YouTube content Audit with n8n and Gemini

Built with n8n
, , , , , ,
1

The Problem: 2-Hour Videos That Could Be a 2-Minute Read

It’s hard to distinguish high-quality content from someone just rambling for hours. You waste time watching, hoping to get to the point.

The Solution: Automated YouTube Intelligence Pipeline

A clean web app backed by an n8n workflow that audits YouTube links, saves your time, and protects your attention span.

4
42864×1676 396 KB

How It Works (The Blueprint)

  • Link Drop: User pastes a YT link on the website → n8n extracts the Video ID.

  • Smart Cache: Checks Supabase first.
    If watched before: Returns saved data instantly.
    If new video: Starts the analysis pipeline.

  • Transcription: Grabs the full text from transcriptapi.com and formats it into one long string.

  • Dual-Agent Audit (Parallel):
    Agent 1: Extracts key summaries and categories.
    Agent 2: Rates the video (1-10) on Sources, Logic, and Objectivity.

  • Save & Deliver: Saves results to Supabase and sends the clean data back to the user frontend.

1
12758×1596 366 KB

Things I don’t get quite right:

  1. Gemini is taking forever to reply (sometimes)

    The Issue: Long transcripts mean heavy payloads, making Gemini stall. The frontend user is left staring at a loading spinner for too long.

You can actually check the website: checktube.ue
It is personal project so it may interrupt but I will work on it.

1 Like
2

Welcome @MateuszCholewa!

The dual-agent parallel audit approach is smart - splitting summary extraction and credibility scoring into two concurrent Gemini calls cuts response time significantly compared to doing it sequentially. One thing worth exploring as your user base grows: if you hit Supabase rate limits on concurrent webhook calls, you can add an n8n queue (using the Wait node or a simple Data Table buffer) to smooth out burst traffic before hitting the DB layer.

1 Like
3

This is so helpful, thank you! I will play with it on the weekend for sure!