Auto-summarize YouTube videos with n8n and GPT-4o-mini

YouTube URL in → transcript extraction → GPT-4o-mini structured Markdown summary → JSON response with optional Telegram push

Who

Researchers, students, and content operators who need to digest video content quickly

You get

POST a YouTube URL and receive a structured Markdown summary with topics, key points, and technical terms within seconds, plus a Telegram notification

You need

n8n · youtube-transcript · telegram

Steps

1. Receive URL and extract video ID

   After importing the template, confirm the Webhook node path is ytube, method is POST, and response mode is responseNode. The Get YouTube URL node reads youtubeUrl from the webhook body. The YouTube Video ID node runs this JavaScript to extract the 11-character video ID from youtu.be or youtube.com links:

   Execution prompt · Source text。YouTube Video ID node jsCode field

   Copy prompt

   const extractYoutubeId = (url) => {
     // Regex pattern that matches both youtu.be and youtube.com URLs
     const pattern = /(?:youtube\.com\/(?:[^\/]+\/.+\/|(?:v|e(?:mbed)?)\/|.*[?&]v=)|youtu\.be\/)([^"&?\/\s]{11})/;
     const match = url.match(pattern);
     return match ? match[1] : null;
   };
   
   // Input URL from previous node
   const youtubeUrl = items[0].json.youtubeUrl; // Adjust this based on your workflow
   
   // Process the URL and return the video ID
   return [{
     json: {
       videoId: extractYoutubeId(youtubeUrl)
     }
   }];

2. Fetch video metadata and full transcript

   The Get YouTube Video node calls the YouTube Data API with the videoId to retrieve title, description, and other metadata. The YouTube Transcript community node extracts the full transcript text. Note: this is a community node and only works on self-hosted n8n.

3. Concatenate transcript segments

   The Split Out node splits the transcript array into individual records. The Concatenate node uses summarize aggregation with concatenate to join all text fields with spaces, producing a single concatenated_text string for the downstream LLM node.

4. Generate structured summary with GPT-4o-mini

   The Summarize & Analyze Transcript node (Basic LLM Chain) connects to the gpt-4o-mini model with promptType set to define. The prompt instructs the model to break content into Level-2 topics, use bullet points, preserve technical accuracy, and organize by Definition/Background → Main characteristics → Implementation details → Advantages/Disadvantages.

   Execution prompt · Source text。Summarize & Analyze Transcript node text field (promptType=define)

   Copy prompt

   =Please analyze the given text and create a structured summary following these guidelines:
   
   1. Break down the content into main topics using Level 2 headers (##)
   2. Under each header:
      - List only the most essential concepts and key points
      - Use bullet points for clarity
      - Keep explanations concise
      - Preserve technical accuracy
      - Highlight key terms in bold
   3. Organize the information in this sequence:
      - Definition/Background
      - Main characteristics
      - Implementation details
      - Advantages/Disadvantages
   4. Format requirements:
      - Use markdown formatting
      - Keep bullet points simple (no nesting)
      - Bold important terms using **term**
      - Use tables for comparisons
      - Include relevant technical details
   
   Please provide a clear, structured summary that captures the core concepts while maintaining technical accuracy.
   
   Here is the text: {{ $json.concatenated_text
    }}

5. Build response object

   The Response Object node (Edit Fields/Set) assembles the final JSON with: summary (from LLM output text), topics (empty array), title (from YouTube snippet.title), description (snippet.description), id (video ID), and youtubeUrl (original input URL).

6. Return webhook response and push Telegram

   The Respond to Webhook node returns the Response Object as the HTTP response. The Telegram node simultaneously sends a message to the configured chat with the video title and original URL, parse_mode set to HTML, and appendAttribution set to false.

   Execution prompt · Source text。Telegram node text field

   Copy prompt

   ={{ $json.title }}
   {{ $json.youtubeUrl }}

You are an n8n workflow setup assistant. The user wants to recreate the "AI-powered YouTube video summarization & analysis" template. Assist in this order:
1. Confirm the user runs self-hosted n8n (community node restriction).
2. Check that three credentials are configured: YouTube Data API v3, OpenAI API Key, and Telegram Bot Token (if using Telegram).
3. Guide the user to import the workflow JSON from https://n8n.io/workflows/2679/ or install from the template library.
4. Verify the Webhook path (ytube) and response mode (responseNode).
5. Suggest testing with a public YouTube URL to validate transcript extraction and LLM summary output format.