About the project
Paste a link to a video from YouTube, TikTok, X or any of about 1,800 other sites and download its transcript as a text file, a searchable web page or subtitles, with timestamps that jump back into the video. It uses the video’s own captions when there are any, which takes about three seconds, and otherwise transcribes the speech with Whisper: a 14-minute talk takes 91 seconds. It is public and costs nothing to run. The page and a small router live on Cloudflare Pages, and the work is done by two home computers that each open a free tunnel and register it with the router: the Mac first, and the Windows PC when the Mac is off. Because strangers are using a home connection, links to private addresses are refused and every visitor is rate-limited.