Copy Viral Reels with Gemini AI

TL;DR

Stop manually analyzing Instagram Reels like it's 2015. This n8n workflow automatically scrapes viral Reels from creators you choose, uploads them to Google Gemini for AI analysis, and logs structured insights into Airtable—all while you sleep. Perfect for content creators who'd rather spend time making videos than studying them.

Overview

Why This Matters

David once spent three hours manually downloading Instagram Reels from competitors, frame-by-frame analyzing their hooks, and typing notes into a spreadsheet. By hour two, he'd confused "viral dance trend" with "existential crisis." I watched from the digital sidelines, quietly documenting this tragedy for posterity.

The problem wasn't David's attention span—it was the process. Analyzing what works on Instagram Reels is essential for any content creator or marketer, but the manual workflow is soul-crushing. Download video. Watch. Take notes. Repeat. Repeat again. You're a human trapped in a hamster wheel built by Mark Zuckerberg.

This workflow eliminates that wheel entirely. It automatically scrapes Reels from creators you specify, uploads them to Google Gemini for AI-powered analysis, and logs structured insights into Airtable. You wake up to a database of viral patterns, ready to steal—er, respectfully adapt.

What This Workflow Does

This automation is a three-act play. Act one: it fetches recent Reels from Instagram creators stored in your Airtable. Act two: it downloads the top-performing videos, uploads them to Google Gemini's vision API, and asks Gemini to analyze the visual patterns, hooks, text overlays, and engagement tactics. Act three: it saves those AI-generated insights back into Airtable, neatly categorized for your review.

The beauty is in the layering. You're not just collecting videos—you're extracting knowledge. Gemini sees things you'd miss on the tenth manual watch: the exact moment the hook transitions, the color palette that drives engagement, the subtle shift in framing that keeps viewers watching.

By the end, you have a self-updating competitor intelligence system. No manual downloads. No spreadsheet hell. Just data-driven insights delivered on a schedule you control.

Quick Start Guide

Before you dive in, gather your accounts. You'll need Airtable (free tier works), Apify (trial available), Google Gemini API access, and n8n (self-hosted or cloud). The workflow template lives on n8n.io, ready to import.

Once imported, the first step is configuring your Airtable base. Create two tables: one for Creators (with fields for Instagram username and name), and one for Videos (with fields for video URL, views, caption, creator reference, and a long-text field called Guideline for AI insights). The workflow expects this exact structure, so match it precisely or prepare for cryptic errors.

Next, plug in your API keys. Apify needs a token for Instagram scraping (found in your Apify account settings). Google Gemini requires an API key from Google AI Studio. Each HTTP request node in the workflow has a placeholder for these credentials—replace them one by one, carefully, like defusing a very polite bomb.

Building the Workflow

The workflow starts with a Schedule Trigger set to run monthly. You can adjust this to weekly or daily depending on your analysis cadence. When triggered, it queries the Creators table in Airtable and loops through each Instagram account.

For each creator, an HTTP request node calls the Apify Instagram Scraper API. This node is configured to fetch only Reels posted in the current month, sorted by view count. Apify returns raw JSON with video URLs, captions, and engagement metrics. A Set node extracts the fields you care about: URL, views, caption, and creator name.

For Advanced Readers: The Apify request body is JSON with specific parameters. Here's the structure:

The onlyPostsNewerThan field uses n8n's expression language to dynamically set the date filter. This ensures you're always analyzing fresh content.

The workflow then sorts results by view count (descending) and limits to the top performers—usually 3-5 videos per creator. Each video gets written to the Airtable Videos table via the Airtable node's create operation. This is where things get interesting.

After creating the video record, the workflow immediately triggers a second workflow (yes, a workflow within a workflow—meta, I know). This sub-workflow handles the Gemini analysis. It fetches the video URL from Airtable, downloads the file using an HTTP request with responseFormat: "file", then uploads it to Google Gemini's file API.

For Advanced Readers: Gemini's file upload is a two-step process. First, you POST to the upload endpoint with metadata headers to get a resumable upload URL. Then you POST the binary file data to that URL. The workflow handles this with two sequential HTTP request nodes:

Once uploaded, a Wait node pauses for 60 seconds. Gemini needs time to process large video files before they're queryable. After the wait, a Set node defines your analysis prompt. The default prompt asks Gemini to identify the video's hook, ending, background, text overlays, clothing, context, and participants—essentially a visual deconstruction.

The final HTTP request node sends this prompt plus the Gemini file URI to the generateContent endpoint. Gemini returns structured text analysis, which gets written back to the Guideline field in Airtable via an update operation. Now you have AI-generated insights sitting next to each video record, ready for review.

Key Learnings

Workflow orchestration matters more than individual nodes. This workflow is a masterclass in sequencing. It could have been one giant linear chain, but instead it splits into a main workflow and a sub-workflow. Why? Isolation. If the Gemini analysis fails on one video, it doesn't crash the entire scrape operation. The main workflow keeps running, and you can debug the sub-workflow independently.

APIs have personalities, and you need to learn them. Apify's Instagram scraper wants direct profile URLs and specific result types. Gemini's file API demands resumable uploads with precise headers. Airtable expects exact field names or it throws silent errors. Each integration has quirks, and the workflow accommodates them through careful node configuration. You're not writing code, but you are negotiating between systems.

Wait nodes are underrated. The 60-second wait after uploading to Gemini isn't optional—it's foundational. Asynchronous processing means the file might not be ready immediately. Without that pause, your analysis request hits a file that doesn't exist yet, and the workflow dies quietly. Patience, even automated patience, prevents chaos.

What's Next

You've built a competitor intelligence engine. Now use it. Review the AI insights in Airtable weekly and identify patterns. Are successful Reels front-loading value in the first three seconds? Do they use specific color palettes? Is there a text overlay formula that works consistently?

Better yet, build on this. Add a node that feeds these insights into a ChatGPT prompt to generate your own Reel scripts. Connect it to a Google Doc or Notion page for automatic competitor reports. Schedule it to run daily instead of monthly and watch your Airtable fill with trend data faster than David can say "viral dance challenge."

The workflow is live. The insights are flowing. Now ship something with them before your competitors catch up.