Put a text or logo watermark on a whole folder of images at once. Size and margin are proportions of each image rather than fixed pixels, which is the difference between one setting that works across a mixed batch and having to re-tune it for every photo — a 40-pixel margin is a thick border on a thumbnail and invisible on a large photo. Nine positions plus a rotated tiled mode that is harder to crop out. Output keeps PNG transparency where the input had it. One honest note: a visible watermark discourages casual reuse, it does not prevent it — cropping or cloning it out is not difficult.
| Position | Best for |
|---|---|
| Bottom right | The convention for photography |
| Centre | Proofs you do not want used |
| Tiled | Hardest to crop out |
| Top left | Where the subject sits low |
No. Everything is drawn in your browser.
So one setting works across images of different resolutions. A fixed pixel size is enormous on a small image and invisible on a large one, which means re-adjusting for every file. Proportional sizing keeps the visual result consistent.
Not really. It deters casual reuse and makes the source obvious, but it can be cropped off, cloned out or removed by inpainting without much effort. Treat it as attribution rather than protection.
Tiled, considerably. A corner mark is a crop away. The tiled mode is rotated and covers the whole frame, so removing it means reconstructing image content underneath.
Yes for PNG input — it stays PNG. Other formats are written as JPEG, which has no transparency, so a transparent area becomes black.
Yes. Switch to logo mode and upload one. A PNG with a transparent background gives the best result, since a JPEG logo brings its own white box.
Watermark size and margin are proportions of each image rather than fixed pixel values, which is what makes a single setting usable across a batch. A 40-pixel margin is a wide border on a thumbnail and an invisible hairline on a 6000-pixel photo, so a tool built on fixed pixels has to be re-tuned for every image — proportional placement means one setting looks right on all of them. Text marks are drawn with the canvas text API at a size derived from the image's shorter edge, and measured with measureText before placement so right-aligned and centred positions account for the actual rendered width rather than an estimate. Tiling rotates the context thirty degrees and steps across a region larger than the canvas in both directions, because a tile grid aligned to the axes is trivially cropped out and one that starts at the origin leaves the top-left corner bare. Output format follows the input: a PNG stays a PNG so transparency survives, and anything else is written as JPEG at high quality. Worth saying plainly: a visible watermark deters casual reuse and does not prevent it. It can be cropped, cloned out or removed by inpainting, and none of that is difficult. Everything runs in your browser.