Redact a scanned document or photo
A scan is just a picture of a page, so many tools can't find anything to redact in it. Sumi reads the text in your scan first, on your own device, then finds the personal data and lets you black it out. Photos taken with a phone work too.
What works
- scanned PDFs with no text layer;
- photos in PNG, JPEG or WebP, which Sumi turns into a one-page PDF;
- tilted pages, straightened before reading;
- faded paper and uneven backgrounds, cleaned up before reading;
- 30 languages, detected automatically for Latin-script text (pick the language yourself for others, such as Arabic or Chinese);
- mixed PDFs: turn on "Also read text inside images on every page" to catch text in logos and stamps.
How to do it with Sumi
- Drop the scan or photo into the app. Text recognition starts on its own and shows its progress page by page.
- Review the findings as usual: names, IBANs, ID numbers, addresses, dates of birth.
- Words the reader wasn't sure about are flagged, so check those against the page.
- Use Draw box over anything the reader missed.
- Preview, then Redact & download.
How to check the result in 10 seconds
Zoom in on each black box in the downloaded file: no letters show at the edges. Photos also lose their hidden metadata, including GPS location.
Limits
Very blurry or low-light photos can be misread. Retake the photo in daylight, flat on a table, if many words are flagged as unclear.
Questions
Is the scan sent to a server for text recognition?
No. The recognition engine runs in your browser and works offline after your first visit.
How long does it take?
About a second per page on a recent laptop, longer on phones and older computers.
Does it read handwriting?
It doesn't read it, but it can box it for you. See redacting handwriting and signatures.