Screenshot Auto-Redactor

What auto-redaction finds for you

Drop in a screenshot and the page reads it with on-device OCR (tesseract.js), then pattern detectors flag the things people leak most: API keys (OpenAI, AWS, GitHub, Slack, Google, JWTs), email addresses, phone numbers, Aadhaar-style 12-digit numbers (checksum-verified), PAN-style IDs, card-style numbers (Luhn-checked), and IP addresses. Each find becomes a checkbox in the review list with a red preview box drawn over the screenshot — so you see exactly what will be covered.

Why you review every redaction (mandatory by design)

OCR is a suggestion engine, not a guarantee. Small text, stylized fonts, and low-contrast UI can all slip past it — and a redactor that silently "handled it" would be dangerous. So the review step is mandatory: nothing is covered until you check its box and press the button. The page says it plainly: suggestions, not a guarantee — look at the screenshot yourself before sharing.

Solid blackout vs blur — which is safer?

Solid black boxes, always. Blur and pixelation average the pixels underneath, and researchers have recovered text from both in practice — a "blacked out" marker that's even slightly transparent can leak under brightness adjustments too. A solid opaque box destroys the information; nothing of the original survives inside it. That's the only cover this tool applies. For passwords, keys, IDs and financial details, blackout isn't the cautious choice — it's the only real one.

How to use this redactor

1. Drop in your screenshot. Click the box or drag the file in — PNG, JPG, WebP, anything. OCR runs in your browser; nothing uploads.
2. Wait for the scan. The progress bar shows OCR reading the text, then the detector pass.
3. Review every suggestion. Each flagged item is a checkbox with a red preview box. Uncheck anything that isn't actually sensitive — and eyeball the whole screenshot for things OCR missed.
4. Black out the selected areas. One click paints solid black boxes over everything you checked.
5. Zoom in and verify. Check every black box fully covers its text, then download. Keep your original separately.

Real-world use cases

• Dev screenshots for Stack Overflow / GitHub issues — terminal output and .env files love to leak keys into bug reports.
• Support tickets with dashboards — customer emails and order IDs hide in admin panels.
• Sharing error dialogs with IT — usernames and machine names ride along in screenshots.
• Social posts of funny chats — phone numbers and profile details in message screenshots.
• UPI / banking receipts to landlords — account numbers and balances in payment screenshots.

Tips

• Rotate a leaked key; don't just redact it. Redaction doesn't un-expose a key that already went out — revoke and regenerate first.
• Crop before you redact. Anything outside the frame can't leak; tight crops shrink the review job.
• Check the small text twice. Timestamps, tooltips and status-bar text are OCR's blind spots.
• Dark mode is fine. OCR reads the text regardless of theme.
• Keep the original. Downloaded blackouts are permanent — keep the unredacted screenshot if you might need it.

The bottom line

OCR finds what your eyes skip; you confirm what gets covered; solid black boxes make it permanent. Auto-suggest plus mandatory human review — that's the whole design.