🎬 Extract viral-worthy clips from YouTube videos with Gemini AI & FFmpeg editing

⚡ 3,239 views · 🎬 Content Creation & Video

💡 Pro Tip — Since TikTok doesn’t have a public API, getting data usually means fragile HTTP scraping. ScraperNode is a community node that gives you clean JSON for TikTok profiles, videos, and comments — drop it into your workflow and skip the parsing.

View All Scrapers

Description

Workflow Overview

Who’s it for

This workflow transforms hours of manual video editing into an automated AI-powered pipeline. Perfect for anyone looking to repurpose long-form content into viral short-form clips.

Ideal users include:

Whether you’re a solo creator or managing content at scale, this workflow saves 5-10 hours per video while maintaining professional quality output.

How it works

This workflow combines AI analysis with professional video editing tools to automatically identify and produce viral-ready clips from any YouTube video.

The process flows through three main stages:

Stage 1: Download and Analysis

Stage 2: Clip Selection and Extraction

Stage 3: Professional Editing Pipeline

Final Delivery: The workflow processes clips with configurable wait times to match your system’s capabilities. When all clips complete processing, you receive an email notification and find your social-ready clips in the /data/clips/ directory, ready for upload to any platform.

Requirements

⚠️ Self-hosted n8n only - This workflow requires command-line access and cannot run on n8n Cloud due to its dependency on system-level tools.

System dependencies you must install:

Credentials needed:

Storage requirements:

How to set up

Step 1: Install system dependencies

SSH into your n8n host and install required tools. For Ubuntu/Debian systems, run:

apt-get update
apt-get install ffmpeg
pip install yt-dlp

Verify installations by running ffmpeg -version and yt-dlp --version.

Step 2: Configure directory structure

Create the clips output directory with proper permissions:

mkdir -p /data/clips
chmod 755 /data/clips

Step 3: Import the workflow

Download the workflow JSON and import it into your n8n instance. You’ll see several sticky notes color-coded by stage: yellow for description, blue for download/analysis, pink for editing operations, and green for clipping.

Step 4: Set up credentials

Navigate to the “viral clips identification” node and add your Google Gemini API credentials. The workflow uses the gemini-2.5-flash model for optimal speed and quality balance. Then configure Gmail OAuth2 in the “Send a message” node following n8n’s authentication wizard.

Step 5: Update email notification

Open the “Send a message” node and replace habeebmohammedfaiz@gmail.com with your email address.

Step 6: Create the editing subworkflow

The workflow references a separate subworkflow for the editing pipeline. Create a new workflow in n8n, copy all nodes from the “EDITING” section (between the Execute Workflow Trigger and the final output), and save it. Note the workflow ID from the URL.

Step 7: Link the subworkflow

In the main workflow, open the “Call subworkflow” node and update the workflow ID to match your newly created editing workflow.

Step 8: Test with a short video

Start with a 5-10 minute YouTube video for your first test. Use the manual trigger or form submission. Monitor the execution to ensure all nodes complete successfully and clips appear in /data/clips/.

Step 9: Adjust performance settings

Based on your system’s performance during the test, modify the Wait node durations. Systems with 8GB+ RAM and modern CPUs can reduce wait times to 30 seconds. Limited systems should keep 60-second waits or increase them.

How to customize the workflow

Adjust clip quantity and quality thresholds

Open the “filter out top clips according to score” node. The code currently uses .slice(0, 10) to select the top 10 clips. Change this number based on your needs: use .slice(0, 5) for only the best clips, or .slice(0, 20) for more options. You can also add score filtering by adding results.filter(c => c.score > 0.7) before the slice operation to only include clips with virality scores above 70%.

Customize subtitle appearance

Navigate to the “calculate relative subtitle size” node. The JavaScript code defines several styling variables you can modify:

Modify AI analysis prompts

In the “viral clips identification” node, edit the Gemini prompt to target specific content types. For educational content, add “Focus on key teaching moments and actionable tips.” For entertainment, emphasize “Identify funny moments, reactions, and unexpected events.” For podcast clips, specify “Extract controversial opinions, storytelling segments, and quotable statements.”

Change aspect ratios

The workflow defaults to 9:16 for vertical video. To create horizontal clips for YouTube or other platforms, open the “Analyze the actual whole video” node and change the aspect ratio in the JSON schema from "aspect_ratio": "9:16" to "aspect_ratio": "16:9". The AI will automatically adjust cropping coordinates accordingly.

Enable audio normalization

By default, audio normalization is disabled for faster processing. To enable it, open the “extract all actionable operations” node, find the audio_normalize task object, and change enabled: false to enabled: true. This ensures consistent volume levels across all clips but adds processing time.

Add custom editing operations

The editing pipeline is modular. You can add new operations like:

Add these as new task objects in the “extract all actionable operations” node following the existing pattern.

Customize notification content

Open the “Send a message” node to modify the email subject, body text, or add clip details. You can include clip metadata like timestamps, scores, and descriptions using expressions like {{ $json.hook }} or {{ $json.score }}.

Integrate with cloud storage

Add nodes after clip generation to automatically upload finished clips to Google Drive, Dropbox, AWS S3, or any n8n-supported storage service. Use the Loop Over Items1 output to access completed clip file paths.

Schedule automated processing

Replace the Form Trigger with a Schedule Trigger to automatically process videos from a spreadsheet or RSS feed. Combine with Google Sheets integration to maintain a queue of videos to process overnight.

đź”— Nodes Used

Execute Sub-workflow, Gmail, Execute Workflow Trigger, Filter, n8n Form Trigger, Read/Write Files from Disk

đź“Ą Import

Download workflow.json and import into n8n: Workflow menu → Import from File

📖 Importing guide · 🔑 Credential setup