the voices are stored in the esm then added to from the sexlab extension and the dwarven suits plugin. with about 1000 dialogs in already, if each voice type has an equal share of that, you'll need to add about 200 dialogs for each new voice type you add then those new voices need to be registered in the bootup and reboot quests using the pahboot script and pahcore script properties - the properties are added to the quests from the ck once you have enough dialogs to get a functional voice type.