Tamara Tirjak has been working in Frontier’s localization department since 2015. Since then she helped localize company’s biggest projects – Elite Dangerous and Planet Zoo into dozens of languages. We talked to Tamara and found out when it’s best to start working on localization, what’s the point of adding an artificial language to your game and in what way having a proprietary game engine helps with translation.
When NASA announced the discovery of the Trappist-1 system, our game already had a very similar one
— Frontier Developments works with proprietary game engine Cobra that’s been evolving since 1988. Why, do you think, Frontier spent time and resources on developing their own engine when there were ready-made solutions?
— Back in 1988 there were no commercial engines available to support making the games Frontier wanted to make. Unreal was only introduced in 1998, and Unity did not come around until 2005. Therefore, Frontier had to create a unique engine for our unique games; an engine that can be scaled and customised as needed, in order to achieve the outstanding gameplay and visualisation we endeavour to deliver to our players. By having our own internal engine team, we can develop custom tools and add new engine features as needed, and we do not need to wait for a commercial engine team to consider and develop our feature request. We can stay focused, since our own games drive our engine development, and we can optimise Cobra to the specific tasks it needs to perform, be it simulating park visitors and zoo animals, or rendering star systems.
Let me share a fascinating story about the Stellar Forge, the part of Cobra which is being used to create the world where Elite Dangerous takes place, by generating the roughly 400 billion star systems in the Milky Way galaxy, using existing astronomical data available and filling in the gaps using predictive calculations. This simulation has turned out to be so realistic that when NASA announced the discovery of the Trappist-1 system in 2017, our game already had a very similar system with seven planets nearly at the same location!
— A lot of game companies struggle with implementing localized files because of their game engines. How do Cobra’s resource management systems work with lockits? Would you say ingame texts are conveniently systemized?
— Text resources are handled separately from the code in a convenient and flexible xml or txt format. A huge advantage of having internal build engineers and tools programmers is that they can create bespoke solutions for us to improve our work processes. For example, we have recently developed and implemented a localisation asset management system, which allows us to sort and filter text strings, add metatags and comments, and of course track their status in the workflow. This system has streamlined the communication between the people who create, review, translate and test the text in our games. Of course, this system is also able to work with the voice files of the recorded dialogs. We coordinate hundreds of voice lines and thousands of text lines across multiple languages, so we need a robust system to know what has been finalised, what is currently in progress, and what needs rework because it has become out of date.
— What kind of tasks can be currently automated with the technologies Frontier uses? How do you and your department work with the game engine?
— The string database mentioned above is connected to our memoQ server, which is Frontier’s CAT tool of choice. This means that once we select the lines for the next translation batch and press Export, they appear in memoQ within a minute or so, and we are but one click away from sending them straight to our outsource translation teams, or start working with them in-house.
The localisation team works directly with the game build and assets; we can check out and run our own local game builds to verify the new translations in context before committing them into the repository.
It is crucial to have an international mindset throughout the development process
— A lot of localization evangelists nowadays talk about having an international mindset when developing the game and even thinking about internationalization and localization when creating the narrative side of the game. What’s Frontier’s approach?
— I absolutely agree. If a developer aspires to market their game globally, it is crucial to have an international mindset throughout the development process; not just in the narrative, but also in the visuals, environments, UI design, everything.
The localisation team at Frontier is well integrated in the Production teams from quite an early stage, which means that we can provide input about the localizability and potential cultural implications, and this also gives us visibility, encouraging the production teams to proactively reach out to us, as they frequently do, if they have any concerns about global appeal.
— On what stage of game development would it be ideal to involve the localization department? Some companies only wake up so to say, when the game is already in beta or even after release.
— It depends on the length of the development cycle and the size and complexity of the game, but I believe that the Beta is rather late. I would normally prefer to have the translations already in game for the Beta if that is possible at all. It is a great opportunity to gather some player feedback not just about the translation itself, but to gauge the reception of the game outside the English-speaking territories. Even if the actual translation work does not start before the game enters Alpha, it is important to include the internal localisation specialists in the loop much earlier than that, because we can help making sure that the UI design, the narrative, the environment and the items are culturally appropriate and relevant to our global player base, and we can also prevent design decisions causing globalisation problems further down the line, when they are more costly to fix. Of course, I understand that we are in quite a privileged position at Frontier, where the Production and Publishing teams are under the same roof, and this can be much more challenging for a development studio to loop in the localisation experts working for their publisher, often in a different country!
