To craft a nice machine translation of this game, I pretended it was originally released in another language (it wouldn't have made sense to process English directly through Google Translate).

Therefore, the first step was to translate the script into Italian, my target language. After importing everything into MemoQ and begging my trusted developer to help with Regex, all that was left was to move on to translation.

But, since I finally had the means and time and didn't have to rush because of deadlines, I wanted to assess whether (working on such a simple and linear text) it would actually be convenient to use the Text-to-Speech feature in MemoQ.
It's a function I use occasionally, only with extremely simple files, but until today I've never been able to assess whether it saves me time or not, so this was the right opportunity to give it a try.

I reached the end of the 3361-word file in 2 hours and 20 minutes. "Reaching the end of the file" doesn't mean I've finished localizing it: there are still blank sentences and parts to review, but it's at least a first draft that allows me to track the time invested.

So, we're at around 1440 words per hour. For files of this kind, I usually hover around 1100 to 1300 words per hour, so despite the difficulties, it seems slightly more advantageous to use Text-to-Speech, but it's still estimates based on other files, so I should re-translate this same file by hand to compare the speed at the same level of difficulty. Something I'll probably do in a few months, when I've forgotten most of the rest and it will be like starting from scratch.

Things I learned using Text-to-speech (but I already knew anyway)

It's vital to have a quiet environment when using such a function, as even the sporadic typing can interrupt the recording. Not to mention when the cats scream loudly or knock things over.

Another downside is not being able to play some music, something I usually listen to in background when possible.

A good deal of editing and proofreading is still necessary, to find and correct issues (whether they're caused by pronunciation or by MeomQ not having great listening skills).

Furthermore, it must be said that hand-typing stresses me out less, because when the tool doesn't understand a very simple word for the third time , all I wanna do is tossing the PC and the app out the window.

However, using Text-to-Speech can provide some gems, like this one, where I reminded my partner that it was very late and it was time to go to sleep (I had stopped the recording but as soon as I reactivated it, the program still filled in the string with what I had said).

Or this other one, from another game I often translate by dictating, where the word "scovare" just couldn't be understood no matter how carefully I pronounced it.

The misspelled version of Scovare (to discover) is basically... Füçk

Or when the cat claws you to climb on your lap...

What we have learned from this little experiment is: always proofread thoroughly, because misunderstandings can result in NSFW content!