"Fan Translation"

[princessoftrash]

What is the definition of "Fan Translation"? I have always thought it means manual translation, but apparently that's not the case since there are plenty of MTL that are labeled as "Fan Translation" on this site, this among the most recent examples. So what exactly is a Fan Translation and what makes it different from Edited MTL?

 

[VegitoHlove]

Basically anyone whom does a translation project without the support or directive of the actual developers/publishers of the creative work in question.

It doesn't matter how or even why they do it.

 

[princessoftrash]

Basically anyone whom does a translation project without the support or directive of the actual developers/publishers of the creative work in question.

It doesn't matter how or even why they do it.

I understand, I just find it weird that some translations are labeled as "Edited MTL" while others as "Fan Translation" even though there is often no difference between them. I absolutely loathe MTL and try to avoid it as much as possible, so it's a pretty big deal to me.

 

[♍VoidTraveler]

Basically the same thing as edited MTL, sometimes slightly better than edited MTL.
Not very often tho. If you want more details check my custom scale in my profile, i use it to rate TL quality when i review.

 

[Emerald_Gladiator]

A bit old, but I'll speak up on this, at least in my opinion and experience.

"Fan Translation": In the literal definition, it's a translation created by a fan (a non-professional). In reality, usually that means someone's gone through the game, whether using an MTL for a base or not, and used their own words to translate. Ideally, they're fluent and/or proficient in the language into which they're translating the game. And oftentimes, fan translations are worked on pretty heavily and more carefully than edited MTLs because an actual person is going through the translation, hence the term "fan translation".

"Edited MTL": In the literal definition, it's an MTL that's edited by someone. In reality, it's usually poor quality because most edited MTLs are just the base game dumped into an AI/auto-translating tool and then the "translator" might fix any bugs that are caused to the scripts or maps breaking due to the translation tool, or do a rough run-through of the translation to change anything that is utterly incomprehensible. Very rarely do you get much quality or actual editing, because edited MTLs are meant to be quick and dirty to appease an impatient demographic. For instance, I could change a single word in an MTL and it'd be considered an "edited MTL" in the literal sense of the word.

As a fan translator myself, I do take a bit of umbrage when others call my translations an "edited MTL", because I go line-by-line, for thousands or tens of thousands of lines of text and translate into English. It takes weeks or months of work at least to create a fan translation, at least for me. IMO, my translations look almost completely different than the original MTL if I use them as a base, because I need to take into account proper game or cultural context, the differences in syntax/slang/verbiage between the two languages, and much much more.

Edited MTLs on the other hand can vary greatly in quality, but the vast, vast majority are just MTLs and read like gibberish. I honestly have no idea what is actually changed or "edited" in edited MTLs because in most cases, I've dumped those translations into the GUI of a translation tool and they looked completely unchanged from as if I'd just created a raw MTL in the first place - as in I can see the EXACTLY which AI/Auto-translation API the MTL was created from. In fact, I imagine that a good number of "edited MTLs" are simply MTLs labeled differently as to provide credibility that there's actually some translation QA going in the background when in reality, they're so egregiously bad that nobody could tell the difference between the two.

If that's the case, and "edited MTLs are the same as fan translations" to y'all, then I'd be more than happy to save myself the unimaginable amount of time and energy, just start creating MTLs, and then maybe edit a few lines of text to call them "edited MTLs". Making the distinction between the two terms of "edited MTL" and "fan translation" not only gives respect to the time and effort of the fan translator over someone who just spent 5 minutes dumping the game into an auto-translator, but it also allows for an easier differentiation between "good translation" and "bad translation". That way, you don't NEED a sliding scale to rate or evaluate a translation, or at the very least, it'll be a much smaller scale. Just seeing that a game is "fan translated" should usually mean it's superior to an edited MTL, unless the fan translator isn't proficient in their own language. We're not at the level where general APIs available to the public are better than native speakers who actually understand the context of what they're translating (not yet at least).

 

[desmosome]

A bit old, but I'll speak up on this, at least in my opinion and experience.

"Fan Translation": In the literal definition, it's a translation created by a fan (a non-professional). In reality, usually that means someone's gone through the game, whether using an MTL for a base or not, and used their own words to translate. Ideally, they're fluent and/or proficient in the language into which they're translating the game. And oftentimes, fan translations are worked on pretty heavily and more carefully than edited MTLs because an actual person is going through the translation, hence the term "fan translation".

