I don't understand why they didn't translate the dialog outside the game, as a sort of post-process graphical filter which detects speech bubbles and other text in the game and overlays the translated version. Apart from a few diehards flashing custom ROMs to SNES cartridges, I'd expect almost everyone playing it is doing so through an emulator, so this should be easy.
I can see an argument against it that this affects the "purity" of the game compared with the translation being done in the ROM itself, but I can also see a counter-argument that running the original unaltered ROM with what amounts to subtitles is a "purer" approach.
Edit: Just to be clear, I'm talking about a fan translation which is overlaid on the game by pattern matching the image of the equivalent original text on-screen, not an attempt at automatically detecting and machine translating all dialogue.
I imagine that would be hard and slow. You'd have to typeset everything yourself, to boot. And modify the emulator to handle it. Patching the rom would be far more portable.
You still have to do all of that for a ROM patch, plus reverse engineer the (apparently horrible) ROM code. At least the emulator's source is probably available.
It's more than just dialog and speech. All the UI text needs to be rendered correctly, the menus, etc. Often there's modal dialog boxes on top which obscures partial words.
And games don't read very well with machine translation, so you'd still have to teach the system to overlay the correct English localization that was hand made.
All that might still have been easier than dealing with spaghetti assembly, but then you are faced with the daunting task of getting someone to actually download your executable. Patching a ROM is pretty well established process. Downloading an extra executable just to run a ROM in your native language is much less so.
I have recently been reading three books by Clyde Mandelin. Clyde is better known online at Mato. He runs the excellent 'Legends of Localization' web site (https://legendsoflocalization.com).
The two books I suggest are…
• Legends of Localization Book 1: The Legend of Zelda
• Legends of Localization Book 2: EarthBound
[To clarify, when I say "Earthbound", I refer to the game that was released in Japan as Mother 2, with the translated release being called "Earthbound" in the US.]
If anyone wants to make the argument that a real-time translation would work, then I have an additional recommendation:
• press start to translate: This is what happens when you let a computer translate a video game?
I have technical concerns and non-technical concerns with the idea.
With Zelda 1 that would indeed work; there was not much text in the first place, and the typeface was blocky and easy to detect.
With Earthbound, things are much more complex. Ignoring text that is "baked in" to the graphics, the dialogue was much more interactive. The text was smaller, with thinner letter shapes, and when one of your characters was dead, the contrast between text and background was alot lower. You also have to deal with changes in coloration and highlighting, for example when you are selecting from a range of options.
Earthbound also had another trick: The dialogue system was not static text. Instead, the game effectively became an interpreter: The game didn't open a dialogue window, the script itself included the code to open a dialogue window, which the game would parse and execute. Check out http://media.earthboundcentral.com/2011/04/a-look-at-the-mot... for an example
So, if you were to make a mapping of "untranslated text" to "translated text", you would have to work out every possible expansion of the interpreted script, translating each one individually.
If your text ends up being longer than the Japanese text, you'd have to work out a way to deal with that. Either making your overlay bigger than the original text, or maybe intercepting a controller button-press to advance the text.
Of course, all of those are technical concerns. But I think there is at least one big non-technical concern: At least for me, it would be too jarring.
It's the difference between a "lean-back" experience and a "lean-in" experience. The lean-back experience is something like watching a movie. Translating a movie via subtitles means expending additional "brain cycles" to line up the subtitles with the speaker, but I'm not doing much else, so that's OK. Translating a movie via dialogue kindof looks weird, because the voices do not line up with the mouth movements.
Playing a video game is a "lean-in" experience. I am doing a lot more work, ingesting and processing dialogue and NPC actions, keeping track of information, making decisions, and taking actions myself. Translating the dialogue is possible, because you can "line up the mouth movements" (update the programming) to work with the translated text. If there is something else—an overlay, say, or a separate translation file—that takes up a lot more cycles, makes it harder for me to keep up with the game, and lessens the experience.
Of course, there are exceptions! A game like Tetris doesn't qualify either way. A game like Doki Doki Literature Club is more like a movie, and so subtitles fit.
> part from a few diehards flashing custom ROMs to SNES cartridges, I'd expect almost everyone playing it is doing so through an emulator, so this should be easy.
But, which emulator?
People play SNES games on Windows PC's, Macs, Android phones, Raspberry Pi's, 3DS's, etc etc. And the emulators in use have shifted: ZSNES was the emulator of choice a decade ago, but using that today would be ill-advised.
Coincidentally, there exist some old translation patches that were only ever tested against ZSNES as opposed to real hardware. Playing them today is a PITA.
There was an Ask HN some time back that asked "How would you write software to last 100 years?" My answer—which I very much stand by—was: "Target the NES".
An aside, I absolutely adore Vice games (formerly Waypoint). They seem to cut this really interesting line between game design, game critique, and social commentary.
I imagine that a lot of folks here might not like their takes, as they tend to bend towards a social justice framing of games as media. (E.g. a recent podcast discussed, among other things, the differences between queer representation in games manifested as 'I can form a queer relationship with my character' vs 'there are queer people in this world.'. Another, not so long ago, discussed games with a specifically decolonialist bent.)
Plenty of discussion about unions in the games industry, and how developers handle bad working conditions as well.
So much games media these days seems to be about reporting on what press releases the games companies ship, or try to strike an inoffensive tone. Or even try to push the idea that games aren't/shouldn't be political (which is a pretty silly idea, imo.) It's nice to have an outlet really trying to dig into stories and present them with a deliberate lens and framing.
