Auto-Summarize Podcasts with Wikipedia Enrichment and Email Delivery in n8n

Podcast transcript → AI summary → topic & question extraction → Wikipedia research → HTML email

Who

Content operators or knowledge managers who need quick podcast digests delivered by email

You get

Turn a podcast transcript into a structured summary with topics, questions, and Wikipedia-enhanced explanations, then email it as HTML

You need

n8n · Gmail

Steps

1. Prepare transcript and trigger the workflow

   Open the template in n8n, locate the Podcast Episode Transcript node (type: Code), and replace the hardcoded sample transcript with your actual podcast text. Save, then click When clicking "Execute Workflow" to run it manually.

2. Configure the summarization chain

   The Summarize Transcript node (type: Summarization Chain) is set to documentLoader mode. It is wired to Default Data Loader, Recursive Character Text Splitter (chunkSize=6000, chunkOverlap=1000), and OpenAI Chat Model (gpt-4o-mini). You can usually run it as-is; for very long transcripts, adjust chunkSize and chunkOverlap. After running, this node produces $json.response.text as the summary.

3. Extract topics and questions

   The Extract Topics & Questions node (type: Information Extractor) receives the summary and uses gpt-4o-mini to output structured JSON with a questions array and a topics array.

   Execution prompt · Source text。Extract Topics & Questions node text parameter (information extractor input template)

   Copy prompt

   =Podcast Summary:  {{ $json.response.output_text }}

   Execution prompt · Source text。Extract Topics & Questions node options.systemPromptTemplate parameter

   Copy prompt

   =Come up with a list of questions and further topics to explore that are relevant for the context. Make sure questions are relevant to the topics but not verbatim. Think hard about what the appropriate questions should be and how it relates to the summarization.

4. AI Agent researches each topic

   The Topics node (Split Out) splits the questions array into individual items, feeding each into the AI Agent. The AI Agent uses gpt-4o-mini and has the Wikipedia tool attached, automatically retrieving and explaining each topic.

   Execution prompt · Source text。AI Agent node text parameter (agent per-item input prompt)

   Copy prompt

   =Question: {{ $json.question }}
   Why: {{ $json.why }}
   
   Context:  {{ $('Summarize Transcript').first().json.response.text }}

5. Format as HTML

   The Format topic text & title node (type: Code) combines AI Agent outputs, the summary, and the original questions list into HTML fragments. It uses JavaScript to read upstream node data and returns { topics, summary, questions }.

   Execution prompt · Source text。Format topic text & title node jsCode parameter

   Copy prompt

   const inputItems = $input.all();
   const topics = [];
   const questions = [];
   const summary = $('Summarize Transcript').first().json.response.text;
   // Format Topics
   for (const [index, topic] of inputItems.entries()) {
     const title = $('Topics').all()[index].json.question
   
     topics.push(`
       <h3>${title}</h3>
       <p>${topic.json.output}</p>`.trim()
     )
   }
   
   // Format Questions
   for (const question of $('Extract Topics & Questions').first().json.output.questions) {
     questions.push(`
       <h3>${question.question}</h3>
       <p>${question.why}</p>`.trim()
     )
   }
   
   return { topics, summary, questions }

6. Send email

   The Send Digest node (type: Gmail) assembles the HTML fragments into an email body. Change sendTo from [email protected] to the real recipient, and confirm the gmailOAuth2 credential is configured. Subject is Podcast Digest.

The workflow automates the process of creating a summarized and enriched podcast digest, which is then sent via email.

Note that to use this template, you need to be on n8n version 1.19.4 or later.