Features -Upload multiple images or a video -Video frame extraction and manual frame selection -Structured extraction (name, title, company, email, phone, website, address) -On-device OCR mode (no image uploads) -Batch processing with concurrency -Review + edit extracted contacts in a table -Dedupe toggle + basic merge strategy -Basic validation highlighting (email/phone/URL) -Export results to .xlsx, .csv, .vcf (vCard)

Tech Stack -React + TypeScript -Vite -Tailwind CSS (local build) -Tesseract.js (on-device OCR mode) -Google Gemini via @google/genai -xlsx for Excel export -Optional Node backend proxy (to keep API keys server-side)

Security / Privacy

Business cards contain PII (names, emails, phone numbers).

Data flow depends on the processing mode: -On-device OCR: images stay on the device (no uploads). -AI (Gemini): selected images are sent to Gemini (either directly from the browser if you use a client-side key, or via the optional backend proxy).

This app does not persist uploaded images or extracted results to a server you control by default: they stay in the browser (in-memory for the current session). The app does store settings (processing mode, OCR language, and optional API key) in localStorage on that device.

This is a client-side app. If you deploy it publicly with an API key bundled, the key can be extracted from the built assets.

If you plan to deploy this publicly, move Gemini calls to a backend (proxy) and keep the key server-side.

See PRIVACY.md for details and tradeoffs.

On-device OCR mode uses Tesseract.js and self-hosted worker/core/language assets (same origin, no third-party CDNs). First use may download ~60MB of OCR assets and cache them in the browser.