Auto-Repurpose YouTube Videos into Social Posts and AI Thumbnails with n8n

Input YouTube URL → extract transcript and metadata → OpenAI generates LinkedIn/Twitter/blog/YouTube description + AI thumbnail → store in Notion database

Who

Content operators, social media managers, YouTube creators

You get

Send one YouTube link, get a LinkedIn post, Twitter thread, blog draft, YouTube description, and a 1024×1024 AI thumbnail, all archived in Notion

You need

n8n · google-drive · youtube · apify · notion

Steps

1. Receive YouTube URL and Extract Video ID

   Configure a Webhook node in n8n (POST method, path yt-repurpose) to receive a JSON body containing a `url` field. A downstream Code node splits the URL to extract the videoId after `v=`.

2. Fetch Video Details and Transcript

   The YouTube node retrieves snippet/topicDetails/contentDetails using the videoId. An Apify Actor (benthepythondev/youtube-transcript-scraper) fetches the video captions. A Code node cleans the transcript: replaces HTML entities (e.g. '→'), removes [♪…♪] music markers, and collapses whitespace.

   Execution prompt · Source text。Clean Transcript Text node JS Code field

   Copy prompt

   const transcript = $json.transcript;
   
   // remove HTML entities like '
   const clean = transcript
     .replace(/'/g, "'")
     .replace(/\[♪.*?♪\]/g, "")
     .replace(/\s+/g, " ")
     .trim();
   
   return [{
     transcript: clean
   }];

3. Generate Social Media Content and Thumbnail Prompt

   The cleaned transcript is sent in parallel to two OpenAI nodes. The first uses gpt-4.1-mini to generate a JSON with LinkedIn post, Twitter thread, blog draft, and YouTube description. The second uses gpt-4.1-mini to generate an English thumbnail prompt based on the video title and transcript.

   Execution prompt · Source text。Generate Social Media Content node Responses > Content field

   Copy prompt

   =You are a professional content repurposing assistant.
   
   Convert this YouTube transcript into:
   
   {
     "linkedin_post": "",
     "twitter_thread": "",
     "blog": "",
     "youtube_description": ""
   }
   
   Transcript:
   {{$json.transcript}}

   Execution prompt · Source text。Create Thumbnail Prompt node Responses > Content field

   Copy prompt

   =Create a professional YouTube thumbnail prompt for this video.
   
   Video Title:
   {{$('Fetch YouTube Video Details').first().json.snippet.title}}
   
   Transcript: {{ $('Clean Transcript Text').item.json.transcript }}
   
   
   Rules:
   - Make it visually attractive
   - Modern YouTube style
   - Bold text
   - High contrast
   - Cinematic lighting
   - AI art style
   - Return only the image prompt

4. Extract and Structure AI Output

   Code nodes extract text content from OpenAI responses, then JSON.parse into structured fields (linkedin_post, twitter_thread, blog, youtube_description). The thumbnail prompt is similarly extracted as plain text via a Code node.

   Execution prompt · Source text。Extract AI Response / Extract Thumbnail Prompt node JS Code field

   Copy prompt

   return [
     {
       content: $json.output[0].content[0].text
     }
   ];

   Execution prompt · Source text。Convert AI Response to JSON node JS Code field

   Copy prompt

   const data = JSON.parse($json.content);
   
   return [
     {
       json: {
         linkedin_post: data.linkedin_post,
         twitter_thread: data.twitter_thread,
         blog: data.blog,
         youtube_description: data.youtube_description
       }
     }
   ];

5. Generate AI Thumbnail and Upload to Google Drive

   The OpenAI image generation node (gpt-image-1, 1024×1024) generates the thumbnail from the prompt. It is uploaded to a Google Drive folder, shared publicly with anyone/reader permission, and the webViewLink is extracted as the public URL.

6. Aggregate Data and Save to Notion

   A Code node aggregates: video title, video URL, first 1900 characters of transcript, AI-generated content, and ISO timestamp. The Notion node creates a database page with title parameter fixed as 'n8n'; field mappings must strictly match database properties, noting the leading space in ' Youtubedescription' and the Timestamp field using timezone rather than date value.

   Execution prompt · Source text。Prepare Final Content Data node JS Code field

   Copy prompt

   return [
     {
       json: {
         video_title: $('Fetch YouTube Video Details').first().json.snippet.title,
   
         video_url:
           'https://www.youtube.com/watch?v=' +
           $('Fetch YouTube Video Details').first().json.id,
   
         transcript:
           $('Fetch Video Transcript').first().json.transcript.substring(0, 1900),
   
         linkedin_post: $json.linkedin_post,
   
         twitter_thread: $json.twitter_thread,
   
         blog: $json.blog,
   
         youtube_description: $json.youtube_description,
   
         timestamp: new Date().toISOString()
       }
     }
   ];

You are an n8n workflow setup assistant. The user wants to replicate the "YouTube video auto-repurposing + AI thumbnail" template. First confirm they have: an n8n instance, YouTube Data API credentials, OpenAI API key (with image generation access), Apify account, and OAuth credentials for Google Drive and Notion. Then guide them to import workflow.json, configure each node's credentials, create the Notion database (fields must match exactly, note the leading space in ' Youtubedescription'), and test the webhook.