Making Games Classics: Localizing Dragon Age: Inquisition

Dragon Age

BioWare puts much time and effort into the localization process. Melanie Fleming explains how the games are transformed into tailored experiences for BioWare’s non-English speaking audience.

Disclaimer: This article was first published in issue 06/2014, which was released in October 2014.

Dragon Age: Inquisition is BioWare Edmonton’s most ambitious and biggest title so far. We have close to one million words of dialogue, codices, operations, and missions for players to explore in the game. To ship localized versions of our games to our high quality standards, we needed good, efficient processes and excellent voice-over and translation teams that would be willing to take that volume of work without breaking down or dropping their quality of work. The EA audio capture team, based in Madrid, was managing the day-to-day with our external partners.Our localization quality assurance team, also primarily at EA’s Madrid office, weigh in on the casting of our actors as well, picking their favorite voices and giving review and feedback on the translations and performance of the voice-over (VO).

Localization testing includes checking the translation of the words and the style and quality of the voice over, the codices, the appropriateness of the art pieces and story themes, and the stylistic consistency of the user interface, the operations and subplots. As it is such a big game, and since BioWare games branch a lot and have a lot of choices and paths the player can take, that meant we needed a lot of eyes on each language to make sure we covered the game well for testing. To give you a sense of scale: At our peak, we had over 50 testers across all the languages on Dragon Age: Inquisition.

Localization is a global affair

This sounds like an obvious, no-brain statement, but the implications are as follows: if you work in the field, much of what you will do in Localization is talking with other people who are working remotely or at least very far away from you across the world, often from a different culture, often in another time zone, often bravely speaking to you in English as their second or even third language, and sometimes with completely different goals or viewpoints than you. There can be gaps in understanding, vast chasms even, so the connection and understanding between people is the most important thing. We make extra effort to develop the trust and the quality of communication between our team members.

We take trips to Madrid to meet and understand our team members there. We tour the VO studios in Germany and France, and we talk on the phone directly and regularly with our translators. We very much care about what is happening with all of our people, and we take an interest in the specific set of circumstances a certain region or language may be facing. We try and understand it, and listen, and we do not dismiss. So rather than feeling burdened by the differences in culture and language, we celebrate it, rejoice in it, sharing stories of our own cultures and experiences with each other to find that common ground.

Obviously, there is an effort and a cost to running a team that way – it is expensive to fly all over the place, and it takes years to build relationships and trust between teams that may not start out understanding what the other one even does. But we find this effort absolutely essential to the creation of our games. We aim to keep raising the high bar of quality with each game we do, learning more about other cultures and about what people want to see. We learn as we go.

The Localization team based in Madrid was managing everyday business. Upper row: Ricardo Santos, Philippe Charel, Daniel Mellar, Allan Smith, Jeremy Teyssier, Michael Kukawski, Roman Hannberger and Holger Hartmann. Lower row: Oleksandr Lebid, Olga Zhuravleva, Oriana Lapelosa, Pablo López, Melanie Fleming and Charles Ulbig.

Game Localization is a highly creative exercise

By this I mean that we create our stories for people, and when they play it, it represents a personal picture for them. It would be naïve of us to think that the picture is interpreted in exactly the same way by every person in every language we use. In addition, not all of our stories are easy to tell across cultural boundaries, and we are not known for shying away from sensitive topics or morally ambiguous decisions in our games. This creates some challenge, so we get both positive and negative reactions from fans over that. We try to listen to fans from everywhere around the world. It is indeed a highly creative process, and there is always a lot of background knowledge we need to gather. There are a lot of questions to ask.

For instance, words, gestures, colors, sounds or grunts, even the same piece of music can be subject to interpretation in different ways by different cultures. We need to catch those things and consider how they will be seen by different people, but when localizing the game and translating and creating the content, we don’t just look for what will offend or what will be taboo, and we certainly don’t just look at the words. It is a deeper process than that.

Our job in localizing a game ends up being very multi-layered because the stories are always very multi-layered. Our stories are expressed equally as much through everything you read, see, and hear in the game. Every flag or banner you see, every painting on the wall, every animal, every plotline and bit of gameplay has been considered from multiple points of view. It is not just a translation of words.

