Watermarking that works: placement, opacity, and what it cannot do

A watermark is a compromise between two things you want simultaneously: an image that looks good, and an image that stays attributable when it travels. Most watermarks fail at one or the other, usually by being either invisible or unbearable.

Be honest about the goal first

Watermarks are asked to do three different jobs, and the right design differs for each.

Attribution. You want your name to travel with the image when it is shared, reposted or screenshotted. Subtlety is fine — the mark only needs to be legible when someone looks for it.

Deterrence. You want casual unauthorised use to be inconvenient. This needs a mark that is genuinely hard to remove without damaging the image, which means it cannot be tucked in a corner.

Proofing. You are showing clients work they have not paid for yet. Here the mark should be obtrusive, because the point is that the proof is unusable as a final image.

Confusing proofing with attribution is the most common mistake. A discreet corner signature does not stop a client from using the proof, and a large diagonal overlay across your portfolio makes your portfolio unpleasant to look at.

Relative sizing is not a detail

A watermark specified in absolute pixels looks correct on exactly one image size. Twenty-four point text is prominent on an 800-pixel proof and effectively invisible on a 6000-pixel master, and a batch containing both comes out looking careless.

Sizing the mark as a percentage of each image's width fixes this. Five percent of width is a reasonable starting point for text; a logo at 15–20% of width reads clearly without dominating. Margins should be relative too, so the mark sits the same visual distance from the edge regardless of resolution.

This matters more than it sounds, because mixed-resolution batches are the norm — a shoot delivered at full resolution, proofs at 1200 pixels, social crops at 1080.

Placement

Corners are the default for attribution because they are the least intrusive, and the bottom-right is conventional enough that viewers know where to look.

Corners are also the easiest thing in the world to crop off. If deterrence is the goal, the mark has to be somewhere that cropping it damages the image — across the subject, or repeated across the frame.

Some practical placement notes:

• Avoid areas of high detail, where a semi-transparent mark becomes illegible.
• Avoid pure white or pure black regions, where a single-colour mark disappears entirely.
• Watch for the safe areas of the platform you are posting to. Instagram's interface overlays the bottom of a story; a mark placed there is covered by a caption.
• If your images vary wildly in composition, a consistent corner is safer than a "smart" placement that lands somewhere different on every frame.

Opacity and contrast

Full-opacity white text on a photograph looks like a mistake. Somewhere between 60% and 85% is the range where a mark reads clearly while still sitting inside the image rather than on top of it.

Contrast is the harder problem: a white mark vanishes on a bright sky and a black one vanishes in shadow. Three approaches work in practice — a mark with its own subtle drop shadow or outline, placement chosen for mid-tone areas, or accepting that one colour will not work for every image in a varied set and processing light and dark images separately.

What a watermark does not do

It is worth being direct about this, because watermarking is sometimes sold as protection that it does not provide.

A visible watermark can be cropped, cloned out, or removed by inpainting tools that have become very good very quickly. It does not prevent copying, it does not prevent scraping, and it does not establish ownership in any legal sense that the file's existence did not already establish.

What it does is raise the effort of casual reuse and ensure that when an image does travel, it carries your name. Those are real benefits. They are not the same as protection.

If provenance genuinely matters, the belt-and-braces approach is a visible mark for attribution plus records of your own — original files with their capture metadata, kept somewhere you control, which is what actually helps if a dispute arises.

Batch consistency

Whatever you decide, apply it identically across a set. A portfolio where the mark moves position, changes size, and varies in opacity between images reads as unprofessional far more than any single choice of placement would.

This is the argument for doing watermarking as a batch operation with fixed settings rather than by hand per image: not speed, but the fact that a human doing it two hundred times will not do it the same way two hundred times.

Fonts, and a limitation worth knowing

If your watermark is your brand's wordmark, use an image file rather than text. Text watermarks in browser-based tools are typically limited to generic system font families, because the worker threads doing the processing have no access to page-loaded webfonts. A PNG or WebP of your logo with transparency sidesteps the issue entirely and guarantees the mark matches everything else you produce.