• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

XSEED gives a glimpse at how their localization process works.

slap me if old.

http://kotaku.com/getting-jrpgs-out-in-english-is-harder-than-you-think-1441094168


original.png

original.png

original.png



more at the link.

doing games translation work seems a magnitude harder than subbing movies/tv show work i've done before.


Also get Rune Factory 4.
 
From what I also hear, it also sometimes involves regularly sending back what you've worked on to the original programmers, so they can program in the text and send the build back to you for playtesting.

It al sounds complex, and I find it endlessly fascinating.
 
I've wondered, do they sort of have to do a "back and forth" sending their Excel files to the original developers, who then put the English text into the game? I remember reading that the localizers don't do the actual programming.
 
Did they make an article out of the tumblr that XSEED employee made last summer? Always nice to see a bit more attention to the process so people can learn how it works.
 
Yep, the normal process is typically Japanese > direct translation > editor clean up. That's the process I use for fan translations too.


Also, 'sending back stuff' to the JP programmers is a recent thing. Working Designs did all that in-house. That was why it often took them a while, but they always got it done. A Grand Knights History debacle would've never happened if XSeed had programmers on staff.
 
Line-by-line, in random order... why? It would produce the lowest quality output possible with such loss of context.

Put them all in order in a database or something instead and have a direct link instead of all this data entry. Yuck!
 
Line-by-line, in random order... why? It would produce the lowest quality output possible with such loss of context.

Put them all in order in a database or something instead and have a direct link instead of all this data entry. Yuck!

Gonna take a shot in the dark and wager they extracted the script and then just slapped it all into Excel. Not organized, but quick.
 
That tabs, that spreadsheet.
Everybody who made FEA have excellent dialogue or even localization team that managed to do decent job, I hereby all of you get my respect for that.
 
Line-by-line, in random order... why? It would produce the lowest quality output possible with such loss of context.

Put them all in order in a database or something instead and have a direct link instead of all this data entry. Yuck!
It is possible to order all story stuff, but how do you handle stuff that can appear in random order in the game depending on the player choices? Item names? Menu options? Stuff might also appear in a different order depending on language context.
 
That tabs, that spreadsheet.
Everybody who made FEA have excellent dialogue or even localization team that managed to do decent job, I hereby all of you get my respect for that.

Oh yes, after getting acquainted with Excel in engineering coursework, and learning the hard way about losing work due to crashes, I would be paranoid if I had to deal with huge Excel files like the ones used during localizations, I'd be saving every minute, making backups of backups of backups, and I'd still feel nervous.
 
Line-by-line, in random order... why? It would produce the lowest quality output possible with such loss of context.

Put them all in order in a database or something instead and have a direct link instead of all this data entry. Yuck!
Not sure I understand exactly what you're envisioning, but generally speaking anything that requires your freelancers scattered around the world to (a) spend time learning new software or (b) have reliable fast internet connections is going to be impractical. Game text volume can be large-ish in terms of sheer word count sometimes, but it's not typically complex to a degree that spreadsheets aren't sufficient as databases for it, and your clients themselves typically supply the text to you in spreadsheet form and want it back in those same files. (Sometimes it's merely output to spreadsheet for the sake of the localization agency, but often the spreadsheets are the exact files the developers themselves are using to make the game in the first place).

As for the line order, it's nice when the client gives you the lines in order. It's just that they simply don't always do that, and they're the boss, so you can't really gripe at them too much. (Lots of stuff doesn't really have an order anyway; you just need to know the context for how it's used, e.g. situation-dependent battle one-liners.)

I mean, usually of course story-event lines are in order. It's just, you also have cases where, say, it's not entirely clear from just the script file itself whether lines 1-39 are all one cutscene, or 1-36 are the cutscene and 37-39 are the lines the NPC repeats if you talk to them again after the cutscene. Things like that may require checking in-game if the client didn't provide clear context descriptions/labels in the script files. Some projects have more scene description/context text in their files than others.
 
Not always. Gaijinworks did Class of Heroes 2's in house.

