A Conversation With Sega’s Localization Team On Censorship, Consistency, And Comedy

Earlier in 2019, we got a chance to sit down with some members of Sega’s fantastic localization team. Sam Mullen, localization director at Sega/Atlus, and localization producer Andrew Davis joined us for a roundtable conversation about the ins and outs of localizing games like Yakuza and Persona. Rather than a traditional interview, Mullen and Davis mostly talked about how these games end up on U.S. shores, so we’ve presented the subject matter and posted what the two localizers talked about.

On localizing content that feels out-of-place or problematic in a different region:

Sam Mullen: I would say that’s not an uncommon thing that happens and it’s one of those really tricky things for us because, if we’re too heavy-handed – like, let’s say we do choose to do nothing. Our staff is really tuned into that kind of sentiment and all the stuff that goes around it, something we definitely think about a lot and talk about a lot. We look at it, we constantly engage in Japanese content and we understand that the things that different cultures react to is just different and we’re very hyper-sensitive about that. We do nothing, though, and it’s like walking into a wall and going “Well, people aren’t going to like that.”

On the other hand, if we’re too heavy-handed, we get dinged with this whole censorship kind of thing. There’s a certain consistent base that will be like “Hey you should just tell us what the–you know, don’t mess with the content.” And that’s if we’re really heavy-handed and aren’t sensitive to the actual true localization process, which is what we are trying to keep it to what it originally was, we’re trying not to change it too much. But there is a space in-between where we can kind of like “slight touch” and make it so certain things don’t pop out quite as hard.

It goes back to, like I was saying on the [PAX West] panel, our intention is to have users react emotionally in the same way that the original creator intended. So if the creator was not intending the user to have some kind of visceral reaction to this content, but if we leave it as-is and they’ll have a reaction to it, well then we’re kinda not really doing our job. That’s the whole thing, find this perfect space in-between where all parties are being addressed. Now, there are some extreme cases where you can’t make everyone happy and at the end of the day you have to just say this is where it’s going to fall. It’s a little bit of hoping that people understand, but that’s not really an excuse that can be made, anyway. We do have conversations with our dev teams, we do make cases and we do have open conversations, but sometimes creators have a hard line on something.

Andrew Davis: But often development can be a global collaborative process, like developers in Japan will listen to our feedback and sometimes they will say “We’re in the middle of design for this, do you think this will be something that could be appropriate globally?” And sometimes we’ll come back with our comments and say “Well, this is how this will be seen in this culture,” whether it’s American culture or whatever. Like I’ve had to say, “In Germany, you have to understand this would be very sensitive.” So we can go back to the dev teams and sometimes we do have the opportunity to have a dialogue and it’s always interesting.

Mullen: I remember recently with [Yakuza] Kiwami this time last year, there’s a character, one of the two hostesses you can find of have a relationship with. One of them identifies as a lesbian. I remember having a lot of discussion with Scott [Strichart, Yakuza localization producer] about how to approach that storyline. The way her story results in Japanese, if translated straight, comes off as that like “Well you just never met a real man” kind of feel to it. But I don’t think that’s the intention they were going for, that’s just the way it comes across in English. So I do remember there being this light feeling of needing to change the language, because that’s not the point here.

That entire story arc is actually really interesting, because it gives people an interaction with a type of character that isn’t represented much elsewhere, but there are definitely some complicated things here. I would say that the Ryu Ga Gotoku [Yakuza] Studio guys are actually currently very hyper-aware of that. Some of the stuff that we haven’t announced or confirmed for the west, like the remasters of 3, 4, and 5, in the Asian and Japanese markets, they did make some content adjustments for that. [Yakuza producer] Nagoshi-san himself has spoken about how back in 2009, attitudes to certain things were different back then. They recognize it and are making adjustments to the content for the remasters.

They are listening and it’s kind of our job on the localization side to make sure that all the overseas talk that is in our regions and our languages is being communicated. That’s part of the localization job, it’s not just text.

On the subject of how much overseas fanbases matter to development decisions:

Mullen: One thing I do want to clarify, because I do see this confusing a lot in the media. When [Nagoshi] says “overseas,” people always interpret that to mean the west, and that’s not what he means. Specifically speaking, he’s talking about Asia as the primary overseas market, he’s not talking about us. We’re included in “overseas,” but right now Japan has a huge push to try and get more uptake in the Asian market, which is why they’re releasing their games with usually Chinese and Korean at the same time. Right now, with Sega Japan, they’re trying to get in that market that’s opening up. You know how markets are, whoever’s there on the ground first is the strongest presence. Those guys are really focused on the overseas market, so when they say that, they’re talking about opening new Yakuza audiences, which of course the west is part of that, but don’t forget Asia. Asia is a huge part of that push as well.

