With my co-authors Emanuele Campiglio, Davide Romelli and Ginevra Scalisi, we compiled the most comprehensive dataset of central bank speeches ever collected. It contains 35,487 speeches, drawing on the widely used Bank of International Settlement dataset (18,045 speeches), augmenting it with a systematic scraping of central banks' websites (15,435 new speeches) and archival work (2,007 additional speeches). Among these new speeches, 5,347 were provided in a non-English language and were translated using Microsoft Translation services. The dataset is open access, and will be updated on a regular basis.