How to Redact PII From Images
Omit redacts PII inside PNG, JPEG, TIFF and BMP files the same way it redacts documents: OCR reads the text, the same detector finds the PII, and the matching regions get blacked out in the image itself, entirely on your machine.
In short
OCR reports where each word sits on the page, so a redaction box lands on that word's own coordinates rather than blurring a region or guessing at one. The formats covered are PNG, JPEG, TIFF and BMP, and an image is a first-class input rather than a side effect of another format's path. A redacted copy does not carry the original EXIF data, so GPS coordinates, device model and timestamp do not survive the re-encode.
Omit redacts PII inside image files the same way it redacts documents. OCR reads the text sitting inside the image, the same detection engine that scans a spreadsheet or a PDF runs against that text, and the regions that match get blacked out directly in the image. Nothing is uploaded anywhere; the whole pass runs on your machine.
It works on PNG, JPEG, TIFF and BMP files, and it is a first-class input to Omit Redact, not a side effect of some other format: drop an image in the same way you would a document, and a redacted copy comes out.
How it works
How the boxes actually line up with the text. OCR does not just extract words, it also reports where on the page each word sits. When a detected span of PII falls inside a word OCR found, that word's own coordinates become the redaction box, so the black bar sits exactly over the text that was there, not a blur over the whole image or a guess at where the sensitive part might be.
What this covers
The kinds of images this covers in practice: a screenshot of a chat or an email thread with a name or phone number visible, a photo of an ID card or a filled-in form, a scanned page from a paper file, a screenshot of an internal dashboard or spreadsheet before it goes into a slide deck. Anywhere the PII exists as text sitting inside a picture rather than as a field in a document.
The same rule the rest of the product follows applies here without exception: the original image is never modified. Redaction writes a new file and leaves yours exactly as it was.
A side effect worth knowing
One side effect worth knowing about, not a feature we built on purpose: when Omit writes the redacted image back out, it does not carry the original EXIF data across the re-encode. A lot of photos carry more than the picture, GPS coordinates, device model, a timestamp, and that layer does not survive the redacted copy either.
Where this is weaker
Where this is weaker, honestly. OCR accuracy depends on the image. A low-resolution screenshot or a heavily compressed photo can lose words the detector never gets a chance to see, and handwriting is a harder problem for any OCR engine than printed text, ours included. If the text is legible to OCR, the same detection that finds PII in a document finds it here. If OCR can't read it, nothing downstream can act on it.
Everything above is measured the same way the rest of this product is measured: an offline pass on a real file, not a claim about a demo.