— Have you ever had to change the gameplay features because of the feedback your localization department gave to the game designers?
— There is precedent for renaming or changing the appearance of an item, and parts of the UI were also adjusted to accommodate our feedback. There has not been any need for changing gameplay features, because Localisation is involved in any major design decisions that can have international or cultural implications, so we can normally prevent problematic features from being developed in the first place.
I am a firm believer in continuous feedback
— From your experience, how to build transparent communication between the localization vendors (if they are in the equation), the internal localization department, and other teams within the company?
— First of all, I believe in seeing each other as human beings. IM, emails and calls are great and efficient to get the job at hand done, but a chat in the tea kitchen or having lunch together can go such a long way in making this online process work!
I was hired by Frontier in 2015 to build up their localisation department and workflows from scratch, and a very important part of that process was to establish myself as “that nice girl with a big smile, who is always helpful even if I need something translated a day before release. She is very knowledgeable about languages and culture, and is very enthusiastic to share her passion with anyone who cares to listen”. My team has been growing since then, but we still work on cultivating this image to build trust with the rest of the internal teams. After all, this is our shared goal, to create games that will enjoy global success, so the least we can do is help each other in this endeavour.
As for the vendors, I feel it is important to meet face to face, if only once for the project kick-off or during industry events. I like to treat them as ‘remote team members’ and share with them our plans, intentions and challenges, and I am also a firm believer in continuous feedback about their work to strengthen our partnership and align them to our goal: to publish a great game, which will be well received all over the world.
— When choosing new localization vendors companies often rely on miniscule test tasks. How representative is an average test job in your opinion?
— It depends on how well the test is put together, whether it is representative of the actual game, and if it tests the translation skills that will be the most relevant for a high-quality delivery. A test will obviously never be able to fully imitate the production scenario, where the translator would have access to extensive documentation, maybe even a game build to check for context, and they would have a chance to clarify their doubts with the client. For a test translation, the linguist might not even be told the title or genre of the game yet! But these tests are definitely useful to have a quantitative score to have a somewhat objective comparison of the candidates and to be able to justify your outsource decision.
Even with these limitations, I still find test translations a very useful tool in assessing how well my potential partner deals with placeholder tags, transcreation and puns, or technical and scientific text, based on my project’s needs. I also invite them to add comments to each line, and I am very happy to see translators adding proof of their research, a justification of their word choice, or taking note of a doubt that they would have queried in a real-life situation. For me, a translation test goes beyond testing simply linguistic skills; I also want to get an impression of their work standards and attitude.
— How do you measure quality? What’s your stance on quantitative analysis of translation?
— As a linguist myself, I understand that the goodness of a translated text is a highly subjective matter. But as a manager of a world-class studio, I also understand the need for data-driven decisions. If I am asked to justify our outsource decisions, I need to be able to present hard data to prove that our partners handed in the best test translation, and they continue to perform to our expectations.
On a day-to-day basis, we use linguistic QA to regularly provide feedback to the translators and improve the quality of our games. We have recently set up our own customised LQA model to make sure we measure and score things that matter to us, and by doing this we also wanted to develop a system that is fair to the translators and do not set up unreasonable expectations.
— What is, in your opinion, last decade’s greatest innovation in localization industry?
— It will have to be neural machine translation, even though I think we expected this technology to be more disruptive than it has actually turned out to be, at least up to this point in time. For now, its uses in our own segment, video game localisation has been limited, at least in the sense of using it to support the work of translators. However, there are interesting use cases in our industry where NMT removes the language barrier in real-time communication within the player community, or facilitates the communication between the publisher and their global player base.
I also know of some very promising NMT projects, so I expect this technology to start having a bigger impact on our translation work very soon.
— What would you consider today’s biggest challenge in localization industry?