But isn't Gaijinworks/Working Design more like the exception than the norm? I remember back in the PS1 days Enix couldn't localize Dragon Quest IV on PS1 since the original team wasn't around anymore. Xseed does have programmer for PC
 
That look like pretty much the same process I used when I was localizing games. One developer can be on the ball and provide updates quickly to speed the localization process, the other can be stupidly slow and make it take forever. Seeing the changing in-game helps ALOT.
 
But isn't Gaijinworks/Working Design more like the exception than the norm? I remember back in the PS1 days Enix couldn't localize Dragon Quest IV on PS1 since the original team wasn't around anymore. Xseed does have programmer for PC

I think they're the ONLY team that does localization programming. Atlus, NISA, and XSeed all are subject to the developers schedule.
 
People also forget that in order to see all the localization changes the localization team needs to play through the game entirely, or at least be able to have access to virtually the whole game. It's a VERY systematic approach going through every line of text and making sure it's right both in translation and in execution.
 
Not sure I understand exactly what you're envisioning, but generally speaking anything that requires your freelancers scattered around the world to (a) spend time learning new software or (b) have reliable fast internet connections is going to be impractical. Game text volume can be large-ish in terms of sheer word count sometimes, but it's not typically complex to a degree that spreadsheets aren't sufficient as databases for it, and your clients themselves typically supply the text to you in spreadsheet form and want it back in those same files. (Sometimes it's merely output to spreadsheet for the sake of the localization agency, but often the spreadsheets are the exact files the developers themselves are using to make the game in the first place).

As for the line order, it's nice when the client gives you the lines in order. It's just that they simply don't always do that, and they're the boss, so you can't really gripe at them too much. (Lots of stuff doesn't really have an order anyway; you just need to know the context for how it's used, e.g. situation-dependent battle one-liners.)

I mean, usually of course story-event lines are in order. It's just, you also have cases where, say, it's not entirely clear from just the script file itself whether lines 1-39 are all one cutscene, or 1-36 are the cutscene and 37-39 are the lines the NPC repeats if you talk to them again after the cutscene. Things like that may require checking in-game if the client didn't provide clear context descriptions/labels in the script files. Some projects have more scene description/context text in their files than others.

Considering all that, I'd do it with a secure web site with a database back end. Just starting with a simple structure like [String_ID] [Text_JP] [Text_EN] and you can add columns with comments for scene grouping, field lengths, etc. Optionally, you can set up your game to pull from the appropriate localization when you do your build, but that requires planning for that from the beginning.

If bandwidth is a problem, you could set up some file-based export/import methods, which is pretty much what they're doing with this file.
 
That was a very interesting read. Wouldn't it be nice to have a dev environment where you can modify the script and have it update/reload in game every time you save? Since games won't be writing anything to the script itself, this should be possible. Hopefully, as game engines mature further, this will become a thing.
 
Was expecting some new Senran Kagura screenshots :/

Nice write-up, tought. Hatsuu already posted some titbits about Xseed's localization process on her tumblr, it's nice to have some screenshots to better illustrate what it's about.
 
Good article. This type of stuff is what I find super interesting, particularly since I started to study Japanese. The differences between English and Japanese are so bloody vast that I feel bad for people who have to do live translations (i.e. press conferences).
 
Also, 'sending back stuff' to the JP programmers is a recent thing. Working Designs did all that in-house.
No, I'm pretty sure WD was always the exception there.

EDIT: If anything what's recent is that they actually talk about localization, but they'd been more and more open about that since the late 90s, which can either be attributed to WD or just gaming growing bigger and localization being more than getting some sap in Japan to translate the few pages of text there are and hope it sounds natural enough.
 
I saw this article at Twitter, but thanks for posting it here nonetheless. I have a lot of respect to people who work as translator/do translation work, and XSEED is one of the reasons I would like to work at localization when I get older someday. This article definitely gives a great inside look at what it's like to translate games professionally. Thanks a lot for showing all this stuff, Hatsuu/Brittany!
 
I remember following the blog for the Mother 3 fan translation and them detailing the entire process of translating and fixing boxes etc. It was fascinating stuff. Always fun to see the inside workings of these things.
 
Lotta work (and, therefore, money) for not a lot of benefit, I think.

