Maybe a good balance could be a human voice actor for main dialogue, supported by ai trained on their main dialogue to voice sidequest and deep lore dialogue. It could enable fully voiced dialogue-heavy games that would otherwise be too expensive to produce, something like generative RPGs or Morrowind, if all the books could be voiced, and more easily translated while remaining fully voiced. But keeping humans to fill the main campaign contributions, emotional beats and determine character personality. I’m just comparing Morrowind to Oblivion, which was voiced, but the dialogue and conversation trees were heavily reduced in volume as a result.

Maybe a good balance would be hiring a person to speak lines into a microphone. It could employ a person and create art with an acceptable bare minimum quality standard. If you can’t afford that and would rather push the costs onto government subsidies for power and emissions, maybe just do text dialogue or pull a classic Banjo and Kazooie single dialogue line randomly jumbled up and pitch shifted for every interaction.

Using a robotic voice could make the game more accessible to blind, partially sighted, and dyslexic individuals. I’m not sure how an AI voice is inherently different than the voice that comes out of a screen reader, especially if it’s trained on the voice of employees or volunteers.