"Edited MTL": In the literal definition, it's an MTL that's edited by someone. In reality, it's usually poor quality because most edited MTLs are just the base game dumped into an AI/auto-translating tool and then the "translator" might fix any bugs that are caused to the scripts or maps breaking due to the translation tool, or do a rough run-through of the translation to change anything that is utterly incomprehensible. Very rarely do you get much quality or actual editing, because edited MTLs are meant to be quick and dirty to appease an impatient demographic. For instance, I could change a single word in an MTL and it'd be considered an "edited MTL" in the literal sense of the word.

As a fan translator myself, I do take a bit of umbrage when others call my translations an "edited MTL", because I go line-by-line, for thousands or tens of thousands of lines of text and translate into English. It takes weeks or months of work at least to create a fan translation, at least for me. IMO, my translations look almost completely different than the original MTL if I use them as a base, because I need to take into account proper game or cultural context, the differences in syntax/slang/verbiage between the two languages, and much much more.

Edited MTLs on the other hand can vary greatly in quality, but the vast, vast majority are just MTLs and read like gibberish. I honestly have no idea what is actually changed or "edited" in edited MTLs because in most cases, I've dumped those translations into the GUI of a translation tool and they looked completely unchanged from as if I'd just created a raw MTL in the first place - as in I can see the EXACTLY which AI/Auto-translation API the MTL was created from. In fact, I imagine that a good number of "edited MTLs" are simply MTLs labeled differently as to provide credibility that there's actually some translation QA going in the background when in reality, they're so egregiously bad that nobody could tell the difference between the two.

If that's the case, and "edited MTLs are the same as fan translations" to y'all, then I'd be more than happy to save myself the unimaginable amount of time and energy, just start creating MTLs, and then maybe edit a few lines of text to call them "edited MTLs". Making the distinction between the two terms of "edited MTL" and "fan translation" not only gives respect to the time and effort of the fan translator over someone who just spent 5 minutes dumping the game into an auto-translator, but it also allows for an easier differentiation between "good translation" and "bad translation". That way, you don't NEED a sliding scale to rate or evaluate a translation, or at the very least, it'll be a much smaller scale. Just seeing that a game is "fan translated" should usually mean it's superior to an edited MTL, unless the fan translator isn't proficient in their own language. We're not at the level where general APIs available to the public are better than native speakers who actually understand the context of what they're translating (not yet at least).

It's just a nomenclature issue. Most people can understand that running a body of text through MTL (GPT these days) can be a first step in doing a full translation. The base of the translation would be edited line by line, paragraph by paragraph.

I did this exact thing for a non-porn related work. If you are fluent in both languages, you can obviously make your own sentences, but it takes a really long time to land on the best sentence structure for each line. When you start with a machine translated base, you can rephrase it as needed, fix contextual mistakes, fix pronouns, fix wrong facts, etc etc. It massively speeds up a translation, and the work you put into the translation after running it through a machine is what separates fan translation and barely edited MTL. Tbh, you do get influenced by what the machine decided for the original translation before you start editing, but GPT actually writes quite well and might phrase sentences better than what a layman could write on their own.

But there is just no easy and clean way to categorize a translation, because it's a sliding scale and not binary (edited or not). The results should speak for itself. Honestly, GPT is good enough that the people who never played the old ass MTL poetry might find it playable. But it will be no where near as good as a real translation (by someone who is actually fluent in the language to be translated into and good enough with the original language + and actually knows how to write). That's the thing though, many fan translators are not actually good writers and/or have a tenative grasp of the languages. And GPT is great these days. The gap between shitty fan translation and good machine translation has shortened dramatically.

 

[Emerald_Gladiator]

It's just a nomenclature issue. Most people can understand that running a body of text through MTL (GPT these days) can be a first step in doing a full translation. The base of the translation would be edited line by line, paragraph by paragraph.

Either we've seen vastly different edited MTLs or I'm not understanding you properly. I don't think I've ever seen anything labeled as an "edited MTL" actually edited by a human being "line by line, paragraph by paragraph", as you say. Again, I can prove that quite a few edited MTLs are just mostly unedited MTLs just by putting them in T++/Deep/*insert auto-translation tool and going through APIs until the text matches; the "editing" is minimal and lazy, but it does qualify as an "edited MTL" because that's exactly what it is.

