Hacker News Reader

That’s pretty cool, how did you collect all the caption data?

With youtube API. It provide captions already split, with timestamps. I store them in db and make search with ElasticSearch.