Clipora - AI Viral Clip Generator
An AI-powered viral clip generator that extracts the best short-form clips from long videos.
Role: Full Stack + ML Engineer. Team: Individual. Status: Complete. Year: 2025.
Overview
Clipora analyzes long-form video and extracts the best short-form clips for TikTok, Reels, and Shorts. It combines shot detection, transcription, and Gemini multimodal scoring to find the moments with the strongest viral potential.
Architecture:
- Mobile: React Native
- API: Node.js and Express on Cloud Run
- Video: Google Video Intelligence API for shot detection
- AI: Gemini 2.0 Flash for multimodal scoring
- Processing: FFmpeg for clip extraction
- Infrastructure: Cloud Run, Pub/Sub, and Terraform
Why This Architecture:
- Video Intelligence API handles shot detection at scale
- Gemini multimodal can evaluate visual + audio virality factors
- FFmpeg provides precise, efficient clip extraction
- Terraform ensures reproducible infrastructure
Key features
- Gemini multimodal virality scoring
- Video Intelligence API shot detection
- Automatic clip extraction with FFmpeg
- TikTok/Reels/Shorts format optimization
- React Native mobile app
- Terraform-managed GCP infrastructure
- Batch processing for long videos
Technical details
- Video: GCP Video Intelligence API for shot/scene detection
- AI: Vertex AI Gemini multimodal for content analysis
- Processing: FFmpeg for precise clip extraction
- Backend: Node.js and Express on Cloud Run
- Mobile: React Native for iOS/Android
- Infrastructure: Cloud Run, Pub/Sub, and Terraform
- Storage: Cloud Storage for video assets
Challenges
- Defining and scoring 'virality' factors objectively
- Processing long videos efficiently in cloud environment
- Extracting clips with precise timing and transitions
- Optimizing output for different platform requirements
Tags: Node.js, Express, Cloud Run, React Native, Gemini, FFmpeg