Tight budgets always make me look at what’s possible. When a client asked about getting their eLearning translated, we talked about cheaper options and how far we could go without a full translation service. That’s when I suggested trying DeepL, an AI translation tool that promised professional results.
I’d heard about DeepL before but hadn’t used it myself. During my deep dive research, I’d seen it described as one of the more reliable AI translation tools, closer to natural language and more accurate than many of the free tools.
I’ve worked on translation projects before, usually alongside professional translators. I’m always a bit wary about using AI for things like this. The marketing often makes it sound like the best thing since sliced bread, but in practice, you still need a lot of human review to get something you’d actually be happy to use.
When a client approached me about translating their eLearning project into another language, I gave them two options. One was the cost of using a professional translation service, the other was the price for going down the AI route. I was clear from the start that I don’t speak the language, so I wouldn’t be able to check the accuracy myself. The professional service included review and quality assurance in the price, while the AI option would mean finding someone on their team to do that part internally. I also made it clear that AI translation isn’t 100% accurate; it can save time, but it still needs a human eye.
With all this in mind, the client chose the AI route. I should note that I used the Business plan, so I was able to access professional-grade features and glossaries.
DeepL is easy to use. Much like other AI tools, you paste in your text, choose a language, and the translation appears almost instantly. Within half an hour, I had a full course translated, something that would usually take days. As mentioned, I couldn’t review the accuracy myself, but I did use Google Translate as a rough checker. Obviously, that’s not perfect either, but when both tools produced similar phrasing, I felt reasonably confident.
The real difference, though, came from having someone on the client side who could review the translations properly. Without that, I wouldn’t have trusted the output completely. It also made me realise that while the tool cut down on turnaround time, the savings weren’t as clear-cut. The client still had to spend time checking everything, and that’s hard to measure in cost terms. It was clear from the remediation work that there were a lot of things that weren’t translated correctly.
I would definitely recommend using DeepL as an AI tool and I would use it again in the future. With the right collaboration and realistic expectations, AI can make multilingual projects possible when they might otherwise be out of reach. However, as with many AI tools I have used, it reinforces the message that AI still don’t replace the need for humans; they just change the processes.
Resources
If you’d like to try using DeepL yourself, they offer a free trial (at the time of writing this post).