Source: https://elevenlabs.io/press
As far as the voices go, I have a good friend that uses Eleven Labs voices for some of his D&D campaigns and he's very satisfied with the results.
This could make it really easy to scale up a variety of voices across all of the different environments. For example, while the same thing may be said in one hospital vs another, you could have a different voice in every location or even based on day/night cycles.
You wouldn't even be limited to having them say the exact same things, you could script locations to say slightly different things and the model would automatically adjust the voice for that.
Considering the business model of Eleven Labs, I wouldn't expect anything to be made on the fly, this would still just generate a bunch of audio files that they would then incorporate into the game.
