LocalSight — Local-first image OCR

Select and copy text from any image on the web, just like in WeChat, without uploading sensitive files

Problem

Turning text in images into editable content is a constant need, but most tools upload to unknown servers (contracts, screenshots, sensitive material) or cannot do direct in-browser selection like WeChat does. One developer built a pure front-end PDF merger and emphasized opening DevTools to see zero upload requests, showing a clear demand for local processing and verifiability.

Solution

A browser extension plus desktop app that starts local OCR on hover, letting users select and copy text directly from images. Enhanced recognition preserves structure for tables, formulas, and code, exporting to Markdown or Word. Everything runs locally with a one-click network-panel proof of zero uploads, plus a screenshot-recognition hotkey.

Users

Developers, students, legal and admin staff who frequently extract text from screenshots, scans, and course materials

MVP

Browser extension with hover-triggered local OCR and selection copy; screenshot hotkey recognition; enhanced structure-preserving recognition with Markdown/Word export; one-click zero-upload verification hint; desktop app for system-level screenshots

Revenue

Free basic local OCR; paid subscription or credits for enhanced format-preserving recognition; enterprise one-time license with offline deployment package

Why now

Local OCR models now run well on everyday devices, while sensitivity about data leaving the machine keeps growing, making a local, verifiable solution both feasible and trust-friendly