Seems like the voices are text to speech? You could have an option to change the voice type as well as regenerate different scripts with a simple llm.