Any wild <snip> ideas as to how I could start to approach this?

In two steps. User1 types the lyrics into an editor and presses a button which starts a player and then a clock as soon as that player produces "now producing audio" output. User1 then then hits the spacebar throughout the song.

Later, User2 with the same audio, a similar enough player, and User1's data gets a scrolling line of text as the song plays.

As for the program? At first it might just read a line from the lyric file and produce a "$milliseconds: o/~ I put my trust in you o/~" line. Later the user interface might change and the program might produce "$milliseconds: o/~ I put my trust in you o/~\n$milliseconds+1: o/` I o/`\n$milliseconds+1200: o/` put o/`". Later you might find that the best solution makes use of something visual with floating text and random-access audio play.

Most of your hard work may be making things easier for user1. But you can still do much with a tedious first pass.