What a mark is for
A visible mark is friction. It does not lock a picture and it was never able to, but it changes what somebody has to do before they can use yours as though it were theirs. Four things it genuinely does:
- It stops casual reuse. Lifting a product shot for a marketplace listing now means either living with your name across it or doing real work to get rid of it, and most people lifting a product shot are not going to do real work.
- It carries attribution when a picture travels. A repost, a screenshot, a scraped feed — the file loses every piece of metadata on the way, and wording drawn into the pixels is the only thing that survives the trip.
- It sets an expectation. A marked picture reads as claimed. An unmarked one reads as stock, whatever the licence actually says.
- On a document scan it makes a leak traceable. A stamp naming who received the copy and when is why the practice exists at all: if that scan turns up somewhere, the stamp says whose copy it was.
All four describe a mark that is meant to be ignored — repeated, semi-transparent, there to be worked around rather than looked at. A mark you want people to actually read is a different job — one element, placed once, sized to be seen — and it belongs in a tool built around placing one element exactly. This site repeats things quietly.
Your files stay on your disk. This page reads them, draws the mark, writes new files, and keeps nothing — there is no server behind it for a copy to go to.
Setting one up
- Open a picture. Drag it in, paste it, or pick it from disk. Several at once opens the queue instead of the single-file editor.
- Write the mark, open a logo file, or make a QR code. Wording takes a face, a weight, a colour, an outline and a shadow. A logo takes a PNG with a see-through background; a flat JPG gets a warning rather than a silent rectangle. A QR code is encoded on the page from the link you type.
- Size it as a percentage. Width is set against the picture, not in pixels, so one setting lands identically on a 900-pixel web copy and a 6000-pixel original.
- Put it somewhere. Nine pins sit on the picture inside a dashed safe area: click one, press 1 to 9 in keypad order, or drag the mark and it snaps as it passes a pin. X and Y take exact percentages, and the corner and stem handles size and turn it.
- Repeat it, or leave it once. Turn repeating on and the mark becomes a lattice with its own gaps, angle and row offset, and a live count of how many instances the frame holds.
- Save it. The export re-draws the mark against the full picture rather than enlarging what is on screen, so the file matches what you approved.
The keys are worth learning if you do this more than once. The mark itself is a focus stop on the picture: tab to it, arrow it around a pixel at a time, hold shift for ten, press Enter to put it back on the grid point it came from.
| 1 – 9 | Jump to one of the nine grid points, laid out the way a numeric keypad is: 7 is the top left, 1 is the bottom left. |
|---|---|
| Arrows | Nudge the mark one pixel while it has focus. Shift makes it ten. |
| [ and ] | Step the opacity down and up by five. |
| T | Turn repeating on and off. |
| ⌘Z / ⌘⇧Z | Step back and forward through every change, including loading a saved setting. |
| ⌘↵ | Write the file. |
Files this takes
| Pictures in | JPG, PNG, WebP, GIF, BMP, ICO, AVIF where the browser decodes it, TIFF through a decoder fetched the first time a TIFF turns up, and HEIC from an iPhone the same way. |
|---|---|
| Mark files | PNG, SVG, WebP, GIF, BMP and JPG. A PNG with a see-through background or an SVG is what you want; a flat JPG is flagged before you save anything, and its background can be knocked out. |
| Fonts | Fourteen faces ship with the site — the two it is set in and twelve open-licence faces chosen for marks, including scripts, a stencil and a Devanagari sans and serif — plus whatever this device has. Your own .woff2, .woff, .ttf or .otf is read into this tab and registered for as long as it stays open. |
| Files out | JPG, PNG or WebP, or whatever the original was. PNG and WebP keep see-through areas, so a cut-out marked here is still a cut-out. |
| Sizes | Up to 120 MB a file, and about 32 megapixels before the browser is offered a lighter working copy instead — 16 on an iPhone or iPad, where the ceiling is far lower and failure is silent. |
| Queue | Up to five hundred files a run, worked on a few at a time so the tab holds only what it is using. |
The three ways out are not interchangeable once a mark is on the picture. JPG throws away detail every time it is written, and a thin outline or a pale mark at low opacity is exactly the kind of detail it throws away first — at quality 70 a 4 % mark starts to fray at its edges. PNG keeps every pixel and every see-through area and costs three to six times the bytes of the JPG it replaced, which matters when four hundred of them go into one archive. WebP sits between the two and is the only one of the three that leaves here without a credit field written into it. Marking a cut-out is the one case with no choice in it: JPG has no transparency, so a cut-out saved as one comes back with a solid rectangle behind it.
Three file types behave in a way worth saying out loud rather than discovering afterwards. An animation is marked on frame one and written out as a still. A CMYK JPEG — which is how a lot of print work arrives — is converted to screen colour to be drawn on, so the saved colours can sit slightly differently from the original. A TIFF with several pages gives up page one and the rest are not carried through. Each of those appears as a line under the picture the moment the file is opened.
