SnapText — On-Device Real-Time OCR App
Krish Ujeniya
An on-device optical character recognition mobile application built to extract text from documents and camera feeds with zero cloud roundtrips.
Core Problem
Most OCR solutions require remote cloud APIs that create latency, incur recurring usage costs, and transmit sensitive client documents over external networks.
System Design & Implementation
On-Device Inference: Embedded Google ML Kit vision models for on-chip text detection and extraction.
High-Performance Frame Pipeline: Streamlined live camera stream processing with background Dart isolates to maintain a smooth 60 FPS UI.
Architecture: Structured using clean repository patterns to support batch scanning, text formatting, and offline export.
Repository
Like this project
Posted Oct 30, 2026
Likes
0
Views
0
