OCR Scanner App

The idea came from a simple, recurring frustration: needing to extract text from physical sources—documents, signs, menus—and do something useful with it, fast. The goal was to build an app that could run entirely on-device, recognize text in multiple languages with high accuracy, and integrate seamlessly into daily workflows.

Mobile OCR presents a unique optimization problem: accuracy versus speed versus battery life. The real challenge was model selection—balancing recognition accuracy across six languages with on-device performance on older iPhones.

SwiftUI formed the foundation of a native interface that prioritized camera access. Core ML and Vision Framework performed on-device text recognition, eliminating latency and privacy concerns.

The app achieved 98% accuracy on printed text across all supported languages, with processing speeds fast enough to feel instantaneous on modern devices.