Hi @Max_Marko,
LR doesn’t create translations exactly.
For human translations, LR takes the subtitles that it finds in the video source — for YouTube it’s the translations that the YouTuber uploads with their video, for Netflix it’s the translations that Netflix paid translators to do for videos.
For human translations, a human did theoretically sit down and caption/subtitle the captions/subtitles — for the Netflix/YouTube video — not LR.
If you ever have a problem with wrong human translations (i.e. mistranslations) it’s a mistake on YouTube/Netflix’s part, not LR’s as they just scrape the data they find from the source for those subs/captions.
For machine translations LR does translate those themselves — using Microsoft translate, I believe. It means an AI of some sort translated the target language subtitles into whatever you selected as your translation language.
These types of translations result in more direct and less contextually-based translations — especially if the translation software used is not trained/advanced enough in the language it is pulling data from to translate.
I find that for contextual languages like Korean, these translations from LR tend to be not so good, so I avoid using these if I can — especially as a beginner.
I’ve had more success using DeepL and translators like Papago to do my machine translations outside of LR for languages like Korean — a.k.a. third party translation services designed for/highly trained on the target language I’m learning, which is just Korean right now.
I hope this explanation helps!
Note: for the human translations, if LR shouldn’t count auto-generated subs from YouTube as “human translations”.