Other (older) games bridge this gap partially well by having a broad variety of voice sets to choose from, and so it's baked into your chosen voice pack whether they might shout "I'll stop you if I have to!" or "Blood makes the grass grow!" when you're about to violence someone.
I can't say how well that approach would stand up in a modern high-detail game, of course, but I do feel that rather than having our MC voice every possible line, we'd be better severed, in terms of game immersion, by having an extensive selection of 'sets' to choose from so that our character sounds consistent to the way we want them to sound, in general, for most of the gameplay experience, even if their specific lines in dialogue scenes are selected silently, and we are presumed to say them in the manner that we imagine.
