Hide text in an image

Names on packages, an address on an envelope, a document caught in the background of a photo — drag a box over each below and download a clean copy. Everything runs on your device; the original never leaves your browser.

The preset is a solid box rather than blur, and this page exists partly to explain why: pixelated or blurred text is one of the few redactions that has been publicly, repeatedly broken. Faces lose their identity when downsampled; text often doesn’t, because attackers know the alphabet and can brute-force what fits the smudge.

Preset: solid boxes for text. Faces in the photo are still auto-detected and get blur.

Drop a photo here — or click to browse, or paste with Ctrl+V

JPG · PNG · WebP — faces found and hidden on your device, never uploaded

How it works

  1. Open this page — the tool is already set up for “Hide text in an image”.
  2. Drop your photo in, click to browse, or paste it with Ctrl+V.
  3. Faces are detected on your device; drag boxes over anything else to hide.
  4. Check the preview (hold “compare”), then download the flattened, metadata-free copy.

Photographed text vs screenshot text

Recovery tools work best on screenshots — perfect fonts, uniform grids. A photographed label at an angle, in shadow, with camera noise, is a harder target, and a strong mosaic on it will usually hold. But “usually holds” is a poor standard when the fix is free: a solid box removes the question entirely, and on a document or label the black rectangle doesn’t hurt the photo the way it would on a face.

Handwriting sits in between: no fixed font helps recovery, but signatures are identifiers in their own right and worth covering fully rather than artistically blurring.

The usual suspects

Shipping labels in “look what arrived” photos. Mail on the kitchen counter in a listing shot. Prescription bottles. Whiteboards behind people in office photos — a classic source of leaked internal plans. Boarding passes, which encode your booking reference in both text and barcode; the barcode needs a box just as much as the letters do.

Tips that save a re-edit

Before you post: zoom in and check every face, plate and label yourself. Auto-detection is a head start, not a guarantee — a missed face in a shared photo can’t be un-shared.

Frequently asked questions

Can I use blur on text if I make it really strong?

At extreme strength the text becomes an unreadable smear and recovery gets unrealistic — but you’re guessing where “extreme enough” is, and researchers keep moving that line. A solid box isn’t a guess. This tool defaults to boxes for text and keeps blur for faces, which is the split the evidence supports.

The text is on a curve or at an angle — do boxes still work?

Yes — draw one box or several overlapping ones; boxes can tile any shape. The export flattens everything, so seams between overlapping boxes don’t matter.

Is the covered text still in the file somewhere?

No. The exported image is rebuilt from pixels with the boxes already applied — there’s no layer to peel off and no metadata carrying the original. That’s also why the download can’t be “undone”: keep your original if you’ll need the unredacted version.