Add a Watermark to Images

Yes, and that is not incidental — it is the entire reason this tool exists. The drawing goes through your browser's own text engine, which shapes the letters, joins them and reorders the two directions exactly as it does for any text on any page. The same feature on PDF files would have cost between 600 kB and a full megabyte, because the PDF library does no shaping and no reordering, which is why our "add text to PDF" tool is deferred. Here it costs nothing.

No, and we would rather say so. A visible watermark is a deterrent, not protection: it can be cropped out, painted over with content-aware fill, and sometimes removed in a single click. What it actually does is tie the image to your name everywhere it travels, and make unauthorised use obvious rather than quiet. If you need real proof of ownership, the high-resolution original on your own disk is the evidence — not the mark.

"32 px" means nothing on its own. The same number disappears entirely on a 6000 px photograph and covers a 400 px thumbnail edge to edge. Every measurement here — type size, margin, and the spacing between tiles — is a percentage of the longest edge, so the composition is identical at any resolution. It is also what makes the preview honest: the preview is a few hundred pixels wide and the exported file is thousands, and if the maths were in pixels the preview would resemble the result instead of predicting it.

Colour alone cannot solve this: white is the right choice on a dark photograph and invisible on a snow scene. The outline draws a thin edge around each letter in the opposite tone — black or white, chosen from the lightness of your colour — so the text stays readable on any background without you trying one colour after another.

Because they stop meaning anything. Tiling covers the whole frame, so there is no "bottom right" and no distance from an edge to compute. Hiding the two controls is more honest than leaving them on screen doing nothing.

Yes, and that is expected. Drawing text over an image means re-encoding the whole image, and the result depends on the format and the quality you chose — asking for PNG from a JPEG original will multiply it. The tool does what you asked and tells you where the size went rather than hiding the increase. If size matters, run the result through our Compress Image tool afterwards.

No. Everything happens inside your browser — the image never leaves your device, and there is no daily limit. And because the tool re-encodes the image, EXIF data and GPS coordinates are stripped along the way, which works in your favour if you are about to publish it.

Yes — this is exactly the case it was built for. Choose any number of images (up to 20, and up to 200 MB combined), set your text, position and size once, and we apply it with the same precision to every image — each one calculated from its own dimensions, so the composition stays identical regardless of size. The results come back as a single ZIP archive.

The preview represents only the first image in the batch — enough to confirm the text, position, size and colour all look right before running the whole batch, since these settings apply literally to every image. What the preview cannot predict is each image's own content, such as background contrast.

Each image is decoded on its own, so decode memory never grows with the batch — but every result accumulates in memory before the archive is built, and stamping can grow a file's size (the same reason a single file can grow, above). 20 images and 200 MB combined keeps the tool running smoothly on any device.

Yes. Switch the watermark type to "Logo (image)" and upload the logo file — it gets the exact same position, size, opacity and tilt controls text does, because the placement math never looked at what it was placing. Colour and the contrasting outline disappear in logo mode, since both are text-only concepts.

Any image format your browser can open — PNG, JPEG and WebP are guaranteed in every current browser. The logo goes through the same validation your target images do (a 50 MB limit, a real decode check), so an unsupported or corrupted file shows a clear error instead of freezing the tool or stamping a blank box.

Yes — transparency is fully respected. The canvas draw call composites the logo's own alpha channel automatically, so a transparent background stays transparent rather than turning black or white. The overall Opacity slider multiplies on top of the logo's own transparency, so a half-transparent logo at 50% opacity comes out fainter than either alone — which is expected, not a bug.

Yes — the one logo file you upload is stamped onto every image in the batch exactly, the same way one text string is. You upload the logo once, not once per image; each file still gets its own real position and size, calculated from its own dimensions.