Changes in the translation software landscape

Greetings, Training for Translators subscribers, I hope you’re all doing well and enjoying the Northern Hemisphere summer! 

New books!

Beginning and aspiring interpreters: I’m excited about the positive reviews of my new book, Getting Started as a Freelance Interpreter; “the book I wish was available when I started,” according to a number of experienced interpreters! The print and electronic editions are available on my website (link above) or from any online bookstore! 

If you enjoy the Training for Translators blog posts, perhaps you’d like a PDF of all of them! For $10, you can buy my new PDF compilation of posts from January 2024 through March 2025. This gives you everything in one place, and it’s fully searchable. 

This week’s topic: Changes in the translation software landscape

Take this with the grain of salt that I’m not a translation technology specialist. However, a few readers have sent me questions about translation tools in the current technology environment, wondering:

• How is AI affecting the translation software market?
• If I’m looking for a translation memory tool right now, what should I buy?
• Are more companies (whether agencies or direct clients) moving to server-based versions of translation memory tools, instead of single-license tools that you have to buy? 

These are really good questions, and here’s my number one tip for all things translation technology: join techforword (not an affiliate deal; I’ve done training for them but I pay for my own membership). Josh Goldsmith and Nora Díaz are the people you want to listen to on these topics!

How may AI be affecting the translation tool landscape?

Interesting topic, and I honestly don’t know. It’s a question for people who do industry analysis (looking at CSA or Nimdzi). In one sense, AI-based tools like DeepL are doing a lot of the work that translation memory tools used to do, and tons of agencies have switched to machine translation with a human editor. 

I would also suspect (again, based on my own hunch and nothing more scientific), that “back in the day,” owning a fairly expensive translation memory tool was seen as a mark of professionalism. I’ve definitely talked to agency owners who said, “When we have to take a chance on someone new for an important project, if they own [insert name of market-leader translation memory tool], we know that at least they’re serious about the job because that’s a big investment.” 

These days, I think that agencies are more often part of a faster, lower-margin workflow, and they’re more likely to be serving end clients for whom good enough is good enough. And I don’t mean that in a critical way, I mean that’s just the reality of where the agency market has gone. As such, I think agencies are more likely to use translators who charge less, and who don’t necessarily own a market-leader translation memory tool. For example, I’ve talked to more translators whose agency clients use server-based versions of tools like Phrase. I personally like this system, because it frees me from owning and managing the TM tool; I just translate, and the agency compiles the final file. Translation memory tools are notoriously temperamental and buggy (I like Trados, and I still feel like I’m going to scream every time I see the error message, “Object not set to an instance of an object”). 

What translation tool should you buy now?

I recently had to address this question: I’ve been a Trados user forever, but I bought a new computer (always a good nudge to think about your software installs), and I never (literally, never) use the project management or collaboration features in Trados, and I’m going to need to upgrade it soon, so it was time to re-evaluate.

Unrelated to AI, many of the major translation memory tools have switched to a subscription model in the past few years, and none of these tools are inexpensive. An installed license for Trados will now run you US $795, upgrades start at $395, and the subscription with ongoing upgrades is $54 per month. 

I translate almost exclusively for direct clients, so the TM tool is for my own purposes. For projects where the main goal is a smooth, native-sounding translation, I prefer not to use TM. For something that’s strictly informational or where the document is heavily formatted (i.e. a “clean” PDF of an official document translation), I like using TM because I can focus on the text instead of the formatting, and I’ve calculated that I work about 10% faster even if the text isn’t repetitive. I also like the automated glossary feature versus manually looking up glossary terms. 

I went hunting for alternatives to Trados, and settled on Wordfast Anywhere, and I’m happy with it so far. It costs $1 for the first month and $10 per month thereafter, so the cost is very reasonable. A key consideration with any online tool is whether it stores your documents, and if so, whether they’re on a secure server. Wordfast Anywhere does store your documents (so I don’t use it for anything confidential) but they are on a secure server. I’ve been really impressed with its PDF conversion abilities, and the TM feature fits my needs. I really like that it’s a simplified tool that does mostly translation memory and glossary, so I’m not getting bogged down with a million features that I never use. This isn’t an affiliate deal, and I’d recommend Wordfast Anywhere if you want an alternative to the more full-featured tools. 

Is MT a replacement for TM? 

MT has long been confused with TM. Back in the early 2000s, I started exploring TM tools, and I first tried OmegaT (a free and open-source took that I still love, and I find its matching algorithm to be better than commercial tools’). I loaded up a few source segments, and then wondered, why isn’t it spitting out any English??? The problem wasn’t the tool, it was that I didn’t understand the difference between TM and MT. 

To be clear, MT and TM serve totally different purposes. I like TM because it saves me from translating the same text twice. And even though I have the Pro subscription to DeepL, I actually find it very inconsistent, even within the same document: I use DeepL mostly for interpreting prep into French, and it often uses two and even three terms for the same English word, in the same context, in the same file, for example translating the English expression “big data,” as “données massives,” “gros volumes de données” and “mégadonnées” within the same document. 

However, MT and TM have a commonality: for better or for worse, they speed up the translation process. And if you’re looking to really speed up the translation process, MT beats all the other options. In that sense, MT is a real threat to the TM market. I’ve written many times about how my translation work is almost completely unaffected by the AI boom. Why? My sense is it’s pretty simple: I work almost exclusively for direct clients, and they were never that tech-focused to start out with. They didn’t really care about TM, and they don’t really care about MT, they just want a good translation that’s ready when they need it, and they want a personal relationship with a translator who they trust. But for clients who are the opposite: preferring a faster, cheaper translation, then MT is way ahead of TM. 

Again, I’m not a translation tech expert, and these are just anecdotes from the translation ether, but I hope they’re helpful as we all navigate this rapidly changing landscape!