How do I install the Youtube Captions skill?

Run npx skillmds@latest add third774/youtube-captions in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Youtube Captions safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Youtube Captions?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Youtube Captions free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Youtube Captions?

third774 (@third774) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Youtube Captions by third774 — Agent Skill

Name: Youtube Captions
Author: third774

Youtube Captions

Extract captions and transcripts from YouTube videos for agent context. Tries manual subtitles, then auto-generated, then falls back to audio transcription via Whisper. Use when a user provides a YouTube URL and wants to understand, summarize, reference, or search video content.

third774 443d86c 2 files · 4.3 KB Updated Sep 17, 2026

File contents

youtube-captions

Extract timestamped captions from YouTube videos. Output is VTT format (timestamps preserved).

Prerequisites

yt-dlp — REQUIRED (brew install yt-dlp)
openai-whisper — REQUIRED only if video has no subtitles (pip install openai-whisper)
- First run downloads the small model (~500MB)
- Transcription is significantly slower than subtitle download

Usage

bash scripts/get-captions.sh <youtube-url> [language]

youtube-url — any valid YouTube video URL
language — subtitle language code (default: en)

Output goes to stdout. Status messages go to stderr.

Fallback Chain

Manual subtitles — human-uploaded captions (fastest, most accurate)
Auto-generated subtitles — YouTube's speech recognition
Whisper transcription — downloads audio, transcribes locally with whisper --model small

The script tries each step in order and exits on the first success.

Example

# Get English captions
bash scripts/get-captions.sh "https://www.youtube.com/watch?v=dQw4w9WgXcQ"

# Get Spanish captions
bash scripts/get-captions.sh "https://www.youtube.com/watch?v=dQw4w9WgXcQ" es

Output Format

VTT (Web Video Text Tracks) with timestamps:

WEBVTT

00:00:01.000 --> 00:00:04.000
First line of dialogue

00:00:04.000 --> 00:00:08.000
Second line of dialogue

Notes

All temporary files (audio, intermediate subtitle files) are cleaned up automatically
If neither yt-dlp subtitles nor Whisper are available, the script exits with an error and clear instructions
Long videos with no subtitles will take time to transcribe — Whisper processes roughly at 1x realtime on CPU

---name: youtube-captionsdescription: Extract captions and transcripts from YouTube videos for agent context. Tries manual subtitles, then auto-generated, then falls back to audio transcription via Whisper. Use when a user provides a YouTube URL and wants to understand, summarize, reference, or search video content.---# youtube-captionsExtract timestamped captions from YouTube videos. Output is VTT format (timestamps preserved).## Prerequisites- `yt-dlp` — REQUIRED (`brew install yt-dlp`)- `openai-whisper` — REQUIRED only if video has no subtitles (`pip install openai-whisper`) - First run downloads the `small` model (~500MB) - Transcription is significantly slower than subtitle download## Usage```bashbash scripts/get-captions.sh <youtube-url> [language]```- `youtube-url` — any valid YouTube video URL- `language` — subtitle language code (default: `en`)Output goes to stdout. Status messages go to stderr.## Fallback Chain1. **Manual subtitles** — human-uploaded captions (fastest, most accurate)2. **Auto-generated subtitles** — YouTube's speech recognition3. **Whisper transcription** — downloads audio, transcribes locally with `whisper --model small`The script tries each step in order and exits on the first success.## Example```bash# Get English captionsbash scripts/get-captions.sh "https://www.youtube.com/watch?v=dQw4w9WgXcQ"# Get Spanish captionsbash scripts/get-captions.sh "https://www.youtube.com/watch?v=dQw4w9WgXcQ" es```## Output FormatVTT (Web Video Text Tracks) with timestamps:```WEBVTT00:00:01.000 --> 00:00:04.000First line of dialogue00:00:04.000 --> 00:00:08.000Second line of dialogue```## Notes- All temporary files (audio, intermediate subtitle files) are cleaned up automatically- If neither yt-dlp subtitles nor Whisper are available, the script exits with an error and clear instructions- Long videos with no subtitles will take time to transcribe — Whisper processes roughly at 1x realtime on CPU

Frequently asked questions