Turn fresh news into ready to publish audio in minutes. This build pulls articles from a news homepage, chooses the best stories, writes a clean script, and produces a voice ready file. It helps content teams ship daily news podcasts without manual copy and paste work.

Here is how it runs. A manual trigger starts a fetch of the news page, then HTML nodes extract article cards, titles, links, and short descriptions. Items split out and a limit node caps volume. A Gemini based classifier scores each headline to decide if it fits a podcast. Only suitable items move on. The workflow fetches each full article page, extracts the body, removes empty results, and aggregates the text. A Gemini LLM then turns the set of stories into a structured podcast script using an output parser. If a script exists, a Hugging Face text to speech node generates the audio file.

You need access to Google Gemini and Hugging Face and the right credentials in n8n. Expect production time to drop from hours to minutes and more consistent story tone. Great for news roundups, brand updates, or internal briefings. Adjust CSS selectors if your source site layout changes and tune the classifier to match your editorial style.

• Manual start to control when new episodes are produced
• HTTP fetch of the news homepage to collect fresh article cards
• HTML parsing to extract titles, links, and short descriptions
• Split and limit controls to manage volume and keep runs fast
• Gemini powered text classifier to mark stories as suitable for audio
• Detail fetch and HTML extraction to pull full article text
• Filter and aggregate steps to remove empty items and combine content
• LLM chain with a structured output parser to produce a clean podcast_script field
• Conditional check to confirm script exists before audio generation
• Hugging Face text to speech to output a ready to use audio file

• Reduce manual production from 2 hours to 10 minutes per episode
• Automate up to 90 percent of script drafting and formatting
• Handle up to 10 times more stories without adding staff
• Improve consistency of tone and structure across episodes
• Connect editorial selection, writing, and voice output in one flow

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 Google Gemini and Hugging Face. See the Tools Required section above for links to create accounts with these services.
3. Open the Gemini node: double click the node, then on the 'Credential to connect with' dropdown, click 'Create new credential' and follow the on screen instructions to connect your Google Gemini account. Name the credential clearly, for example Gemini Prod.
4. Open the Hugging Face Text to Speech node: create a new credential using your Hugging Face access token from your account settings. Store it as HuggingFace TTS Prod.
5. In the news fetch HTTP Request node, confirm the URL points to the homepage you want to monitor. If your site is different, paste the correct homepage URL.
6. Open the HTML extract nodes and check the CSS selectors for article cards, titles, links, and descriptions. If your source site layout differs, update selectors to match the page elements.
7. In the Limit node, set the maximum number of articles to process per run. Start with 10 to keep tests fast.
8. Open the News Classifier node and adjust the category descriptions to reflect your editorial rules. For example, prefer clear, timely, and brand safe topics.
9. Check the LLM chain Output Parser and confirm the schema includes a podcast_script field. Keep the output simple so the text to speech step reads cleanly.
10. Run the workflow with the manual trigger. Inspect the News Classifier output to see which items are marked Suitable, then confirm the Fetch Detail and Extract Detail nodes return article text.
11. If extraction returns empty arrays, update CSS selectors or add alternative selectors. If TTS fails, verify the Hugging Face token and selected model supports text to speech.
12. Optional: add a Schedule trigger in n8n to run daily at a set time once manual tests are stable.