Well, it all depends on the economy of scale. Not worth it for a single game, sure, but if a studio is doing a bunch of games (or, taking that a step further, multiple studios use a common central site) then maybe that could be worth it.

I'm also thinking of things like concurrent updates with multiple teams working on the same project. You get version problems with Excel files being passed around, where with the database approach you can do things like having QA following along checking translations shortly after the first pass, tracking that review status within the same system, reporting progress and so on.

It would also be handy for opening things up for crowdsourced translation after release.
 
Well, it all depends on the economy of scale. Not worth it for a single game, sure, but if a studio is doing a bunch of games (or, taking that a step further, multiple studios use a common central site) then maybe that could be worth it.

You'd need the spare budget to actually develop it (and to account for lost productivity in the time it takes to get people up to speed on it), and even getting to the point where that was plausible would probably require you to be an internal localization team only -- there's no way you'd ever get multiple external teams on board for this sort of thing when most Japanese games are still being developed with the assumption that they'll never be localized.
 
All my respect goes to the people translating those huge JRPGs! I can definitely imagine how much time consuming it must be to check every little hidden corner of a game with every possible combination, just to make sure you didn't miss that line that only happens if it's October 3 and you have armor A and sword S and the blue thingy that you could only get if you have so and so in your party.

Having said that, I'm also surprised at how archaic their process is. I mean, there would be a million ways to do this better than with swapping Excel files, of all things. God, I don't want to imagine the versioning issues they run into!

I can understand when I see people swapping Excel files for things like scientific studies. Software is not their field. But I'm surprised a software company is still organized like it was 1992.

My guess is, as with all things software, the problem is of human nature. For example, I wouldn't be surprised if smaller software companies in Japan write their games with the Japanese market in mind only, in contrast to a company that would write their code with the express intent of making it internationalization friendly.

It looks like an interesting challenge though! If we could take a look at lost time due to human errors, lost files, version confusion, and just plain Office licenses, I'm sure companies that do translation could save a lot of money with the appropriate software.
 
For example, I wouldn't be surprised if smaller software companies in Japan write their games with the Japanese market in mind only, in contrast to a company that would write their code with the express intent of making it internationalization friendly.

This happens much more often than it should, in fact a lot of the bigger Japanese companies still do this.
 
This happens much more often than it should, in fact a lot of the bigger Japanese companies still do this.

Yeah, the most disappointing aspect of this article was how it reconfirmed just how badly Japanese games are prepared for localization. We would probably see a lot more titles brought over, and more profitable niche localization companies, if these games were built with alternate languages in mind from day one.
 
This happens much more often than it should, in fact a lot of the bigger Japanese companies still do this.
And frankly if your games reliably get localized it seems kind of insane to not make the process smoother. It's one thing if this were the first Rune Factory, but this is the sixth brought over. I can understand games like 7th Dragon that were new IPs and likely had enough hurdles for localization to write it off early on, or something like Mother 3 that was so late in the GBA's life that honestly it was a bit late to bring over anyway, but doing it in RF4 is just... what the hell?
 
I am assuming something like Pokemon X/Y would has a easier process? It is getting 7 languages with a worldwide release.
 
I am assuming something like Pokemon X/Y would has a easier process? It is getting 7 languages with a worldwide release.
Nintendo's probably smart enough for many of their bigger titles (and possibly even some of their smaller, look how Xenoblade successfully held English and Japanese VA tracks) to make games easy to release internationally. Not that they necessarily apply this to all games as the Mother 3 fantranslation blog indicated, but for Mario, Zelda, and Pokemon? Without a doubt.
 
Yeah, the most disappointing aspect of this article was how it reconfirmed just how badly Japanese games are prepared for localization. We would probably see a lot more titles brought over, and more profitable niche localization companies, if these games were built with alternate languages in mind from day one.

The fun really begins when they use 'double dipping' lines for multiple situations, which works fine in Japanese, but doesn't fly in English at all.

