YouTube Transcript Analyzer Skill

Description

This skill fetches transcripts from YouTube videos and provides structured analysis including summaries, key points, topic segmentation, and timestamp-indexed content. It supports multiple languages and auto-generated captions. Perfect for researchers, content creators repurposing video content, and students processing lecture recordings.

Instructions

1. Fetch Transcript: Extract the video ID from the URL and retrieve the transcript
2. Process: Clean up auto-generated caption artifacts (timing glitches, repeated words)
3. Analyze: Generate summaries, extract key points, identify topic transitions
4. Format: Output in the requested format (summary, full transcript, timestamped notes)

Rules

• Always attempt auto-generated captions if manual captions are unavailable
• Clean up common auto-caption artifacts: duplicated words, missing punctuation, misheard terms
• Preserve timestamps for key topic transitions
• Support multiple output formats: summary, chapter markers, full transcript, Q&A extraction
• When summarizing, maintain the original speaker's intent without adding interpretation
• Flag sections where the transcript quality is poor (low confidence, unintelligible)
• Handle multi-language transcripts by specifying language preference

Transcript Fetching

Using yt-dlp (Recommended)

Copy
# Fetch transcript as JSON with timestamps
yt-dlp --write-auto-sub --sub-lang en --skip-download \
  --sub-format json3 -o "transcript" "VIDEO_URL"

# Fetch manual subtitles if available
yt-dlp --write-sub --sub-lang en --skip-download \
  --sub-format srt -o "transcript" "VIDEO_URL"

# List available subtitle languages
yt-dlp --list-subs "VIDEO_URL"

Using youtube-transcript-api (Python)

Copy
from youtube_transcript_api import YouTubeTranscriptApi

def get_transcript(video_id: str, lang: str = "en") -> list:
    try:
        # Try manual captions first
        transcript = YouTubeTranscriptApi.get_transcript(video_id, languages=[lang])
    except:
        # Fall back to auto-generated
        transcript = YouTubeTranscriptApi.get_transcript(
            video_id, languages=[f"{lang}-auto", lang]
        )
    return transcript

# Each entry: {"text": "...", "start": 12.5, "duration": 3.2}

Output Formats

Executive Summary

Copy
# Video Summary: [Title]
**Channel:** [Channel Name] | **Duration:** [Length] | **Published:** [Date]

## TL;DR
[2-3 sentence summary]

## Key Points
1. [Point 1] (timestamp: 2:34)
2. [Point 2] (timestamp: 8:15)
3. [Point 3] (timestamp: 14:02)

## Topic Segments
| Start | End | Topic |
|-------|-----|-------|
| 0:00 | 2:30 | Introduction and context |
| 2:30 | 8:00 | Main argument: [topic] |
| 8:00 | 14:00 | Supporting evidence |
| 14:00 | 18:30 | Practical applications |
| 18:30 | 20:00 | Q&A and conclusion |

Chapter Markers

0:00 Introduction
2:34 What is [Topic]?
5:12 Three core principles
8:45 Real-world example: [Company]
12:30 Common mistakes to avoid
16:00 Step-by-step implementation
19:45 Summary and resources

Timestamped Notes

Copy
## Notes: [Video Title]

### [0:00-2:30] Introduction
- Speaker introduces the topic of [X]
- Key stat mentioned: "[quote]" (1:15)

### [2:30-8:00] Core Concepts
- **Concept 1**: [description] (3:02)
- **Concept 2**: [description] (5:30)
- Important quote: "[exact quote]" (6:45)

Examples

"Summarize this YouTube video: https://youtube.com/watch?v=abc123"
"Extract chapter markers from this lecture video"
"Get the full transcript with timestamps for this tutorial"
"What are the key takeaways from this conference talk?"
"Create study notes from this 2-hour lecture"