Will IA and automated translation services replace handmade localization? | Stillman Translations
Will IA and automated translation services replace handmade localization?

Translation and localization, as we’ve previously stated, can be a tricky business. It’s not as easy as word for word or having a good synonym at hand.

Photo by Alex Knight on Unsplash  

Not long ago, an article was published that gave room to some debate. Here’s a peek into what we think about it. 

Many problems arise when we’re knee-deep in this alphabet soup. Resizing, breaching character limits, syntax, grammar issues, interpretation issues… sometimes even actual gibberish when translating to other scripts such as Arabic. And all of this doesn’t go unheard. On the contrary, it has been widely talked about throughout the gaming world. 

An article titled “As translation technology improves, game localizations are getting worse” materialized all of these concerns to host an interesting debate on Reddit.  

Can machines do the job of trained and paid humans? 


The article’s opening line is that “game developers are becoming increasingly dependent on fans and enthusiasts to quality test their work”. And it further adds that “localization quality is taking a critical hit. When Tabletop Simulator announced that they had added new languages via Google Translate, users from all over the world pointed out errors and explained why machines can’t do the job of trained and paid humans”. This is where it all begins. Why, during the peak of technology, virtual reality, 4K graphics, collaborative work, streaming, is something as basic as this failing? The lines of argument can be reduced to three. 

One speaks about negligence. Some believe “most problems have to do with bad practices rather than the technology used”. Bad practices such as (lack of communication and poor management). And that the use of MT (machine translation) is “mostly an issue of companies not being willing to spend the money.” 

Many times, the translation of the storyline is deemed an unnecessary expense. And resources are allocated elsewhere. This leads to fans themselves doing the dirty work. Or to the rise of specific tools to help fans play in their preferred language. Such as the Universal Game Translator created by Seth A. Robinson. This app used Google’s API to scan screenshots of a game, identify text, and present the translation. 

Another argument that makes its way is excessive regulation, not negligence, as the reason why MT is chosen by default. Take Apple Arcade, a gaming platform that demands at least 14 languages available for the game to be accepted and distributed. Companies feel forced to invest in markets they were not open to and therefore go for sloppy but readymade translations. It means a poor quality product, but a product that they didn’t intend to sell anyway.  

And thirdly, some underestimate the impact of localizing instead of translating. Some companies rely completely on machine translation. Then hand it off to a proofreader, which is less costly. They check for errors, such as typos and wrongly composed sentences. But as Argentinean game critic Diego Nicolás Argüello highlights, that still leaves translation “a bit rusty in places” and “quite literal”. The craft is writing in a way that sounds natural to native speakers. 

It’s not all bad news though. There are developers trying to hire people who are also players of their games. Which is a monumental shift. Since these are people with content, experience, and attachment to the final product. But the standard still has a long way to go. 


This phenomenon happens everywhere. Especially when we leave the classic roman alphabet. Nope, Not Arabic, for example, is an archive of things that are not Arabic. It documents instances of butchered Arabic script in public. From major motion pictures to billboards to video games and everything in between. 

The site’s owner explains that computers are designed to handle the subset of the Latin script. But that every other writing system is poorly served. This archive is what happens when a non-Arab graphic designer pastes Arabic text into mainstream graphic design. And doesn’t check with a native speaker. 

This is true for all non-Latin scripts, but the Arabic script, in particular, seems to present something of a pathological case.  

Take 5 to digest everything you’ve read up to now. It’s no wonder fan translation communities are a thing. They used to be Japanese-exclusive computer and video games but have since expanded to include other languages as well. 

Fans cover the technical aspects, such as literacy, language, and IT-related practices. But also acquire and put into practice sociocultural skills, while enabling interesting conversations online with fellow gamers. Which derives into meaningful metalinguistic discussions on translation strategies. 


Japanese to English translator Katrina Leonoudakis explains that “no two games are approached the same” and that “every language pair has its own difficulties. Japanese doesn’t distinguish between singular and plural, but English does. English doesn’t distinguish between male and female for adjectives, but Spanish does. Each of these difficulties comes with a choice – do we ask the dev to reengineer something? Do we avoid pronouns or use abbreviations?” 

When we localize, we’re taking a big say on the impact the game will have on users from around the world. We decide on story-specific terminology, character accents, what amount of background knowledge should be explicit, underlying concepts, and more. These issues influence the number of sales in a given country. 

Do we want players to feel at home? Or, on the contrary, are the foreign cultural elements what make it rich? Do we want localization to erase the ‘cultural odor’ or boost it? For example, Kirby is a well-known Nintendo game. This cute little monster goes from a cutesy, innocent figure to an angrier warrior-like version. The Nintendo marketing department believed an aggressive Kirby would do better for the art cover in American markets, rather than the innocent, sweet-faced one. These are the professional localization decisions involved in any well-planned strategy. 

There may even be legal issues. For example, in 2000 Square released Final Fantasy IX, whose main character was Zidane, in a literal Japanese translation to English. But Zidane had to be changed, because of a famous French footballer with the same name. So Zidane was changed to Djidane for the French translation, and to Yitán for the Spanish translation. 

Machine Translation has no sensibility or judgment or humor. Not even common sense. So if you want to make a game special, we believe the road less trodden is the road to success. 


The availability of talented translation teams can produce high-quality translation for any market. When should they be involved? If possible, as early as the initial development phase of the game’s characters, storyline, and technical aspects. At STILLMAN we have language professionals that get involved as any other member of your team. 

Project managers, linguists, voiceover dubbing specialists, technicians, legal specialists, translators. Anything you need can be provided. It’s this collaborative team of experts that can produce top-quality video game localization. 

Game translation typically includes text, visuals, audio, product packaging (instruction manuals, user guides, policies), and various other digital materials. Even websites and social media can be considered.  

The result of an engaged team is consistency. For example, a given skill, or potion must always have the same translation in any given part of the game. MT does not keep track of this. Nor of the many twists and turns previously mentioned make a game fun.  

Want to know more? We’re only a few clicks away