Skip to content
Swap Meet

Free

No. 027Free

YT-Condense: a long YouTube video becomes a shorter listen

A YouTube link comes back as spoken prose at about a fifth of the length, ready to read or to play in a text-to-speech reader. The same prompt works on a pasted transcript.

by Lucas Cooper-Bey

About it

Give YT-Condense a YouTube link and it writes spoken prose at about a fifth of the length. Captions are used when they exist; otherwise it transcribes on your machine. --provider none writes a file you paste into any chat, and calls no model. The prompt in the file also works on a transcript you paste yourself. The post says a fifth of the length is not a promise of listening time. Limits: A model can drop or bend a fact while it shortens, so check that part of the video before you act on it. The length target is about a fifth, not a promise of listening time. Private, removed, age-gated and region-locked videos cannot be read. Transcribing a video with no captions takes a few minutes, and the first download of the speech model is about 460 MB, or 75 MB for tiny. --speak is for a Mac. Door 26 of 100 Doors.