Random shower thought this morning:
On Overdrive (and other Library ebook/audiobook lending systems) right now, pretty much all of the Harry Potter books are available for free in more than a dozen languages, in both audiobook and ebook formats. It wouldn’t be that much work to build out an NLP model that took them both and built out timecode/transcription metadata for the audio file and tie it to the text. You could then write a bit more code to interleave the audio from two different languages, so that each sentence is said first in one language, then another, reading the whole book this way.
How amazing would that be for kids in foreign language classes, or for ESL learners?