Personally, I decided using a local LLM was acceptable for generating translation strings in my open source application. It had already been manually translated into a bunch of different languages by various contributors over the years, but I just did a major UI rework and so most of the existing translations now had a ton of missing entries. Qt6’s translation tool had an AI Translation function so I decided to try it out after setting up ollama and the recommended qwen3 model. It did a pretty decent job as far as I can tell, I am only fluent in English but I did some cross checking by translating back to English via Google Translate and it seemed to do a decent job. It at least got all the strings translated to something usable, once it’s merged then I expect native speakers will contribute cleanups and rewordings if needed.

I don’t really consider translation purely generative as it is more of a conversion task. I think AI is pretty useful for this. Same for things like automatic captioning and even mundane text to speech (if not being used to replace voice acting).