Hi Bonnie,
Welcome & glad to see you here! That's an interesting idea... do you have a candidate algorithm in mind?
I'm thinking that the first sustained audio peak to cross a certain threshold will be pretty unreliable (depending on how trained the singer is, and whether the parts start at the same time. Also, the naive algorithm will probably get thrown off quite easily by background noise...).
Maybe it's good to try a proof of concept in python / a bash script first to check feasibility.
Anyways, in the meantime perhaps your friend would find success in trying to use the site herself. Having done a number of "1-person choir" projects myself on the site, I can say that it's the easiest way that I've found to do it so far :)
Allan