Image Watermark

A watermark has two possible jobs, and they want opposite arrangements. Signing your work wants one small mark in a corner where it does not spoil the picture. Deterring reuse wants a mark that cannot be removed without removing the image. Most watermark tools offer the first, call it protection, and never mention the difference.

How to use it

1. Drop an image onto the panel, press Choose images, or paste a screenshot. Pick several files to do them all at once.
2. Type the text you want.
3. Choose Tiled across the image to deter reuse, or a corner for a signature.
4. Adjust opacity until it reads without dominating, then press Download.

Tiling is the only arrangement that works

A single mark occupies a small, known region of the frame. Whatever else is true of it, it can be cropped away, and the remaining picture is still perfectly usable. That is not a flaw in the implementation; it is the geometry of putting one thing in one place.

A tiled mark repeats across the whole frame at an angle. To remove it by cropping you would have to crop until nothing is left. Removing it by inpainting is possible but tedious across dozens of overlapping instances, and the angle makes it worse — a horizontal mark can sometimes be patched by copying a clean row, and a rotated one cannot.

The angle also helps legibility. Text at −30 degrees crosses both light and dark regions of most photographs, so no single instance disappears entirely into a bright sky or a dark shadow.

Everything is relative to the image

Font size, margin and tile spacing are all percentages rather than pixel values. This is the difference between a tool you can point at a folder and a tool you have to re-tune for every file.

Four per cent of the shorter edge is a comfortable tiled mark. On a 1000×800 photo that resolves to 32 pixels; on a 4000×3000 photo it resolves to 120. Both look the same on screen, which is the point. The interface shows the resolved pixel size for whatever file you have open, so the abstraction never becomes mysterious.

Margins work the same way, which matters for corner placements: a fixed 20-pixel inset is a generous margin on a small image and hairline-tight on a large one.

Opacity, shadow, and staying readable

A watermark that cannot be read is decoration, and one that dominates the image ruins it. The working range for a tiled mark is roughly 20 to 25 per cent — visible when you look for it, ignorable when you look at the photograph.

The drop shadow is on by default because opacity alone is not enough over mixed content. A white mark at 22 per cent is clear over a dark subject and effectively gone over a bright sky. A shadow gives it an edge in both cases, which is why professional watermarks almost always have one.

Batch, and what the tool does not claim

Select several files and the batch button processes each one in turn and hands back a ZIP. Processing is sequential rather than parallel on purpose: decoding and encoding are both memory-hungry, and forty large photos in flight at once is how a browser tab gets killed.

Nothing is uploaded, and nothing is added but your own text. A watermark records a claim; it does not establish one. If you are watermarking work to assert ownership, keep your originals and their metadata, because those are what actually evidence it.