It probably won’t make sense for LibreTranslate soon (since the appeal is running your own server) but I think the way to do the highest quality translation will be to use a very powerful remote language model fitted to your task.
Argos Translate currently supports doing few shot translation with the OpenAI API. It doesn’t work very well now but probably will with GPT4+. This could potentially be useful if people want the option of translating on a powerful remote and proprietary server or locally all through the LibreTranslate interface.
Yes without self-hosting (especially on low-powered hardware), the appeal of the project decreases significantly.