Many games also have incredibly unwieldy file set-ups. For example, instead of say, one main dialogue file and a few separate NPC files/system files, you'll get 400+ files of a bunch of different things thrown together (dialogue, system text, random programming codes, etc.). It can be a real... joy. :)
 
Off-topic, but could someone create a new thread for this? I would I do if I could, but I'm still not a full member. All this Falcom music...
71t0N72.gif
 
I've wondered, do they sort of have to do a "back and forth" sending their Excel files to the original developers, who then put the English text into the game? I remember reading that the localizers don't do the actual programming.

It doesnt involve programmers copy pasting everything from those excel files, its much easier, they just export the needed date into xml files from these excel sheets which a programmer can import. in code they simply refers to the correct text that's in another file. there have been cases in games (often in betas) where suddenly the text is replaced with stuff like EN_menu_OptionVolumeButton, or stuff like that, and thats the text field being unable to access the proper text.
 
The fun really begins when they use 'double dipping' lines for multiple situations, which works fine in Japanese, but doesn't fly in English at all.

Many games also have incredibly unwieldy file set-ups. For example, instead of say, one main dialogue file and a few separate NPC files/system files, you'll get 400+ files of a bunch of different things thrown together (dialogue, system text, random programming codes, etc.). It can be a real... joy. :)

that sounds awful!
Do you ever have to go deep into the game's code in order to accommodate the necessary changes?

edit: another thing I always wanted to know about: creating alternate textures and graphics when needed - do you have an in-house artist do it or do you have to like sketch up a rough draft and send it back to the developer?
 
It doesnt involve programmers copy pasting everything from those excel files, its much easier, they just export the needed date into xml files from these excel sheets which a programmer can import. in code they simply refers to the correct text that's in another file. there have been cases in games (often in betas) where suddenly the text is replaced with stuff like EN_menu_OptionVolumeButton, or stuff like that, and thats the text field being unable to access the proper text.

Sometimes they do. :) We had a case where new text bugs kept cropping up in the Eng builds because they were doing just that and sometimes accidentally cutting stuff off or inputting individual text changes by hand which resulted in terrible typos. Every team is different and have their own methods that we roll with accordingly.


that sounds awful!
Do you ever have to go deep into the game's code in order to accommodate the necessary changes?

edit: another thing I always wanted to know about: creating alternate textures and graphics when needed - do you have an in-house artist do it or do you have to like sketch up a rough draft and send it back to the developer?

We've had to adjust things slightly before, but usually to remove code (like replacement codes) in order to fix stuff like "the Christmas" to "Christmas," but generally we don't mess with them as it could add programming bugs.

For graphics, usually the development team has an artist that makes those changes we suggest in the Graphic files, but there has been occasion where what we want is a burden on the dev team and so we did it in-house. All the little elemental graphics were completely redone by Tom in Unchained Blades, by way of example. And in our STEAM releases, Hatsuu does all the graphics by herself. :) Our PR guy is great at photoshop too and he's made adjustments to logos for us before as well.

I keep saying "we" and "our" still... :(
 
As a fellow J<->E translator (in finance and academia, where the number of characters in a line or the width of a text box is almost never an issue), I love seeing this stuff. It's heartening to see that as time has gone by, developers and programmers have become more attentive to the needs of translators.

And Hiro, the dreaded "double dip" is the one thing I wish all original writers would be cognizant of. What a headache those can turn out to be.
 
really curious how they can do a worldwide release of pokemon in idk how many languages. Even though pokemon is light on story, its got tons of shit.

would love a iwata ask for treehouse.
 
really curious how they can do a worldwide release of pokemon in idk how many languages. Even though pokemon is light on story, its got tons of shit.

would love a iwata ask for treehouse.

It's probably a lot easier to translate as you build your game, meaning each time you write text you just do it as many times as you need to for each language. It lets you preserve context easier rather than de-compiling text trying to figure out what goes where and all the jokes and stuff then putting it back together.
 
It's probably a lot easier to translate as you build your game, meaning each to you write text you just do it as many times as you need to for each language. It lets you preserve context easier rather than de-compiling text trying to figure out what goes where and all the jokes and stuff then putting it back together.

seems so. curious how much harder it is for devs to do it like that. Like for rune factory, with every entry localized, you'd think preping it for localization while making it would be better for everyone.
 
Top Bottom