....except flexibility in story line, branching progression, customization options, paying Actors Guild rates for each voice actor who is a member (and you might be surprised to find out about some of the contractual clauses in things like that), multiple story paths, over investment in big names, etc.
Synthetic speech is nowhere -near- capable of bridging the uncanny valley. That option is not on the table except for a science fiction game where everyone is =expected= to sound like a robot. And you can't program synth speech in words; you have to be able to break the word into distinct phonemes that the synthesis software recognizes, and stitch those strings of phonemes together, changing emphasis as you go, to even get close. And since the characters are modeled, you would have to have the appropriate visemes animated and articulated so that the facial structure move correctly......otherwise the uncanny valley will swallow you and your efforts whole.
From a strict audio perspective, there are a lot of atmospherics that are not present. Notice that you don't hear echoes? The larger jarl chambers are all natural echo chambers......and not one reverberation. The same goes for many of the dungeons; sword on sword sounds exactly the same if in a narrow, long tunnel or a large natural cavern. If water is beneath you or stone or soft earth. All that applies to voices, as well.
Well spoken, well acted dialogue has its place......but at this stage, only in the mouths of significant characters; otherwise, it is nothing more than a gimmick that limits the storytelling ability of text based dialogue, and assures that released content can't really be added to or tweaked out with added options.