With all this in mind, it is our primary aim to get across the main points of character and story in a way each region will relate to, and love, and understand. Even with our careful study, some plots, themes, or characters can resonate very strongly in some regions, while we find other characters or concepts can fall flat on their faces. In addition, some characters need a little extra effort to be interpreted so they are inside the same general bubble of intent as was meant in the original language. A good example of this is the character Sera in Dragon Age: Inquisition.

The elf Sera posed a challenge during translation because he is extremely colloquial and referential in her speech.

Sera is a follower, which is a major role, and a central one. She is a powerful figure in the story of the Inquisition, but a challenge lies for us in localization because she is also extremely colloquial and referential in her speech. This makes her much more difficult to translate because her speech patterns and even her physical mannerisms draw upon not only modern Western references but also IP-specific inside jokes. To make matters more interesting, she has sometimes been written so she is purposely supposed to be difficult to understand, even for a native English speaker. So in the translation, she still needs to come across as »difficult to understand«, but in a very specific, understandable way, one that intrigues players without flat-out confusing them.

Some of Sera’s references can relate indirectly to Western pop culture or more modern sayings, some to past BioWare games, and some to deep, deep internal Dragon Age lore, so we had to find a way to capture that flavor and complexity and help it shine through the translation. We knew that with her, if we just threw Sera’s text over the wall to our translators with no other support and simply dusted off our hands, we knew the translation would not go so well. We had to figure out a method of deeper interpretation and get a better connection between our teams.

To do this, we arranged call after call where the translators could talk directly with Sera’s writer and ask questions. This direct contact was important. The method also helped with other characters, such as the references in many of the Iron Bull’s … rather … saltier sayings, as well as Cole’s completely random non-sequiturs and mysterious hints that are scattered throughout his dialogue. The creativity in interpretation of these characters must start with the translation team, so they are one of the most important connections to the original English meaning. They lay the groundwork for what we record later in our voice-over sessions with our actors, so we spend extra time and review on dialogue. If what the translators produce is not good, then the voice-over will automatically also not be as good, no matter how talented the actor.

Trust in the translators is important

Once the translators have achieved an understanding of the quirks and background of each follower and plot arc, they inevitably have a lot of questions. I mean a LOT of questions. So far on Dragon Age: Inquisition, we have currently tracked over 5,400 individual questions from all the translators together since the beginning of the project.

This is more than twice the number of questions the translation team asked during the entire development of »Mass Effect 3«. That is not only because Dragon Age: Inquisition is a much larger game, but we also intended to stimulate the increased number of questions for this project. »Mass Effect 3«’s process was okay, but for Dragon Age: Inquisition, we wanted a more direct and open discussion with our translators than we had ever had before.
On previous projects, it sometimes used to take five days or a week to turn around questions on batches of translation text, so we started to see that translators were simply not bothering to ask. It felt like too much of a hassle and too much paperwork for them. Now we have aimed to get the time to answer questions down to next-day or same-day, despite the time difference between our groups in Canada and Europe. We reduced the number of go-betweens and connected the asker and the answerer directly.

We want to make sure the translators are not blocked or discouraged from asking questions or from understanding a scene properly. Many of the questions are answered directly by the writer, level designer, or the cinematic designer who owns the plot or the character. We pass that knowledge to the translator, and then they use the native knowledge of their region and language to match the translation as best they can to the intended game design. They are like designers themselves in that they interpret the English story to each region, aiming to present it in the most appealing and tailored way they can.

That must start with the translators’ complete and total understanding of the story plus the background history of the franchise, so we spend a lot of time on sharing that when first starting production. This method, coupled with the constant and direct contact between the writers and translators, increases the trust and drive we have together to create a great game.

Choosing Voice-Over actors for our localized games

After translation, we move on to voice-over sessions to bring the characters to life. A build viewed before VO is added and after VO is added feels like night and day.

In casting actors to roles, we must be very careful because while there are a lot of very, very good actors out there, the challenge is sometimes in finding them and getting them connected in the right way to the project.

Each country has its own rules about hiring voice over talent. Some have very strong unions, and you obviously have to know how to work with people so they are happy. Some countries have certain actors that are attached or more affiliated with some recording studios or certain agents. It is easier to hire them through a particular studio if you are working with that studio. So the kind of audio capture we do involves a lot of very specific knowledge of industry norms and common practices.

Some countries have rules about the maximum number of roles you can have an actor play in each recording session, rules about how child actors should be hired and handled, rules about lines involving screaming or otherwise taxing an actor’s voice, or rules about explicit content or uncomfortable or sensitive themes. I could go on.