On the subject of the recent resurgence of Japanese-developed games:

Mullen: It’s kind of complicated. There’s a couple of things going on. For one, in the early-2000s, you have to remember that Japan owned everything then. It was like, all the content was basically Japanese, but when the western developers started coming on the scene, Japan saw the uptake kind of starting to decline and that’s really scary because that’s a contraction to them. So they’re thinking, “Why is this happening? Oh it must be that these western developers and these styles of games are taking our business.” And the natural reaction is to after that and run to that, but ultimately you’re trying to sell a product to someone – this product is appealing to cultural underpinnings of what we like in narrative, what we like in storytelling, what we like in settings, and whatnot. It’s very challenging for someone from outside a culture to go in and try to get at that. It’s really hard, it would be like a western developer really trying to make a JRPG that plucks at the heartstrings of a Japanese person who cares nothing about overseas content.

When Japan started doing that, you can kinda tell like this is Japan trying to run after that, not being very organic or natural. For a while, there was the whole [Keiji] Inafune thing and everyone just being down on their luck, and finally you just had other people being like “Man, F that. I’m just going to make the game I want to make. I don’t want to make the game I think other people want to play, I want to make the game I want to play.” And they start making those games.

At the same time that happened, we have the resurgence of accessibility of Japanese content. Back in 2004, if you wanted to access Japanese content, you’d be downloading mpegs off of shady bittorrent sites. Before that, you’re trading VHSes at your anime club, so the access was really hard, right? But with the advent of like Crunchyroll and more easily accessible manga sites, the savviness of users to get Japanese that in a good time – there’s an entire visual novel market now, Danganronpa is on Steam of all places. What it is is that, because these people have such easy access to it at such an early age, understanding the mindset of foreign characters in pop culture is way easier.

It’s still like, you can only experience Japanese culture to a certain extent by consuming the pop culture. You can’t really consume pop culture and be like “I know what it’s like to be a Japanese person.” That’d be a pretty ignorant statement. But you can still draw the lines between different types of pop culture, like the trends and whatnot, so when you step into a game that kind of needs you to understand what Japanese humor sounds like, what the reference, what the pacing is going to be, like Yakuza, you’re already primed with that know-how. So if I drop like “Kiryuu-chan,” you already know, like, okay -chan, that’s a thing that’s a term of endearment, I get that. And that’s kind of common knowledge now. But if you asked a young pop culture consumer in 2004 what’s different between -san, -chan, -kun, and -sama, they’d be like “What the f–what is that?” But now I bet you most people could probably at least take a crack at it, right? So there’s just a cultural kind of intake where people are just a little bit more primed to receive the quirky stuff that Japanese people are making for themselves in a way that was just not there 15 years ago.

On localizing comedy and finding funny translators:

Mullen: The translators on our team I would say are actually pretty serious people. They’re very straight-laced, but that’s what we need. What their job is is to literally read a line and go “Okay. He’s making a joke here. This joke doesn’t make any sense so someone’s going to have to rewrite this.”

Davis: They will literally write that as a comment in the file. “Oh this is a pun. The pun is this. Go nuts.”

Mullen: But they hand it over to the editors and our editing team, some of them are slightly bilingual or have experience speaking Japanese, but they’re not like Japanese speakers or like bilinguals by any stretch. They’re usually writers, people with journalistic backgrounds, people who are just wordsmiths, or funny and interesting people. So when they’re getting this Japanese texts, just like a copywriter, you give them a report and it’s their job to go “Okay well I’m going to make this not sound stupid.”

So they come in and and they make all this dry, translated text, some of it is already funny because the original intention was set up to be funny, but they come in and they make it hot. They find and insert all the funniness, they find the stuff to go in there and go “Oh we can make this more serious or I’m gonna make you cry, so I’m going to write this really cry-worthy stuff right here.” So they play that up. That’s the way our company does this.

Davis: One thing we were experimenting with is different editors taking ownership of certain characters. I came on to Valkyria Revolution on the tail end of that project, but I heard that there was an experiment done where the editing team on that one sort of took longer shifts on certain characters and tried to fit their voice.

Mullen: And in that game especially, if you really pay attention…Valkyria Revolution as a game aside, the localization itself I felt the team did a really good job on it. If you really look at it, each character does have a very unique voice to them, which is very different from other characters, because literally a different person was writing the other characters. Now functionally that’s actually really difficult because the game scripts are not set up that way.

Davis: There’s conversations.

