My understanding is you're looking for landmarks in the audio and then matching/aligning on those.

Shenidam is open source software which does the whole process, site offline, GitHub still up: https://web.archive.org/web/20160309034204/http://shenidam.o... and https://github.com/stendardo/shenidam

A node stream module: https://github.com/adblockradio/stream-audio-fingerprint

Python: https://github.com/dpwe/audfprint

A blog post explaining fingerprinting: https://willdrevo.com/fingerprinting-and-audio-recognition-w...

Commercial software: https://www.redgiant.com/products/shooter-pluraleyes/