Those rules are there for a reason. We maintain that an actor always has to be okay with the character and the role and the themes they are playing because, number one, that’s moral and right to do to someone you hire, and number two, you will obviously also get a poor performance from them otherwise. They have to believe in the character they are playing and understand the intent of it.

Through all of that process and paperwork, we must focus on finding the actors that can take the translation of our game and turn it into believable followers and engaging plots. The balance of voices is important. We have to be mindful when casting — hire too many actors for the various soldiers, farmers, and townsfolk, and you will blow your budget so wide open that you will never, ever have enough money to make another game. But hire too few, and all you get in the game is … sameness. We do not want noticeable sameness with the voices. We also do not want too much campiness or too much flatness in the voices themselves. This is every voice-over producer’s dread and fear, since it takes so much effort and so many resources just to get to the point where we can even record.

The English voice-over had to be recorded first so it could be used as a r eference for other languages, in this case German.

We do have re-recordings sometimes, either if the text somehow changes or if the performance is not right for the scene, but we work really, really hard to get it right the first time. Doing this right in recording includes getting the English VO recorded first so we can use it as a reference when recording other languages. We also show cut-scenes to the directors and discuss plot points and character motivations with the studios. We share as much information as we can with them. We don’t want any noticeable drop in quality and we work hard to prevent it. We must keep the quality of recordings high, both technically and performance-wise. This is always in balance with the amount of time we have to record. It can be very push-and-pull that way.

Localization Producers need to play the game a lot

We play the game a lot and note where all the small characters are standing. We want to ensure that no two that are standing beside each other have similar voices. Each actor we hire has a purpose in the game, and their voices can be scattered through different levels. There are over 800 unique characters in Dragon Age: Inquisition, so working out the logistics of who plays what role matters a lot to us and takes a lot of effort.

Playing the game a lot is important to ensure that no two characters that are standing beside each other have similar voices.

Some actors can specialize in playing the ‘everyone’ voice, like a voice that is not too recognizably unique, so you can use this sort of actor over and over in various places in the game and yet not many people tend to notice it. Another personal style some actors may have is to perform with many different pitches, styles, or different characterizations, so that they can produce a few quite unique-sounding characters for us just by altering their voice slightly. As long as their voices don’t get too cartoonish, this particular type of actor can be extremely useful for small or medium-sized characters. We generally use these voices for farmhands, army recruits, servant elves, scouts, or what have you.

There are also audio or sound design tricks we can use to make things sound unique, like pitching voices slightly up or down, or adding other effects, such as on villains or special supernatural characters. We do this for characters like the spirits and demons from the Fade in the Dragon Age world.

Conversely, some other actors we specifically hire because they have what we are looking for in good delivery and timing mixed with this beautiful one-of-a-kind, extremely recognizable voice, something that fits well with one of our main characters. Those are the actors we generally use for only one major character in the game and for nothing else, since they are immediately so recognizable.

Varric’s German actor is a great example of this. The color and timbre of Varric’s voice in German is quite different to English but every bit as good. The warm personality of his voice fills it very well. In German, Varric’s deeper, sonorous voice allows that storytelling presence to be dialed up past eleven. It’s like James Earl Jones just came in and told you a comforting bedtime story. Varric’s character tends to be a scene stealer as it is, but that incredible voice gives the character in German its own style that is delightful and works really well in the language. I love playing with Varric in my party when we test in German.

Varric’s German voice differs a lot from the English one, the sonor ous voice giving the dwarf a bedtime-story-like quality.

IP consistency across multiple titles

We have a really, really enormous Wiki and a vast pronunciation guide, so that everybody uses our IP terms like »Lethallin« or »Morrigan« consistently across each game. Above all, we ask a lot of questions, document everything, and back-check a lot of facts, since it all depends on what the writers come up with next before we know what terms will need to apply in the next thing we do. I guess this is similar to what a historian or a researcher would do in a way. This again goes back to the resources we build up with our translation houses and VO studios. Dragon Age in particular is an extremely lore-heavy IP. There are vast background stories and codices that even after three games remain top secret and hidden in our internal Wikia. There are storylines that have not even been used yet.

Kinect Audio Localization

We explored using Kinect with paraphrased text and exploration commands in »Mass Effect 3«, but we found most users ignored those types of commands and found it faster and more intuitive to just press a button instead. Combat, however, creates a situation where many things are happening at once, so it can be of benefit to a certain style of real-time player there to shout commands to their followers while they are also executing some other command with the controller.

