My old methods for creating AI voices involved locating the actor speaking with little to no background noise. So from an interview, audio book, etc.
Rogue Leader found out that by taking character voices from the center channel and running it through an AI background music/noise remover, that you can achieve the same thing while still having them speak in character.