GPT actually writes quite well and might phrase sentences better than what a layman could write on their own.

Yes, in a vacuum. But context is extremely important in translation, transliteration, or just plain old writing. The difficulty in using AI to translate is that AI isn't smart enough to understand exactly what it's doing on a grander scale. Sure, GPT and other AI can probably write better than most humans if we give it a defined set of text lines to translate. However, that text might not make sense in conjunction with other text, or even with the rest of the game as a whole.

And in most languages, again, context is important. When I took Latin (aside from memorizing declensions, genders, and so on of words), I was taught that you can't just take a sentence from a passage of text and translate it correctly because there's multiple ways to interpret its meaning. Japanese (language of which is the vast majority of games on this site) is even moreso dependent on context because its alphabet is limited and there's also a great deal of colloquialism and slang that AI might translate literally or incorrectly. In the translation I just uploaded yesterday, I tested GPT with a line of text that it translated incorrectly because it didn't understand a phrase containing a slang term (to be fair, I didn't understand that term initially either and needed to ask the Japanese community, which AI cannot or will not do.)

But there is just no easy and clean way to categorize a translation, because it's a sliding scale and not binary (edited or not). The results should speak for itself.

Well, while I don't think a translation should be binary either, I do think there's an argument for splitting translations into categories of "edited MTLs" and "fan translations", and then subsuming those into their own sliding scales (though I'd probably just say "good/bad"). I've already explained what I believe to be the difference in my previous post, but I think it's important to differentiate. As far as "the results should speak for itself", the issue is that, at least in my experience, "edited MTLs" are so bad or at least so variable in quality that the term's becoming stigmatized as much as "MTL" itself is, to the point that people don't even want to test out edited MTLs, or at least have a negative preconceived perspective when playing one.

I'd rather have the labels so that people can see "fan translation" and understand that at the very least, a human being created this translation and (ideally) went line-by-line and spent the time to improve, if not perfect, their work.

Honestly, GPT is good enough that the people who never played the old ass MTL poetry might find it playable.

I think that's absolutely true. But the bar is incredibly low. It seems that there's droves of non-English speakers on this site who probably wouldn't care if I wrote in Old English or pidgin, or anything in between (at least, that's my assumption based on how they themselves communicate). And yes, GPT is way better than MTL poetry, but again, that's a low bar. The old MTL crap from years ago was literally incomprehensible. Like, not just random errors or weird verbiage, but I couldn't understand what it was trying to say at all.

I think the idea of "good enough" is what's lowering the quality of translations now. Everyone's so impatient for translations that speed is often more important than translation quality; edited MTLs are a reflection of the zeitgeist of the "bare minimum" of this community. The issue is that "good enough" is relative and arbitrary. Hell, to circle back to the previous point, MTLs are "good enough" for the non-English speakers on this site because they likely can't tell the difference between an MTL and a fan translation anyway. I'm guessing you're right in that GPT is "good enough" in the same way for everyone else. But if that's the case, why have fan translators (or even just translators) at all? I mean, I'm not offended if that answer is, "human translators aren't needed", because it saves me loads of wasted time and effort. It just seems that if AI and Machines are as good as you say, I should just be a button-pusher and create MTLs by the dozen rather than spend my precious days on earth toiling in futility.

That's the thing though, many fan translators are not actually good writers and/or have a tenative grasp of the languages. And GPT is great these days. The gap between shitty fan translation and good machine translation has shortened dramatically.

And that's where I'd strongly disagree. I haven't really seen too many fan translators who are poor writers or at least can't outwrite something like GPT on the whole. If anything, some fan translations (usually the worst ones) are just mislabeled edited MTLs, which in themselves are basically just MTLs. Sure, if the fan translator isn't fluent or simply doesn't understand one or both of the languages they're translating, then I'd take GPT any day. But I haven't seen too many of those either (IMO, they just shouldn't be translating if that's the case).

Honestly (and as humbly as I can offer here), after fan-translating over two dozen games, I'd bet money that my worst translation is better than the best GPT-translated version of that game.