Our core combat mandate on Dragon Age: Inquisition is that a player can play in the style they want to play in. This could be heavy use of the tactical camera for a continuous pause-and-play strategy, or a strategy with heavier use of Kinect and controller issuing of faster, more fluid real-time commands. You don’t have to do strictly either one. There is no better or worse way to play, just what is right for that player.

I do feel the first prize goes to the German language for being the easiest language on which to develop for Kinect. It was like Kinect was made for German. Longer, more robust words, lots of hard consonants, and unique, non-repetitive sounds set German far apart from the Romance languages, English, or any other currently possible Kinect language in Europe.

Some languages have a more repetitive cadence to them and fewer unique phonetic sounds, so Kinect can have a harder time distinguishing between those sounds, much like the human ear does. German does have a large number of unique sounds. All of this benefits development in German, making it easier for the engine to interpret what is said and to a higher level of accuracy. Kinect on Dragon Age: Inquisition will be very combat-zone centered.

Dialects are a fascinating part of developing for Kinect

Some regions have hundreds of dialects, but you need to be careful with that or you would be developing forever and ever if you go down that rabbit hole too far. The English UK group of dialects is like that, and there are also a lot of dialects in French too. It is not feasible to get a voice sample from every human being on earth and define a custom phonetic set just for them. What we must do is define a representative sample for a region we are targeting, and then gather voice testers or voice samples from people who are native to that region.

We try to be careful. Pronunciation can even be affected by where you lived for five years during college, where your mother is from, what you do for a living, things like that. So if you are originally from Germany but your mother grew up in France and you went to Australia for an undergrad university degree in biochemistry or something, then that can affect your particular speech patterns and pronunciation slightly, even though you may not think consciously day-to-day about your life experiences having that effect on you.

We tackle this issue with extensive testing to make sure we cover the primary dialects of a region well, but the biggest factor in our success is actually the user. After using Kinect a short time, we find users will actually adapt their dialects to unconsciously fit the triggering of a command. It is almost like a situation you would find in a public park or on the bus, where you can be talking to a stranger with a different background or accent, and you automatically adjust, slow down, speed up, or alter the sound of your speech to be more understood by them. The same thing happens often naturally with players using Kinect. It seems evident with our testing groups that the more often they use it, the better they get at triggering commands, even though we have done nothing else technically to make those commands trigger more easily. It is a very human thing — intuitive, artistic.

What we learned from Dragon Age Inquisition …

… is that scale definitely matters. Having done production for English voice-over and localization of five BioWare games before this, I thought I knew a thing or two about large games and the type of scalable processes one would need to handle the volume of these projects. But Dragon Age: Inquisition has blown all of my expectations away, and I have found myself and my team inventing new ways to deal with the vastness of the levels. We have to figure out new ways to test those thoroughly.

We figured out new processes for how to deal with the enormous volume of text that needed to be translated, recorded, and then tested and checked. All of this as well was done on a new engine for us, which any developer knows can be really hard and can create its own set of challenges. Despite what you might think, the experience of all of this together has felt more like an opportunity to do so many things we have not been able to do before rather than just drudgery or hardship. With our developments on the Frostbite engine, we can go even further for future BioWare RPGs and other EA games too because we have helped paved the road for their work to be easier.

Languages are like enormous collective works of art. We inherit them and add to them, like we would a quilt or a tapestry. Each language is unique, complex, and beautiful. One of the most interesting parts of localizing games at BioWare is exploring the differences and similarities between languages. We work to find creative ways to overcome conceptual incompatibilities between cultures. We want to avoid linguistic misunderstandings and help people in our stories to find common ground with one another. Much of Localization production on a BioWare game is also designed to tap into the pure creative artistic expression present in our storytelling and best represent what the story is going to mean to different groups of people.

Melanie Fleming


About the author

Melanie Fleming
is Localization Producer at BioWare.

Melanie is a Localization Producer at BioWare’s Edmonton studio and responsible for the quality of translations and voice-overs produced in non-English languages. Traveling constantly and through many experiences, Melanie has been a musician, a sailor, a high school Chemistry teacher, and a project manager before moving on to the family at BioWare seven years ago. Her most notable games are Mass Effect 1, Mass Effect 2, Mass Effect 3, Dragon Age: Origins, and now, Dragon Age: Inquisition.