— The shift to continuous localisation due to the ‘Games as a Service’ approach. This means that developers are constantly injecting new features and content into their live games – or live products, as this has become a tendency for software developers in general, not only game studios. The release cycles can vary between a few days to a few months, but the result is the same: We no longer have the luxury of the good old days of the waterfall method to wait until the previous process step is finished before starting ours – now everything has to happen at the same time, and the handoffs have become smaller and more frequent. This puts extreme pressure on every single actor in the value chain to reduce their overhead and admin time and costs to stay competitive. Little wonder that automation, hands-off project management and lean production have been the hottest topics in the industry lately.
Value of localization is very difficult to ‘measure’I am a firm believer in continuous feedback
— A glaring issue with localization. How to measure the value of localizing the project to certain languages? Does adding more languages significantly broaden the playerbase?
— There are various ways to estimate the value added by localisation, but this is very difficult to ‘measure’; for that we would need sterile, lab-like scenarios, to carefully tweak individual variables, and this is rarely possible in video game publishing.
One of the basic metrics to monitor is the Return on Investment, which compares, let’s say, the Italian localisation costs to the revenues coming from Italy. The next question to answer is whether your players in Italy actually play in Italian, and most publishers would have in-game telemetry to measure and answer that.
And finally, my favourite question: Would these players still buy our game and accept to play in English if we had not localised it? This can sometimes be measured by an A/B comparison, if your portfolio has two similar games, but one was translated into Italian, while the other was not. Given that their performance was similar on every other market, you can compare your Italian sales and hope to see that the one that was localised sold x times as many copies as the one that was not.
These would be some of the metrics we would track to evaluate the success of our localisation efforts.
— Who’s in charge of choosing new markets and languages in Frontier Developments? What kind of metrics are you using to assess attractiveness of certain markets?
— This is a shared decision across multiple disciplines. We identify target markets by looking at things like: Is this genre popular in this country? How big is the market there for this platform? Do we see any growth opportunities? Do our target players expect the game to be localised, or are they happy to play in English? Are there any technical challenges or cultural limitations, which would increase the effort required (e.g. we would have to implement support for a right-to-left language or need substantial modifications to get it approved for publishing), and is it still worth it?
— What CAT-tools does your department use? What would be your personal choice? How effective are in your opinion cloud based CAT-tools?
— Our tool of choice is memoQ, which is a perfect fit for the complexity of our operations. I love the automation features and its versatility in handling practically any file formats I throw at it, even if preparing the filter for it requires some basic knowledge of regular expressions, for example.
In general, I am very fond of tools that facilitate online collaboration; I consider them indispensable in today’s agile development environment. The demand for this synchronicity is even more pressing in the world of mobile games, where the development cycles tend to be even shorter than ours.
But as with every online tool, security must come first! We are working in a highly confidential environment, where a data leak can be devastating. I feel that many IT professionals still prefer to host online tools on premises, rather than putting them out into the cloud, even if this solution presents the question of how to open the ports for our outsource partners to access our online server securely.
Synthesised voices will not replace real voice talent
— How does Frontier use Text to Speech in development? What kind of companies could benefit from implementing a similar system?
— At the moment, we use it as a placeholder for voice over lines to test and iterate on the dialogs before the final lines are recorded and delivered by the recording studios, and previously we used it to deliver GalNet Audio, an in-game news service of Elite Dangerous.
Most of the easily available synthesised voices sound quite robotic, which perfectly fits our use cases. However, there are companies out there who have achieved very impressive results to add emotions and richer intonation. I can totally see these becoming a viable option for things like background chatter or crowd walla in the near future, but they will not replace real voice talent to deliver the superb performance that we have become used to in the narrative of AAA titles.
— Other than adding a certain level of depth to the lore what other problems GalNet Audio solves?
— The design intent behind GalNet Audio and the in-game knowledgebase, the ‘Codex’, was that we wanted to deliver a richer narrative and a more immersive experience of the Elite universe to the players, but without requiring them to read all that text. This enabled players to fly around space and listen to the news in the meantime, without needing to take time off from playing.
The two major inhibiting factors to using recording studios was not just the cost but also the short lead times. Since these articles were often written as a reaction to in-world events, they sometimes had to go live within 24 hours. This already made translation tight, let alone having to arrange a recording studio.
