WatermarkPictureStart
A snowy peak over a still lake with “© Alpine Light” set in the bottom right corner
A corner mark inside the safe area, sized as a share of the frame width.

How easily a watermark is cropped out

Measured, not guessed: how much of a photo survives a crop that removes the watermark, for a corner mark, a centred mark and a repeated lattice — read off this site's own crop check.

The quickest way a mark comes off

Most watermarks that disappear are not painted out. They are cropped: someone trims the edge of the picture where the mark sits and keeps the rest. So the most useful single number about a placement is how much of the picture a crop can keep while losing the mark entirely.

This site measures exactly that. Its crop check draws the mark — at full strength, by the same renderer that writes your file — onto a small copy of the frame, marks every pixel the mark touches, and finds the largest crop with the picture’s own proportions that touches none of them. The figures below are what that check reported, not estimates.

The test

One photograph, 2,400 × 1,600 pixels (3:2), a sky-and-trees landscape drawn by a script for this test. The mark is the studio’s default wording, “© Your name”, 24 % of the picture’s width with a 4 % inset; the lattices use the three density presets on the tiled mode. Each figure was read from the crop check on the placement mode on 1 November 2026.

Largest crop at the picture's own proportions that removes the mark
Bottom right corner82 % of the picture kept
Bottom centre82 % kept
Centre24 % kept
Repeated, Light (gaps 2.4 × 3.2)no clean crop larger than 5 %
Repeated, Standard (1.6 × 2.4)no clean crop larger than 1 %
Repeated, Dense (1.1 × 1.6)no clean crop larger than 1 %

What the numbers say

An edge mark leaves most of the photograph behind it: losing 18 % of the frame is a small price for a clean copy. A centred mark is far better — a crop that avoids it keeps under a quarter of the picture — but it sits on the subject, which is the price. A lattice is the only placement where cropping stops being worth doing at all, and even the lightest preset gets there. Beyond that, extra density buys nothing against cropping and costs the honest viewer more of the picture.

The numbers change with your photo’s shape, your mark’s size and its position, so check your own: the crop check is open on the placement and tiled modes, and draws the clean crop on the stage so you can see exactly what survives.

  • One test photograph and one mark size. A larger mark, a narrower inset or a portrait frame gives different figures — that is why the check runs on your own picture.
  • The check measures crops only, at a resolution of about half a per cent of the frame. It says nothing about retouching or AI fill.
  • A crop keeping 82 % of a photo is still a usable photo. A mark at the edge is a credit, not a protection.
Is a corner watermark pointless, then?

No — it is a credit, not a lock. On the test photo below a corner mark could be cropped away while keeping 82 % of the picture, which is a perfectly usable photo. If the job is to say whose picture it is when it is shared as it is, a corner mark does that and costs the picture nothing. If the job is to stop the picture being used without you, it does not.

Why does a bottom-centre mark measure the same as a corner mark?

Because a crop that keeps the picture's own proportions only has to lose the strip the mark sits in. A mark along the bottom edge — in the corner or in the middle — is removed by trimming that strip, and the rest of the frame survives either way.

How dense does a lattice need to be?

On the test photo, the Light preset already left no clean crop larger than 5 % of the picture, and Standard and Dense left about 1 %. Past that point a denser lattice does not make cropping any less useful; it only covers more of the picture you wanted people to see.

Does this measure how hard a mark is to paint out?

No. It measures cropping only — the quickest and commonest way a mark is removed. Painting out or filling a mark with retouching tools is a different question, and no measurement here pretends to answer it.