Turn long episode transcripts into a clear research digest sent by Gmail. Ideal for content teams, podcasters, and marketers who need quick summaries with trusted references.

The system starts on click, takes a transcript from an input node, and splits it into readable chunks. An AI model creates a refine style summary across all chunks. Another model pulls topics and questions in a fixed schema so the output is clean and consistent. A research agent then uses Wikipedia to explain each topic. A code step formats everything into simple HTML, and Gmail sends the final digest to your inbox. It uses OpenAI for both GPT 3.5 and GPT 4 tasks to balance cost and quality.

Setup needs OpenAI and Gmail accounts plus basic n8n access. Expect to cut transcript review from hours to minutes, while improving topic coverage and email-ready formatting. Great for podcast recaps, editorial prep, internal learning notes, and weekly research briefs. Most of the work runs in one click, and chunking lets you handle very long transcripts without breaking the output.

• Manual start to process a new transcript on demand
• Transcript input via a code node for quick paste and control
• JSON document loader targets the transcript field for clean parsing
• Recursive text splitter with 6000 size and 1000 overlap to handle long content
• Refine style summarization across chunks using GPT 3.5
• Topic and question extraction with GPT 4 for better quality
• Structured output parser enforces a JSON schema for reliable fields
• Research agent uses Wikipedia to explain each topic in plain language
• HTML formatter builds a clean summary and topic list for email
• Gmail node sends the digest with subject and HTML body

• Reduce manual review from 2 hours to 10 minutes per episode
• Automate 80 percent of topic research and outline preparation
• Improve consistency of summaries with a fixed output schema
• Handle 5 times longer transcripts using chunking without failures
• Deliver email ready HTML digests directly to Gmail in one run
• Combine high quality and lower cost by mixing GPT 4 and GPT 3.5

1. Import the template into n8n: Create a new workflow in n8n > Click the three dots menu > Select 'Import from File' > Choose the downloaded JSON file.
2. You'll need accounts with OpenAI, Gmail and Wikipedia. See the Tools Required section above for links to create accounts with these services.
3. OpenAI credentials: In the n8n credentials manager, create a new OpenAI credential. Get your API key from the OpenAI dashboard and paste it into n8n. Save and name the credential clearly.
4. Gmail credentials: Double click the Gmail node. In the credential dropdown, click Create new credential. Follow the on screen steps to connect your Google account with OAuth2. Allow the email scope and save.
5. Wikipedia tool: No credential is needed for the built in Wikipedia tool. Ensure the Research and Explain Topics agent shows the Wikipedia tool connected.
6. Transcript input: Open the Podcast Episode Transcript code node. Paste your transcript into the transcript field string and save.
7. Chunking: Check the Recursive Character Text Splitter settings. Keep chunk size at 6000 and overlap at 1000 for long files, or adjust if your content is shorter.
8. Summarization: Confirm the Summarize Transcript node uses refine mode and is linked to the GPT3.5 Summarize language model. Set temperature to 0 for stable results.
9. Topics and questions: Ensure the Structured Output Parser is linked to Extract Topics and Questions and the JSON schema matches fields used by the Topics node.
10. Research agent: Verify the Research and Explain Topics node is connected to GPT3.5 Research and the Wikipedia tool. Run a test and confirm each topic includes an explanation.
11. Formatting: Open the Format topic text and title code node. Confirm it reads the summary from the Summarize Transcript output and builds HTML for topics and questions.
12. Email: In the Gmail Send Digest node, set the recipient email, subject, and make sure Email Type is HTML. Send a test to confirm layout and links.
13. Validation: Execute the workflow manually. Check each node output for summary text, topics array, and research notes. If results are empty, confirm the transcript pointer is /transcript and your OpenAI key is active.
14. Troubleshooting: If you hit token errors, lower chunk size or reduce topic count. If Gmail blocks sending, reauth the Gmail credential and check quota. If content looks off, lower temperature values.