TTS offered an ideal solution to both of these challenges, and the synthesised voice fits the lore of a space game perfectly. The localisation team now takes care of the TTS conversion internally for all our supported languages (UK English, German, French, EU Spanish, Russian, Brazilian Portuguese), making sure that it sounds good, and the terms specific to the Elite Dangerous universe, such as the names of star systems, stations, personalities and brands, are pronounced correctly in every language.
— Frontier Developments uses Text To Speech both internally for development purposes and for GalNet in Elite Dangerous. Is the same technology used in both areas?
— Fundamentally, yes. The main difference is that while the process for internal purposes only is completely automated, any player-facing TTS asset will go through manual tweaking before getting published.
I am very intrigued by the connection between the culture and the language
— Frontier Developments created a fictional language for Planet Coaster and Planet Zoo called Planco. Planco is based on English, but has its own unique vocabulary. Your colleague James Stant talks about Planco in great detail in the 2017 article published on Gamasutra. How was the idea of adding a fictional language first introduced?
— When our Audio team was investigating the best approach for creating the way the crowd should sound in Planet Coaster, they looked at various options. Simple “Ooooh” “Aaah” “Grrrrr” sounds did not give the park visitors enough depth, when they are at the heart of the simulation, and the players should genuinely care about them and try to make them happy. Using English, and localising it to all our supported languages was also dismissed because we wanted the game to be location agnostic; Planet Coaster is meant to be a separate universe in its own right. Also, hearing the same line repeatedly in a language you understand leads to ‘audio fatigue’ much quicker than if you do not understand it.
So James came up with the idea to create our own language, which fits the atmosphere of the game, and gives personality, depth and authenticity to our crowd.
— You’ve helped to compile the Planco Dictionary, have you worked with constructed languages before Frontier Development?
— This was definitely a first for me, and I really enjoyed the challenge. We currently use a tool called Polyglot to manage our ever-growing glossary. Polyglot was developed by Draque Thompson specifically to assist people who want to start building their own constructed language. The same glossary is also plugged into memoQ, so if we need anything translated, memoQ can work as a simple but smart translation engine from English into Planco, and deliver the new dialog lines in Planco along with the pronunciation, in a matter of seconds.
— What are impressions from working on constructed languages? Would you be interested in creating another artificial language for a different project?
— What language geek could possibly say ‘no’ to this question? I am very intrigued by the connection between the culture and the language, and how the way of thinking manifests in the way a language is structured, the words they use and the general approach to communication. At some point in my life, I would love to be presented with a race or a fictitious nation, with a detailed explanation of their background, history and traditions, and then asked to figure out the language these people would speak.
— How did you taking into account release on foreign markets? How did you make sure that a dictionary of 7000 words is not offensive in any language your games are released in?
— We are lucky that Frontier’s staff is very diverse, and we have most major languages represented internally. Once the crowd voice over has all been recorded, we asked our colleagues to listen to all the lines to spot things that might be inappropriate in their language, and we did end up removing a handful of lines, but all in all they were pretty innocent.
— What is your favorite word?
— Huh, that’s a tough one! I love ‘allyooma’ (roller coaster), ‘faieyva’ (forever) and ‘hasswuuf’ (hotdog). And then there is that whole song about the ‘wippy tentifu’ (happy octopus) from the Whirly Rig ride:
There are cultures that we are more cautious about
— From your experience, what should be done to prepare your project for multicultural release?
— For a game that is culturally appropriate and relevant, you need people in the development teams who are sensitised to things like cultural and gender issues. The more diverse the team is, the more this attitude will develop naturally. They need to be able to voice concerns and raise questions if they have doubts about any element of the game. Then you also need experts, either somebody within the studio or an external expert, who is trained and experienced enough to address these concerns and answer these questions.
— Is there any language you especially concentrated on while localizing your projects? Why?
— We tend to treat each language equally, but there are definitely cultures that we are more cautious about. When reviewing our in-game content, we want to make sure we handle East Asia sensitively, as those territories are a huge market for games, and you definitely do not want to offend your future customers! We are also particularly mindful about South Asian and Arabic territories, African tribal culture and the indigenous people of the Americas. They might not be our main target market, but a respectful representation of global cultures and preventing cultural (mis)appropriation are important to us.
