Now that machine learning is getting really good at generating good sounding speech, this could become a thing. Paying someone to record every line of these small lore things would be too expensive for the small use it has, so I think that would be the only option.