Skip to content
VANHONG

Web case study

TextSnap

High-performance client-side OCR web application.

TextSnap interface with image drop zone and extracted text panel

Project overview

Client-side multilingual OCR for English, Khmer, and Korean.

High-performance client-side OCR web application.

Problem or motivation

Scanning confidential physical documents via cloud OCR services exposes private text to external third-party infrastructure.

What I built

  • Client-side optical character recognition running entirely inside the user's browser using WebAssembly-powered Tesseract.js.
  • Multilingual script recognition supporting English, Khmer, and Korean with on-device canvas preprocessing.
  • Drag-and-drop processing pipeline with progress indicators, interactive text editing, and instant file export.

Technology stack

  • Next.js
  • TypeScript
  • Tailwind CSS
  • Tesseract.js
  • WebAssembly
  • 100% client-side execution confirmed; zero backend database or tracking telemetry.

Screenshots

TextSnap interface with image drop zone and extracted text panel

What I learned

  • Delegating heavy OCR matrix extraction to Web Workers to keep browser interactions buttery-smooth at 60 FPS.
  • Optimizing memory lifecycle and language dictionary caching within browser storage.