Mullen: Yeah, there’s conversations, and having three different editors comb through a single file three times, because they’re having to like–someone wrote this character to be interesting, but then the next person comes through and they’re going “I am going to write this line this way so it feeds into my line a little bit better” and it creates a lot of complexity. Ultimately, I don’t think that was necessarily an effective way to do it, it just depends on the project and the scope. But there are certain people who have different skills, like we have people who are better at straight-laced writing, we have funny pun people, we have guys who are tearjerker-type writers, so we have different people for different content, and we try to pair them up with the content that works for them. Going back to Yakuza, Majima’s way of speaking, that’s Scott Strichart’s thing that he made, he built that. It’s actually kind of crazy difficult for us because we can’t just have one of our producers be the only person who can write for this character. That’s like a risk, right? He could be hit by a car one day and then suddenly Majima’s dead in the localization, we have to write him out of the series because we can’t nail his voice anymore.

In Japanese, the way he speaks is somewhat formulaic, the way he crafts and constructs his lines, and you think that he just needs to say these words in this style, but English is not that simple. The person who’s writing has to have access to all kinds of idioms, and turns of speech, and creative ways of writing. It’s a little bit more complicated than that.

Davis: Japanese in particular does have a lot of opportunities for speech styles that are recognizable, like “gimmicks” in Japanese speech that don’t come off as forced as “gimmicks” in English speech.

Mullen: Like having characters that end their lines with “piko” or something like that. Everything they’re saying ends with piko, you made them say that. Like you can change the way Japanese sounds and say “This character sounds like this.”

Davis: Or even the character Kai in Valkyria Chronicles 4, she actually uses masculine speech in her dialogue, which is not something you really…there isn’t like masculine and feminine speech in English, like not really distinguishable. So we have to come up with other ways to get that across that she is blunt, cool, you know, it sort of diffuses into the larger editing rather than a simple grammatical thing.

Mullen: One of the things you actually have to do in Japanese characters, like if you have a character in game or an anime or whatever that has like multiple characters that interact with each other, they actually create these charts that say this character when speaking to this person uses these types of pronouns. When they refer to themselves, they use these types of pronouns, because it changes depending on the person they’re talking to, and you actually have to map that out so it’s consistent.

On writing a consistent voice for games with contrasting styles:

Mullen: One of the things we constantly have to worry about is consistency. Yakuza 5 is like 2.5 million JPCs [Japanese characters], Persona 5 was 50 percent bigger than that. In order to even attempt to get it done on time, so we didn’t have to release the game two years after the Japanese release, we had to start translating mid-development. But text is being churned through, they’re making changes so stuff that they we were already done with was getting ripped out and thrown away. And then like, oh s—, if you change that then this reference I wrote in doesn’t make sense now and now I gotta go back and [frustrated groan].

Keeping up with it is a big mess and keeping character consistency is one of the biggest challenges, so our localization team spends a lot of time doing what we call pre-production, which is where they map all the characters, they say this is a sample of their speech, this is what they sound like, the kind of stuff they say, they don’t say things like this, they don’t have this kind of attitude, but they do these things, and they say things like this. With that understanding, the editing team tries to walk toward a single goal, but just by virtue of two different people writing for the same character, they’re going to sound different at different parts of the game. Ideally, someone comes through and does a whole pass over the text and says it’s all fitting in, but sometimes that’s just like not possible.

Davis: Not with three million JPCs.

Mullen: If you’re an incredible editor and you can move through fifteen thousand JPCs a day, which is like 7500 words–

Davis: That’s basically proofreading at a publisher, which, I’ve done that, but…

Mullen: But if you’re moving at that speed, it’s like, what’s the math on that? So a single person going through a game that’s like three million JPCs, let’s work that out. How long would it take them to do that?

Davis: Even at 15,000?

Mullen: Yeah, which is already breakneck speed. Like they do nothing other than that, no email, no nothing, no coordination. [Mullen uses his phone’s calculator] It would take them 230 working days to do that. So, working weeks…46 weeks. That’s from the point that it stopped being translated, because if it’s still being translated, you might end up with other inconsistencies there. The translation team, as they’re translating part of the game, they may have to go back to the beginning part and be like “Ah, now I understand this.” Consistency across characterization and translation is definitely a huge challenge for everyone.


2018 saw the release of Yakuza 6 and Yakuza Kiwami 2, as well as PC releases of Yakuza 0 and an announced release of Yakuza Kiwami. Valkyria Chronicles 4 was released earlier this year, as well as a Switch port of the first Valkyria Chronicles. Further projects in the Persona 5 universe have also been announced.

Content source