- Home
- Skills
- Michalparkola
- Tapestry Skills For Claude Code
- Youtube Transcript
youtube-transcript_skill
- Shell
194
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill michalparkola/tapestry-skills-for-claude-code --skill youtube-transcript- SKILL.md13.4 KB
Overview
This skill downloads transcripts, captions, or subtitles from YouTube videos when given a YouTube URL or when asked to transcribe a video. It prefers human-created subtitles, falls back to auto-generated captions, and offers an optional Whisper-based audio transcription if no subtitles exist. The workflow reports progress, asks for confirmations for large downloads, and can output both VTT and cleaned plain-text transcripts.
How this skill works
It first verifies that yt-dlp is installed and lists available subtitle tracks for the provided YouTube URL. It attempts to download manual subtitles (--write-sub), falls back to auto-generated captions (--write-auto-sub), and, only if none are available, offers to download the audio and run Whisper for transcription. After download it confirms file locations and can convert VTT to deduplicated plain text for easy reading or analysis.
When to use it
- User provides a YouTube URL and requests the transcript
- User asks to download captions or subtitles from a YouTube video
- User requests a transcription of a video lacking subtitles
- You need a VTT file with timestamps for a video player
- You want a cleaned plain-text transcript for reading or text analysis
Best practices
- Always run yt-dlp --list-subs first to see available subtitle types and languages
- Prefer --write-sub (manual) over --write-auto-sub (auto-generated) for quality
- Ask for user confirmation before downloading large audio files or installing Whisper/models
- Sanitize video titles when creating filenames to avoid filesystem issues
- Deduplicate and strip VTT timing/HTML when converting to plain text to preserve speech order
Example use cases
- Save a human-created English subtitle .vtt for a lecture video for use in a video player
- Fetch auto-generated captions to get a quick searchable transcript for research
- Transcribe a podcast-style YouTube video with Whisper when no subtitles exist (asks first)
- Convert a downloaded .vtt into a deduplicated .txt file for summarization or indexing
- List available subtitle languages to choose a specific language track before downloading
FAQ
The skill attempts a platform-appropriate install (Homebrew, apt, or pip) and informs you if manual installation is required with links and commands.
When is Whisper used and what does it require?
Whisper is a last-resort option when no subtitles exist. It requires installing the openai-whisper package and models (roughly 1–10 GB depending on model) and explicit user approval before large downloads or installs.