The catarrhine who invented a perpetual motion machine, by dreaming at night and devouring its own dreams through the day.
Being able to turn it off makes it even less cost-effective - because as soon as a player turns it off, the cost spent on voice acting is wasted on them.
Your mention of models brought me some idea though - text-to-speech could make this considerably cheaper, thus more viable. And if the voice actors are decent enough, they could be even used for multiple main cast chars, to bring costs down further.
I think that adding voice acting to older games, as in the proof of concept, is a bad idea. Players already have mental “images” on how characters are supposed to sound like, so there’s a high chance that even good voice acting would rub them off the wrong way.
The picture changes for newer games - like, it would be possible to develop new GBA games to use with an emulator. Then voice acting becomes a matter of cost vs. benefit - good quality voice acting tends to be expensive, but it makes wonders for immersion; while poor quality voice acting would probably make a game worse.
Just my two cents.
I believe that translation is partially to blame, since FE7 was the first Fire Emblem game translated to English.