What a watermark does not do
This is the part most tools in this category leave out, and it is the reason to trust the rest of the page. Five things a visible mark will not do for you:
- It does not survive a crop. A mark tucked into a corner is gone the moment someone drags a selection box around the middle of the picture, and no setting here changes that.
- It does not stop anyone determined. Taking a mark off a picture is its own job with its own tools, and they exist. A corner mark comes off in seconds. A lattice across the subject takes real effort and leaves real damage, which is the strongest thing you can honestly claim.
- Tiling is not free. The lattice that is hard to lift out is also the lattice that is in the way of the customer, the editor or the client you wanted to see the picture. There is no setting that is both invisible and stubborn.
- It does not create a right. Copyright starts when the picture does. Wording drawn on top of it is a claim about that right, not the right itself, and it does not register anything anywhere.
- It is not a forensic mark. There is no invisible payload here and no detector, because a detector needs a service running somewhere to check against, and this site has no server behind it by design. What you get is what you can see.
Why this site will never take a mark off
It is worth stating plainly, because the search results for anything to do with watermarks are full of the opposite offer. This site will not remove a watermark. Not as a page, not as a guide, not as an answer in this list, and not as a demonstration with a working control on it.
The first reason is that it would be an argument with itself. Every honest thing above — that a lattice is real friction, that coverage is a trade worth making — is advice about how much protection a mark buys. A site that sells you that on one page and defeats it on another is worth listening to on neither, and the person who came here to protect their work cannot tell which page it is being run for. The second is that the two visitors want opposite outcomes from the same file: there is no arrangement of this interface that serves both, so serving one means choosing, and the choice was made when the domain was registered. Removal tools exist and they are not hard to find; they are simply not going to be built here.
What replaces it is the honesty already on this page: a mark is friction rather than a lock, and you choose your coverage knowing exactly what it buys.
The subject check, and what it costs
One control on this site downloads something: “Suggest a spot” fetches a 5.9 MB subject finder — 6,163,938 bytes, Apache-2.0 licensed — the first time it is pressed, and then scores the nine grid positions against where a person is standing in the frame. On the folder route the same check runs over a sample of the queue and reports how many of the samples disagree with its answer. Your picture is never part of that request; the model travels to you. Press nothing and nothing is downloaded, and every other control on the page works exactly as it does now. It is a convenience worth 5.9 MB to some people and worth nothing to others, which is why it is one line here rather than the first claim in the title.
The working position, which the guides argue at length: mark across the subject rather than in a corner if being copied is what actually worries you, keep the opacity high enough to read and low enough to live with, keep your originals somewhere you can produce them, and treat the mark as friction rather than as a lock.
Will the mark come off?
A mark in a corner comes off in seconds — content-aware fill is in every phone's photo app now, and a crop gets rid of it without any tool at all. A mark repeated across the subject is much harder to take out cleanly, and it is also in the way of the people you wanted to look at the picture. That trade is the decision, and it is yours rather than ours.
Does putting my name on a picture give me the copyright?
No. The copyright exists from the moment you take the picture, in almost every country, whether or not anything is written on it. A mark is a claim written on top of that right, not the right itself. In an argument it is your originals, your raw files and your registrations that carry weight; the mark makes the claim visible, which is a different and smaller job.
What opacity should I use?
It starts at 35, which is readable on most photographs without dominating them. Below about 15 the mark stops being legible once the picture is resized for a feed, and above about 60 it is the first thing anyone sees. Move it with the slider or step it five at a time with the square-bracket keys, and judge it on the picture rather than on the number.
Why is the size a percentage instead of pixels?
Because a folder is rarely one size. Forty pixels of wording is a banner across a 600-pixel thumbnail and an illegible smudge on a 6000-pixel scan, so a pixel setting is wrong on at least one of them. A percentage of each picture's own width lands the same everywhere, which is what makes a saved setting worth saving.
Can I stamp a scanned document with a diagonal line of text?
Yes — turn repeating on, set the lattice angle to about −30°, drop the opacity to around 20 and write whatever the copy is for. That is exactly the diagonal stamp people put across a scan before sending it on. What this does not do is cover up the numbers on it: a translucent mark leaves every digit readable, and hiding part of a page is a different job that this site does not do.
Do saved settings follow me to another computer?
No. They are kept in this browser on this device, because there are no accounts here and nothing of yours is stored anywhere else. Clearing site data for this page takes them with it, and a second machine starts empty.
What happens to an animated GIF or WebP?
The mark goes on the first frame and the file is written out as a still picture. Every later frame is dropped. The tool says so next to the picture as soon as it sees an animation, rather than after you have saved three hundred of them.