← Back to Portfolio
Project 06 · AI · Data Engineering

AI Video Recipe Extraction

Automated recipe extraction and database management system that analyzes cooking videos using Google's Gemini AI and builds a structured, searchable recipe database.

Overview

AI Video Recipe Extraction is an intelligent system that automatically analyzes cooking videos and converts them into structured recipe data. Using Google's Gemini AI model, the system extracts ingredients, quantities, cooking steps, and preparation methods directly from video content.

The pipeline includes fuzzy string matching to deduplicate ingredients across multiple recipe sources, ensuring a clean and normalized recipe database. All extracted data is stored in a searchable SQLite database via SQLAlchemy ORM.

This project demonstrates the practical application of multimodal AI — combining video understanding with structured data extraction to solve a real-world content processing challenge.

Architecture
Cooking Video Input ↓ Google Gemini AI — Video Analysis ↓ Structured Recipe Data Extraction ↓ Fuzzy String Matching (Ingredient Dedup) ↓ Data Normalization & Validation ↓ SQLAlchemy ORM → SQLite Database ↓ Searchable Recipe Collection
Key Features
Tech Stack
Frontend
Web Dashboard HTML5 / CSS3 JavaScript
Backend
Python FastAPI / Flask SQLAlchemy ORM
Database & Storage
SQLite Recipe Store
AI & Multimodal Engines
Google Gemini AI RapidFuzz Data Deduplication