Image Watermark in the browser is not a generate
/free-tools/image-watermark never spends a credit. Do not pay a model to do an in-browser file job.
/free-tools/image-watermark never spends a credit. Do not pay a model to do an in-browser file job.
The watermark tool overlays text on an image with position, size, and opacity, then exports a PNG. The file is read into memory on your machine, drawn on a canvas, and offered back as a download. Nothing is sent to a server. There is no account. There is no limit. Generation on Versely is the opposite of that: every model call costs credits, and there is no free generation tier. Mixing the two is how a proof mark becomes a 6-credit still you cannot undo.
Canvas is not a model
The free-tools surface has one rule: nothing there calls a model, an API, or the network. Image Watermark is Canvas and the File API. The work happens on the device. That is why it stays free, and why it keeps working with the network off.
Images up to 40 MB. Canvas holds the decoded bitmap, so a 40 megapixel photo needs roughly 160 MB of memory regardless of how small the JPEG is. That is a browser ceiling, not a credit ceiling.
A generate is a catalog row: Ideogram, Flux, Recraft, a video model asked to "put © in the corner." That row is billed. It will also restyle the picture, restype the mark, and give you a cousin file. Watermarking is not a style. It is a stamp on this bitmap.
If you needed a new image, open a stills row and pay the quote you see before confirm. If you needed a mark on an image you already have, stay in the browser tool.
What the stamp actually is
Type the line. Pick a corner (or centre): top left, top centre, top right, centre, bottom left, bottom centre, bottom right. Size is 12–240 px. Opacity is 0.1–1. Colour is a picker. Optional dark halo for contrast. The font is the browser's generic sans-serif — no webfont is loaded, because that would be a network request. On a 400 px-wide image a 48 px font is already large and will show canvas rasterisation, not a hinted desktop font.
The text is rasterised onto the bitmap and exported as PNG. It cannot be toggled off later. That is the point of a watermark and also why you should keep an unmarked original if you will ever need one. The download name is {original}-watermark.png.
This is not a layer in a PSD. This is not C2PA. This is not the platform watermark a social app burns on a cross-post. It is pixels you chose, burned in, locally.
Credits are for generates
The cost hub is the map of how jobs bill: per second, per 1,000 characters, per megapixel, per call, per export. Every generation shows its credit cost upfront. Unused credits roll over while a subscription is active. None of that arithmetic applies to Image Watermark, because no job is dispatched.
The expensive confusion is sending a pack shot through a text-to-image or image-edit model with the instruction "add a watermark." You will pay the model's meter. You may get a mark. You will also get a restyled product. Then you will still want the unmarked master, which you no longer have if you overwrote it.
Do the stamp last, on the file you are willing to mark, in the free tool. Do the generate first, on purpose, on a row you chose.
FAQ
Is Image Watermark actually free?
Yes. No account, no credit, no limit, because it costs nothing to run — the work happens on your device. Generation on Versely is not free. This tool is not generation.
Do my files get uploaded?
No. The file is read into memory on your own machine, processed there, and offered back as a download. Nothing is sent to a server.
Can I remove the watermark later?
Not from the PNG the tool exports. The text is part of the pixels. Keep an unmarked original.
Why not generate the watermark with Ideogram so the type looks designed?
Because that is a 6-credit stills job that rebuilds the picture. If the picture already exists, stamp it in the browser. If you need designed type in a poster you do not have yet, that is a generate — and it is a different page from this one.