m new around, and after almost 20 years, I’m getting my hands back into game making with RPG Maker.
I see localization from a technical standpoint is no big deal with the right tools, BUT… it’s A LOT of work for sure, and obviously machine translating is NOT an option to me.
If you put your heart and soul to your game, why hand it over to a machine to crush your beautiful wording?
So… on my end I can pull English, Spanish and Portuguese at top quality, but I’m thinking about the “hard ones”, which I would love to implement to reach heartfelt hardcore gamers that can appreciate the games in Asia, specifically thinking about Japanese and Simplified Chinese.
My questions are:
- Have you done it before?
- How did you engage it? Directly reaching a translator/native speaker, or machine translation and style correction by a native speaker?
- What tangible results did you get from making that effort?
My games will have strong narrative elements, so if jokes or concepts are lost in translation, the soul of them will dilute.
Thanks first of all for reading, and MULTIPLIED thanks to those who reply!
Quick question on the machine translation part, does that include AI tools like DeepL? Asking because for Japanese and Simplified Chinese I think I’d be tempted to do a first pass with DeepL and then hand it to a native speaker to fix the jokes and tone, which is close to your second option. If AI is off the table too then it really comes down to finding the right translator.
When I think about machine translator I think about tools like Google Translate, more than AI, I want my games to be 100% AI Free.
I was commissioned a content and style correction of a Russian-to-English book translation several years ago, and believe me, it was A LOT of work, but the Publishing House got scammed so many times with Google Translate stuff, that when they requested the service from me they said “Google Translate it, but fix it well”.
The book is now published in both languages, and the English translation (my work) is top class quality, even though I’m not fully well versed in Russian. In other words, a tool, rather than a substitute.
Now, what I DO KNOW, is how much you can lose in those Google Translation, some cases were hilarious like “workhorse computer” (which was more in line with the meaning of “powerful”).
For Japanese and Simplified Chinese, I wonder if that would cut it, it would surely be a shortcut for translator effort and time.
Just so it’s clear before you pick a tool, DeepL is AI, and as far as I know Google Translate has been AI based for years now too. So if 100% AI free is the bar, both of those are out and it’s a human translator from the start.
If you’re ok with the machine draft plus fix it well approach you used on the book, I think the difference this time is that you can’t be the one fixing it. You caught the workhorse computer moments because you know English. For Japanese and Simplified Chinese that catch falls entirely on the native speaker, so I’d put most of the effort into finding a good one and treat the tool as the smaller decision.
Generally speaking, I would suggest you don’t touch machine translation, unless it’s between similar languages. For example: if you can manage Spanish and Portuguese, it might be okay to use machine translation to produce Italian or even French from the Spanish translation. Similarly, if you can get a German translation, it might be okay use machine translation to turn that into a Dutch translation.
The problem is that machine translation inevitably loses nuance. This is especially bad when translating from pro-drop languages like Japanese – it will make up pronouns, and typically it’ll choose the wrong pronoun, so you’ll see things like a girl being referred to as “he”, or the speaker being referred to as “you” instead of “I”.
So, again, I would avoid machine translation – not because it’s AI, but because it just isn’t very good. It’s much better to get an actual human to look at it, preferably someone who is fluent in both the source and target language. Even an amateur should be able to do an okay job if they at least understand both languages.
I’m just noting that, when people say they want to be AI-free, they are usually referring specifically to generative AI such as ChatGPT. As far as I know, neither DeepL nor Google Translate use generative AI as a basis. So, I think using DeepL wouldn’t bother most people on the basis of AI (though it might bother people for other reasons).
Agreed on the Japanese pronoun problem, that’s a good reason on its own to keep a human on it.
One small update on the AI part though, since it matters for FZIdeas’ 100% AI free goal. That distinction was true a couple of years ago but both tools have moved. DeepL switched its translator to an LLM based “next gen” model in July 2024, and the first language pairs it shipped for were English to Japanese and Simplified Chinese:
Google Translate started running Gemini for text translations in December 2025, again with Japanese and Chinese in the first batch:
So for exactly the two languages in this thread, both are generative AI under the hood now. That doesn’t change your advice, it just shows how difficult it is to truly stay free of generative AI.
I guess that at some point, everything (or almost everything) will run generative AI at some level whichever it is, and it will not be considered actual AI intrusion/disruption in creative processes, but just an “organic” tool to every process.
I mean… we have the same deal but at a different level when using character or asset designers/creators, you’re not coding in Assembler or pushing hexas into a file to paint those pixels in your sprite, right?
Thanks for taking the time to investigate and share, much appreciated!
