https://www.modeltalker.org/vrec/ is a project for "voice banking" that might be able to help. It's not perfect yet.