Open jzohrab opened 11 months ago
I love this idea. I recently got into using whisper for subtitle generation. I think this feature would be super useful for working through audiobooks and podcasts.
I'll try to look into the all the audio stuff sometime soon, but my gut says it would make a lot more sense to persist the timestamps in another table, so you could easily jump to anywhere in a longer audio according to the the associated subtitle.
I'm really not sure how to link up audio timestamps with arbitrary sentences. With Text
database entries, it's possible, but sentences aren't created until the page is first read.
The current audio timestamps are stored as denormalized data in the books
table, as BkAudioBookmarks TEXT NULL
. These are just shown on the player as bookmark bars, with no reference to the texts
entries in the book. There's no tie between the texts
entries and the bookmarks, since the two aspects of text and sound/player are totally decoupled. Having them automatically be in sync -- e.g. so that turning a page automatically changes the player, or fast-forwarding changes your current page -- would be quite hard for the user to manage, I think!
But maybe you have a clever idea that makes sense for you. I've never needed this idea, and so haven't put the hours into thinking about it.
From Discord:
Challenges I can see with this request: