How Datablurai works
A privacy risk analyzer that scores your image before you share it — and offers one-click redaction. Every byte stays in your browser.
On-device pipeline
When you drop in an image we run a small chain of models and parsers entirely inside your browser: a WASM face detector, Tesseract.js for OCR, regex classifiers for email / phone / ID numbers, jsqr for QR codes, and exifr for GPS metadata. There is no server round-trip, no analytics on your image, no caching outside your tab.
What we detect
- Faces+15WASM face detector spots faces in group shots and selfies.
- License plates+15OCR + rectangle heuristics flag visible vehicle plates.
- Email addresses+20Regex over OCR text finds inboxes hiding in screenshots.
- Phone numbers+20International phone formats are matched on extracted text.
- QR codes+25Decoded with jsqr — even a small QR can leak a lot.
- ID numbers+50Country-specific patterns (e.g. TW national ID) carry the highest weight.
Risk scoring
Each finding adds its weight to a running total capped at 100. The total maps to a level that tells you, at a glance, whether the image is safe to share.
Built for any device
Heavy work runs in a Web Worker so your UI stays responsive. Models lazy-load only when you upload the first image, and Tesseract workers are terminated when they are no longer needed.