— Have you ever had to change something in translations because of the feedback given by community?
— At Frontier, we take pride in listening to our community and being approachable by our players. This is especially true for Elite Dangerous, where the original Spanish and Portuguese translation of the game was done as a community initiative, and although the updates are no longer handled by them, they still reach out to us if they spot an error in the translation. We do keep our feelers out to pick up issues raised by our community. Not all of them get implemented in the end, but they all get discussed and considered.
— Can you remember a case when the missteps in culturalization led to some controversy?
— My favourite incident happened after we released Planet Coaster back in 2016. There are various shops that the players can place in their theme park, one of them being Monsieur Frites, a stall selling French fries with a handful of different condiments. The day after the release, we received a bug logged by a French player saying something along the lines of: “I understand that you are a British developer, and you would normally put vinegar on your fries (yuck!), but if you create a French-themed shop in your game, then it should be mayo”. We now offer curry sauce, ketchup, mayonnaise, vinegar and others to accommodate the varied taste of our theme park guests and players!
I have one eye on localisation, and one on the games industry
— How many people are in your department? What’s the structure of your department?
— As the Senior Localisation Manager, I am in charge of overseeing all aspects of localisation. My team has three (soon four) Localisation Coordinators, who are in charge of the day-to-day tasks of managing handoffs and handbacks to and from our external partners, finding and fixing localisation bugs and meeting the deadlines of our release schedules. We also have two localisation testers performing language checks internally and coordinating our outsourced test runs.
— What your typical workday looks like?
— The typical tasks are carried out by the rest of the team, whatever is escalated to me tends to be atypical and non-trivial. As we are still a small team, I wear many hats as the senior localisation specialist. I work with the Producers to agree on high level timelines and budgets; take care of the administrative tasks related to tracking our tasks, costs and invoices; attend tons of meetings; work with the engineers or IT on developing or purchasing new tools to improve our work practices; write memoQ file filters, batch files and anything else required for process automation; select vendors; hire new team members; and motivate my current team and help them grow professionally.
But I am not complaining, it is precisely this diversity that makes me love my field. When you are a localisation manager, you are a bit of a linguist, a bit of an engineer, a bit of a business strategist, and a bit of a team lead.
— How do you stay updated on the things happening in the world of localization? What magazines or sites do you read?
— I tend to have one eye on localisation, and one on the games industry, as I try to stay up-to-date with both of them.
For localisation, I am subscribed to the Multilingual Magazine, and enjoy listening to the Globally Speaking podcast while cycling to work. I also love Miguel Sepulveda’s ‘yolocalizo’ blog, he has a lot of interesting insight to share about the various aspects of our profession.
I also follow Slator, Nimdzi Insights and Common Sense Advisory for the latest news and research results in the industry.
When it comes to games, I tend to have a look at Gamesindustry.biz every day, and I follow Gamasutra.
Apart from these two areas, I am nowadays interested in global marketing, and have recently discovered a podcast called ‘The Worldly Marketer’, which I enjoy very much, and I read a lot of books on culture, mainly on cultural dimensions and cultural intelligence. I should try to find a good podcast for that too. Spring is coming, time to spend more time on my bicycle!
— What conferences are worth attending, when the pandemic is over?
— Just like with my media consumption, I try to share my time between events focusing on localisation and game development.
I have been a regular attendee of memoQfest since 2010 or so, I simply love the atmosphere of the event, and there are very knowledgeable people there to share ideas with. And of course it is the perfect medium to stay up-to-date with the latest memoQ features, and to have a say in the direction the tool takes, because the developers are very open to feedback from their user base.
In the last 3 years, I also attended the Game Quality Forum Global, where I led the interactive workshop on the Localisation track. This event features video game industry professionals from the world of QA, Localisation, Player Support and Community Management, and I find this event a great opportunity to learn from my peers in the field.
I also attended DevCom 2018 and Game Connection 2019 as a speaker, and I head down to sunny Brighton almost every year for Develop.