youtube-transcript

Use yt-dlp with --write-sub for manual subtitles or --write-auto-sub for auto-generated captions, combined with --skip-download to get only the subtitle file. The result is a .vtt file that can be converted to plain text.

Parse the VTT file with a Python script that strips WEBVTT headers, timestamp lines, and HTML tags, then deduplicates repeated caption lines. Auto-generated YouTube captions overlap progressively, so deduplication is required for readable output.

Download the audio track with yt-dlp -x --audio-format mp3 and transcribe it locally using OpenAI Whisper with the base model. Confirm the audio file size with the user before downloading, since Whisper models require 1-3GB of disk space.

No, yt-dlp cannot access private videos and may fail on age-restricted or geo-blocked content without authentication. The tool reports the specific error, which should be relayed to the user.

Failures usually stem from network issues, insufficient disk space, or certificate problems. Retry with --no-check-certificate for SSL issues and verify the URL format matches https://www.youtube.com/watch?v=VIDEO_ID.