As much as a good idea as it could be, technically, at this stage, I do not think it is a good venture. If it were some earlier phase when you could implement this, code wise, without it being a huge undertaking, I would think yeah, perfect idea. Have the pre-generated names have spoken lines. Or better yet, build some tech into the system that would take typed names, maybe asked for some pronunciation, and used voice samples and AI to build pronunciation of the name you enter, to make it sound natural.
