"Jestem dowódcą Shepard, a to jest mój ulubiony sklep na cytadeli." - Polish (machine translation)
In a GDC 2010 session entitled "Localizing Large RPGs," which forms part of an ongoing localization summit, BioWare localization project manager Ryan Warden eloquently explored the revered developer's process of adapting a huge, dialogue-driven game like Mass Effect 2 for alternative markets and languages. It may surprise you to learn that BioWare's processes are significantly more elegant than hiring an army of workers to copy and paste lines of text into Google Translate.
With 450,000 words and 30,000 lines of voice-over in the English version, Mass Effect 2 demanded an active approach in its translation to eight languages, including Russian, Czech and Spanish. "Trying to manage this scope is almost unfathomable," said Warden. "For BioWare titles, we don't have the luxury of waiting for the title to be fully complete before starting localization."
The concurrent process was designed to provide as much information and context to translators as possible, allowing them to focus on the job without having to request further information for each new conversation. BioWare compiled a complete localization kit, complete with a pronunciation guide (that's crow-guhn, not kroggin!), an IP glossary, a collection of translator Q&A documents and an extensive character database. "Any time that a translator spends time asking questions and waiting for feedback ... that's wasted time," Warden added. The goal, he said, was to "eliminate any doubt in the confidence of the translator."
The character database categorizes each of BioWare's characters, listing things like gender, species, personality and significance to the overall plot. Even minor characters are accounted for, including a Hanar that features in a humorous commercial you might hear on the planet of Illium:
"A character in an upcoming film -- a hanar (alien jellyfish creature) who plays a badass action-movie role. The joke is that hanar are both frail and extremely polite, so having him deliver badass lines in polite eloquent speech sounds absurd."
In addition to providing reference material, BioWare's localization was designed to follow right behind the English version's voice-over recording -- by the time dialogue is recorded, the text is likely close enough to being final and ready for translation. Should changes be made, BioWare has a system for flagging changes as "minor" or "major," with the latter prompting a re-translation by a trio of translators assigned to each different language.
Chris Christou, BioWare's lead localization tools programmer, provided a rough overview of BioWare's complete translation pipeline, which incorporates not only contextual notes on dialogue (he said, as he punched a hapless guard through a window), but an ability to click through the game's dialogue trees. It's a bit like playing the crudest, most basic version of a typical BioWare conversation.
BioWare also had to compartmentalize its translation process, taking care to ensure that work being done on downloadable content and patches did not interfere with the final release candidate. "At some point, development just has to end on the title," Warden said. "You have to ship that, you have to get that on the shelf."
Thankfully, the concurrent localization was nimble enough to adapt to a last-minute change to Mass Effect 2, just weeks before certification. "As incredible as it may seem, we did bring back some characters and add additional VO," Warden explained. It was feared that one part of the game didn't make it clear enough what players had to do next, or where they had to go. "It was absolutely the right call. It made the game better."
By the end of the project, Mass Effect 2's 450,000 words had ballooned to 2.7 million after localization, along with 140,000 lines of voice-over work that required over 350 actors and 300 days in the recording studio.