The package uses an external text-to-speech engine (like flite) to pronounce the word at or near point, the selected region or a whole buffer. M-x read-aloud-this-- pronounce either the selection or a ...
I found a site where you can listen to people read the sentences in different accents and tried to see if it was possible to hear a series of lines aloud without them gathering meaning.