I definitely appreciate any journalism that centers the people behind the things that make up our reality. This article is definitely a terrific example of what I like about this approach to writing.
>For years, the community settled for a rushed localization that frequently crashed and was full of memes. (You will not be surprised to learn this version of the game was made by a high schooler).
Sadly that is not exclusive to high schoolers, plenty of well known localization companies (for example NISA) do a horrible job(Ys8 is such a clusterfuck) and insert plenty of memes(Danganronpa3), and worse - change the script for no reason.
EDIT:
I am not complaining about localization - but as a translator your role is to convey the thought and style behind it. Not just one or other, but both.
10 comments
[ 2.9 ms ] story [ 28.1 ms ] threadI can see an argument against it that this affects the "purity" of the game compared with the translation being done in the ROM itself, but I can also see a counter-argument that running the original unaltered ROM with what amounts to subtitles is a "purer" approach.
Edit: Just to be clear, I'm talking about a fan translation which is overlaid on the game by pattern matching the image of the equivalent original text on-screen, not an attempt at automatically detecting and machine translating all dialogue.
And games don't read very well with machine translation, so you'd still have to teach the system to overlay the correct English localization that was hand made.
All that might still have been easier than dealing with spaghetti assembly, but then you are faced with the daunting task of getting someone to actually download your executable. Patching a ROM is pretty well established process. Downloading an extra executable just to run a ROM in your native language is much less so.
The two books I suggest are…
• Legends of Localization Book 1: The Legend of Zelda
• Legends of Localization Book 2: EarthBound
[To clarify, when I say "Earthbound", I refer to the game that was released in Japan as Mother 2, with the translated release being called "Earthbound" in the US.]
If anyone wants to make the argument that a real-time translation would work, then I have an additional recommendation:
• press start to translate: This is what happens when you let a computer translate a video game?
You can find all of the above books at https://www.fangamer.com/collections/books/partner_legends-o...
I have technical concerns and non-technical concerns with the idea.
With Zelda 1 that would indeed work; there was not much text in the first place, and the typeface was blocky and easy to detect.
With Earthbound, things are much more complex. Ignoring text that is "baked in" to the graphics, the dialogue was much more interactive. The text was smaller, with thinner letter shapes, and when one of your characters was dead, the contrast between text and background was alot lower. You also have to deal with changes in coloration and highlighting, for example when you are selecting from a range of options.
Earthbound also had another trick: The dialogue system was not static text. Instead, the game effectively became an interpreter: The game didn't open a dialogue window, the script itself included the code to open a dialogue window, which the game would parse and execute. Check out http://media.earthboundcentral.com/2011/04/a-look-at-the-mot... for an example
So, if you were to make a mapping of "untranslated text" to "translated text", you would have to work out every possible expansion of the interpreted script, translating each one individually.
If your text ends up being longer than the Japanese text, you'd have to work out a way to deal with that. Either making your overlay bigger than the original text, or maybe intercepting a controller button-press to advance the text.
Of course, all of those are technical concerns. But I think there is at least one big non-technical concern: At least for me, it would be too jarring.
It's the difference between a "lean-back" experience and a "lean-in" experience. The lean-back experience is something like watching a movie. Translating a movie via subtitles means expending additional "brain cycles" to line up the subtitles with the speaker, but I'm not doing much else, so that's OK. Translating a movie via dialogue kindof looks weird, because the voices do not line up with the mouth movements.
Playing a video game is a "lean-in" experience. I am doing a lot more work, ingesting and processing dialogue and NPC actions, keeping track of information, making decisions, and taking actions myself. Translating the dialogue is possible, because you can "line up the mouth movements" (update the programming) to work with the translated text. If there is something else—an overlay, say, or a separate translation file—that takes up a lot more cycles, makes it harder for me to keep up with the game, and lessens the experience.
Of course, there are exceptions! A game like Tetris doesn't qualify either way. A game like Doki Doki Literature Club is more like a movie, and so subtitles fit.
But, which emulator?
People play SNES games on Windows PC's, Macs, Android phones, Raspberry Pi's, 3DS's, etc etc. And the emulators in use have shifted: ZSNES was the emulator of choice a decade ago, but using that today would be ill-advised.
Coincidentally, there exist some old translation patches that were only ever tested against ZSNES as opposed to real hardware. Playing them today is a PITA.
There was an Ask HN some time back that asked "How would you write software to last 100 years?" My answer—which I very much stand by—was: "Target the NES".
I imagine that a lot of folks here might not like their takes, as they tend to bend towards a social justice framing of games as media. (E.g. a recent podcast discussed, among other things, the differences between queer representation in games manifested as 'I can form a queer relationship with my character' vs 'there are queer people in this world.'. Another, not so long ago, discussed games with a specifically decolonialist bent.)
Plenty of discussion about unions in the games industry, and how developers handle bad working conditions as well.
So much games media these days seems to be about reporting on what press releases the games companies ship, or try to strike an inoffensive tone. Or even try to push the idea that games aren't/shouldn't be political (which is a pretty silly idea, imo.) It's nice to have an outlet really trying to dig into stories and present them with a deliberate lens and framing.
Sadly that is not exclusive to high schoolers, plenty of well known localization companies (for example NISA) do a horrible job(Ys8 is such a clusterfuck) and insert plenty of memes(Danganronpa3), and worse - change the script for no reason.
EDIT: I am not complaining about localization - but as a translator your role is to convey the thought and style behind it. Not just one or other, but both.
It is hard. very hard - but doable.