Openakita Skills Youtube Summarizer

pip3 install youtube-transcript-api

Verify installation:

from youtube_transcript_api import YouTubeTranscriptApi
print("youtube-transcript-api is ready")

Supported Python Versions

• Python 3.8+
• No additional system dependencies required

Instructions

Step 1: Extract the Video ID

YouTube URLs come in many formats. Parse all of them to extract the 11-character video ID:

import re
from urllib.parse import urlparse, parse_qs

def extract_video_id(url_or_id: str) -> str:
    """Extract YouTube video ID from various URL formats."""
    url_or_id = url_or_id.strip()

    # Bare video ID (11 chars, alphanumeric + dash + underscore)
    if re.match(r'^[A-Za-z0-9_-]{11}$', url_or_id):
        return url_or_id

    parsed = urlparse(url_or_id)

    # youtu.be short links
    if parsed.hostname in ('youtu.be',):
        return parsed.path.lstrip('/')

    # Standard, mobile, embed, live, shorts
    if parsed.hostname in ('www.youtube.com', 'youtube.com', 'm.youtube.com'):
        if parsed.path == '/watch':
            qs = parse_qs(parsed.query)
            return qs.get('v', [None])[0]
        for prefix in ('/embed/', '/live/', '/shorts/', '/v/'):
            if parsed.path.startswith(prefix):
                return parsed.path[len(prefix):].split('/')[0]

    raise ValueError(f"Cannot extract video ID from: {url_or_id}")

Step 2: Fetch the Transcript

from youtube_transcript_api import YouTubeTranscriptApi

def fetch_transcript(video_id: str, preferred_langs=None):
    """
    Fetch transcript with language fallback.

    Args:
        video_id: The 11-character YouTube video ID
        preferred_langs: Ordered list of language codes, e.g. ['zh-Hans', 'zh', 'en']
                        Defaults to ['en'] if not specified

    Returns:
        List of transcript segments with 'text', 'start', 'duration' keys
    """
    if preferred_langs is None:
        preferred_langs = ['en']

    try:
        transcript_list = YouTubeTranscriptApi.list_transcripts(video_id)

        # Try manually created transcripts first (more accurate)
        for lang in preferred_langs:
            try:
                transcript = transcript_list.find_manually_created_transcript([lang])
                return transcript.fetch()
            except Exception:
                continue

        # Fall back to auto-generated transcripts
        for lang in preferred_langs:
            try:
                transcript = transcript_list.find_generated_transcript([lang])
                return transcript.fetch()
            except Exception:
                continue

        # Last resort: get any available transcript and translate
        for transcript in transcript_list:
            try:
                if 'en' in preferred_langs:
                    return transcript.translate('en').fetch()
                return transcript.fetch()
            except Exception:
                continue

        raise RuntimeError("No transcript available in any language")

    except Exception as e:
        raise RuntimeError(f"Failed to fetch transcript for {video_id}: {e}")

Step 3: Generate the Summary

Use the STAR + R-I-S-E combined framework for maximum clarity:

STAR Framework (for factual/process content)

• Situation — What context does the video establish?
• Task — What problem or question is being addressed?
• Action — What steps, methods, or arguments are presented?
• Result — What conclusions, outcomes, or calls to action are given?

R-I-S-E Framework (for opinion/analysis content)

• Reference — What sources, data, or authorities are cited?
• Insight — What unique perspectives or non-obvious points are made?
• Summary — What are the essential takeaways?
• Evaluation — What are the strengths and weaknesses of the arguments?

Prompt Template for Summary Generation

When generating the summary from the transcript, use this structure:

You are summarizing a YouTube video transcript. Follow these rules strictly:

1. TITLE & METADATA
   - Video title (if known)
   - Estimated duration from timestamps
   - Primary topic/category

2. EXECUTIVE SUMMARY (2-3 sentences)
   - The single most important takeaway
   - Who should watch this and why

3. KEY POINTS (bulleted, with timestamps)
   - [MM:SS] Point description
   - Maximum 10 key points
   - Each point should be self-contained and actionable

4. DETAILED NOTES (organized by topic/section)
   - Group related content under headers
   - Include specific examples, numbers, quotes
   - Preserve technical accuracy

5. STAR ANALYSIS
   - Situation: ...
   - Task: ...
   - Action: ...
   - Result: ...

6. ACTIONABLE TAKEAWAYS
   - What the viewer should do after watching
   - Resources mentioned in the video

Workflows

Workflow 1: Quick Summary

User says: "Summarize this video: https://youtube.com/watch?v=..."

1. Extract video ID from URL
2. Fetch English transcript (fall back to auto-generated)
3. Generate executive summary + key points with timestamps
4. Present in clean markdown format

Workflow 2: Deep Dive Notes

User says: "I need detailed study notes from this lecture"

1. Extract video ID
2. Fetch transcript in preferred language
3. Generate full STAR + R-I-S-E analysis
4. Include all timestamps, quotes, and specific data points
5. Add a "Review Questions" section at the end

Workflow 3: Multi-Language Summary

User says: "Summarize this Chinese/Japanese/Korean video in English"

1. Extract video ID
2. List available transcript languages
3. Fetch best available transcript
4. Translate content if needed (use transcript API's built-in translation)
5. Generate summary in the user's preferred language

Workflow 4: Batch Processing

User provides multiple URLs:

1. Extract all video IDs
2. Fetch transcripts sequentially (respect rate limits)
3. Generate individual summaries
4. Create a comparison table if videos are related
5. Highlight common themes and unique points

Workflow 5: Timestamp Navigation

User says: "Find where they talk about X in this video"

1. Fetch full transcript
2. Search for relevant keywords/concepts
3. Return matching segments with precise timestamps
4. Format as clickable timestamp links: [MM:SS](https://youtu.be/ID?t=SECONDS)

Output Format

Standard Summary Output

# 📺 Video Summary: [Title]

**Duration**: ~XX minutes | **Language**: English | **Category**: Technology

## Executive Summary
[2-3 sentence overview]

## Key Points
- **[0:00]** Introduction and context setting
- **[2:15]** First major point with supporting evidence
- **[8:30]** Key technical concept explained
- **[15:00]** Practical demonstration
- **[22:45]** Conclusion and call to action

## Detailed Notes

### Section 1: [Topic] (0:00 - 5:30)
[Detailed notes with specific information]

### Section 2: [Topic] (5:30 - 12:00)
[Detailed notes with specific information]

## STAR Analysis
- **Situation**: [Context]
- **Task**: [Problem/Question]
- **Action**: [Methods/Steps]
- **Result**: [Outcomes/Conclusions]

## Actionable Takeaways
1. [First action item]
2. [Second action item]
3. [Third action item]

## Resources Mentioned
- [Resource 1](url)
- [Resource 2](url)

Common Pitfalls

1. Transcript Unavailable

Problem: Some videos have no transcript (auto-generated or manual).

Solutions:

• Check if the video is a music video, live stream, or very short clip — these often lack transcripts
• Try different language codes: ['en', 'en-US', 'en-GB']
• If auto-generated is disabled by the uploader, there's no workaround via this API

Error handling: