Scans · Images
Redact a scanned document, right in your browser
A scan is just a picture. A computer cannot see the names in it until the text is read. ONYRI Sanitize reads your scan with OCR that runs on your own device, finds the sensitive data, and paints solid black bars over it. The picture never leaves your browser.
Runs 100% in your browser. Your file is never uploaded.
What ONYRI finds in a scan or a photo
- Names printed on forms, letters and ID documents
- Home addresses and postcodes
- Phone numbers and email addresses
- ID card, passport and driving licence numbers
- Social security numbers
- IBANs on bank statements and bank detail slips
- Card numbers on receipts
- Dates of birth and issue dates
How to redact a scan in three steps
- 1
Drop the image
Use a PNG, JPG or WebP file: a scanner export or a phone photo. The OCR engine loads in your browser and reads the text in English, French and German. This takes a few seconds.
- 2
Check what was read
ONYRI shows the bars on top of your image. OCR can misread a blurry word, so look at each zone. Add missing words under “Also mask” and untick what should stay visible.
- 3
Download the redacted image
A single image comes back as a PNG with the bars burned into the pixels. They are part of the picture. There is no layer to lift and no original hidden behind.
Scans · Images
Why on-device OCR matters for scans
Scans are often the most sensitive files you own: ID cards, payslips, signed forms, medical letters. Most OCR services work in the cloud. To read the text, they must first receive the whole picture.
ONYRI ships the OCR engine to your browser instead. The model files are served from our own domain, and the reading happens on your machine. The image, the text that was read and the redaction all stay in the tab.
The result is a true redaction. The black bars are painted into the pixels of a new image. It is like a very good black marker, except that nobody can hold the page up to the light.
What you should know before you share the image
- OCR quality depends on the scan. Blurry, skewed or low-contrast images are read less well. A clean scan at 300 dpi gives the best results.
- Handwriting is not read. Crop handwritten parts out, or cover them, before you drop the file.
- Photos, faces, stamps and signatures are images, not text. ONYRI does not detect them.
- Detection is not exhaustive. Review the preview before download. Redaction reduces exposure; it does not by itself guarantee GDPR compliance.
Redacting scans: frequent questions
Which image formats can I use?
PNG, JPG and WebP work best. A scanned PDF works too: pages with no text layer are read by the same OCR, page by page.
Does the OCR send my image to a cloud service?
No. The OCR engine runs inside your browser. Its files are loaded from our own domain, then all reading is done locally on your device.
Which languages does the OCR read?
English, French and German. Text in other Latin-script languages is often read correctly, but accents and special characters may be less reliable.
Can the redaction be reversed?
No. In black-marker mode the bars replace the pixels. The exported image does not contain the original values. Keep your original scan in a safe place if you need it later.
Why does a scan take longer than a PDF?
A text PDF carries its text. A scan or a scanned PDF does not, so the OCR must first recognise every word. Expect a few seconds per page, depending on your device.
Your document, made confidential in seconds.
Drop a file, check the preview, download the masked copy. The file never leaves your device.