Game localization tools: speech recognition and text readers

A brief summary of the benefits and drawbacks of using speech recognition and text readers for your videogame translations.

Two of the most shamelessly nerdy technologies around. But try to give them five minutes of your attention, because they might prove surprisingly useful

Speech recognition

How?

Dragon Naturally Speaking. Available everywhere, solid and supported. There’s an updated version every year for about 100 EUR, but you can work just as well with older and cheaper ones. Unless you plan to use an up-market laptop with a good integrated microphone, you must also buy a USB headset for good results.

How embarrassing? A lot

The first time you open the box, the Italian version makes you read a couple of chapters of Pinocchio to train the system.

So your neighbors will hear you reading bedtime stories to your pc.

Then, when everything is ready, you will inevitably start acting your own dialogs in order to improve characterization.

So your neighbors will hear you having heated debates. With yourself. For hours.

How fast? Scary fast

When everything is ready, you will feel like in 1950’s movies, using the intercom to call Jenny the Secretary to type a letter for you.

And it just works.

It may understand one word for another here and there, but 90% of the times you just need to glance at the original text, close your eyes to focus on a good translation and when you open them it’s all written down in front of you.

With the right text (long, clear sentences, no research needed) you can cover insane amounts of words.

With a system so efficient, you will find out that the biggest bottleneck is… well, you! After a full day of voice recorded translation, the machine will still be ready to continue, but the little human in front of it will be exhausted.

How good? Better

First: you will be 100% focused on the translation. Even if you are the fastest typist this side of the Dvorak river, you can’t deny that typing takes a bit of mental effort. With voice recognition, your only focus is on the translation, and that’s where your added value is. Second: Your first knowledge of a language is as a speaker, so translation mistakes like calques and missed idioms will be immediately be apparent when you say them out loud. The same goes for alliterations and other tongue twisters.

Finally, talking puts a limit to the length and complexity of your sentences.

At some point, you will either run out of breath or lose your train of thought and be forced to rephrase in a snappier, clearer way. Never a bad choice, when your target audience is reading your text while killing zombies with one hand and eating pizza with the other.

Text readers

How

There’s probably a lot of commercial products out there, but we never really had to investigate, because the freeware DSpeechjust covers all our needs.

How embarrassing

Modern voices are much better than Microsoft SAM, but they are still quite robotic. You need headphones, or the neighbors could finally decide to call the police.

How fast? Not much faster, but much better

First: it makes you more resistant to boredom.

While reading a wall of boring text, you will inevitably start to skim. The reader will robotically endure each singly line. Your part is to keep the headphones on (please do), follow with a finger on the screen and listen. Even if you fall in R.E.M. state at some point, hearing mistakes out loud is guaranteed to wake you up.

Second: like for the voice recognition, you will instantly notice if a sentence is too long, complex or misleading. Also, when listeners grasp the meaning of a sentence, they tend to finish it mentally. When this happens to you, pause the audio and write it down, because what you have here is a nicely idiomatic form that will flow just as effortlessly with the audience.

Conclusions

Voice recognition is probably a matter of taste.

If you enjoy typing and tend to polish separate words in your translation, I wouldn’t recommend it. However, if you enjoy tools that save you time/effort and think mostly in terms of flow and comprehension, go for it. I have been using Dragon every day for the past 6 years and that’s pretty much the first software I install on a new PC. It’s that good.

On the other hand, I can’t see any reason for not trying text readers. Even if you are already O.C.D. meticulous enough to read out loud every single word of your translations, you can at least save your little eyes (just don’t forget the headphones)


Alain Dellepiane

Alain Dellepiane @gloc247 01 July 2011
Alain is the founder of team GLOC. Want to read more about localization? You should probably try this blog's Best of, which has a dozen of the best articles ready to read. (View all posts by Alain ➜)

Stay updated.

Receive a monthly email with the best game localization news, papers and tools.

Check out the latest issue. Free of charge, zero spam, unsubscribe any time.