How to Watermark an Image: Add a Text or Logo Watermark That Survives Reposting

A watermark is the cheapest insurance a photo can carry. It overlays a piece of text or a logo at a position and opacity you choose, and it tells anyone who finds the image — on a marketplace listing, a screenshot, a re-upload — where it came from and who made it. That single overlay does most of the work of discouraging casual theft, and it costs you about fifteen seconds per photo once your settings are dialed in.

This guide walks through the decisions that actually matter: text versus logo, where on the frame the mark sits, how transparent to make it, when to tile it diagonally, and how to stamp a whole batch without uploading a single file. If you want to follow along, the Image Watermark tool runs every step in your browser tab.

Text mark or logo mark

The first fork is whether your watermark is type or a graphic.

A text watermark is the fastest path. Type © Studio Mei 2026 or @yourhandle, pick a font, a size, and a color, and you are done. Text is sharp at any resolution, it reads even when the photo is shrunk to a thumbnail, and it carries a literal credit line a stranger can follow back to you.

A logo watermark is a small image — almost always a tightly cropped PNG with a transparent background — composited on top of the photo. It reinforces brand recognition in a way text can't, which is why product and portfolio shots usually wear a logo rather than a name. One caution that catches people constantly: use a transparent PNG, not a JPG. A JPG logo drags its white rectangular background onto your photo, so instead of a clean mark you get an ugly box. Export the logo with alpha first, and only the shape shows.

Where the mark sits: the 9-position grid

Watermark placement is a nine-anchor grid borrowed straight from photographers: the four corners, the four edge midpoints, and dead center. Each anchor pins the mark to that part of the frame with a margin from the edge — a 24-pixel gap by default — so the watermark never collides with the photo's border.

Two anchors do almost all the work:

Bottom-right is the industry default for a copyright stamp. It is unobtrusive, it sits where the eye finishes scanning, and it reads as "this is signed" rather than "this is defaced."
Center is the choice when you genuinely don't want the mark removed. A corner can be cropped off in two seconds; a centered mark sits over the subject, so cropping it out means cropping out the thing the thief wanted.

The trade-off is honest: the more crop-resistant the position, the more the mark intrudes on the image. Corners stay polite; center protects.

Opacity: visible but not loud

Opacity is the dial that separates a professional mark from a ruined photo. A watermark stamped at 90% opacity buries the subject and screams amateur. The sweet spot depends on the mode:

For a single corner mark, around 60% keeps it legible without fighting the photo.
For a tiled diagonal mark, drop to roughly 18–20% — low enough to read the image through it, high enough that the pattern is unmistakable.

A useful instinct: if the watermark is the first thing you notice when the page loads, it is too strong. Pull it back until it reads as a quiet signature rather than a stamp of ownership slapped across the work.

Tiling: the only mode that resists cropping

Here is the concrete point that changes how you think about protection. A single-anchor watermark only survives if nobody crops it off — and cropping is the first thing a reposter does. Tile mode repeats the watermark diagonally across the entire frame at a fixed spacing, the same way "Confidential" runs across an internal PDF. Because the pattern covers everything, any crop large enough to be worth stealing still contains at least one complete watermark. A semi-transparent diagonal tile is, in practice, the hardest watermark to crop out cleanly.

Pair tiling with low opacity and you get the best of both: the photo stays viewable up close, the mark fades into the background at normal viewing size, and a screenshot-and-repost still carries your name. No client-side watermark is invincible — a determined attacker with content-aware fill can attack a corner mark — but a faint diagonal tile turns "two-second crop" into "thirty minutes of retouching," and most casual reposters simply move on.

Batch watermarking without uploads

Watermarking one photo is trivial; the value shows up at scale. Drop in up to ten images, set your recipe once, and hit "Apply to all" to run the same position, opacity, and mark across every file. Each photo previews live, and when you're happy you export a single PNG or a zip of the whole batch. Settings persist between batches, so stamping two hundred photos is just twenty rounds of the same frozen recipe — a few minutes of clicking, no re-configuring.

A worked example: a portfolio shot with a corner logo

I keep a personal portfolio and got tired of finding my images on scraper sites with no credit, so here is the exact recipe I use. I drop the photo in, switch the watermark type to Image, and upload my transparent PNG logo. I set the logo size to about 12% of the photo's shorter edge — big enough to read, small enough to ignore — anchor it bottom-right, and pull opacity down to 65%. The preview shows a discreet mark tucked into the lower corner, exactly where a viewer's eye lands last. For a gallery of thirty images I do it in three batches of ten, click "Apply to all" each time, and export a zip. The mark survives a right-click save and an Instagram re-upload, which is all I need: it points a curious viewer back to me, and it makes lazy reposting pointless.

A note on privacy

The reason all of this can run without an account is that none of it touches a server. Every image — the original photo, the logo PNG, and the composited result — is read with FileReader and drawn onto an off-screen <canvas> inside your tab, then exported with canvas.toBlob(). You can open DevTools, watch the Network tab, and confirm the request count stays at zero while you watermark. This matters more than it sounds: the things people watermark are exactly the things that shouldn't leave their machine — unreleased product shots, photography proofs, internal screenshots, ID scans. URL sharing carries only the recipe (position, opacity, text), never the pixels, so an unpublished image never lands in a link or on a host you don't control.

If your watermarked exports come out heavier than you'd like, run them through a local image compressor afterward — it works the same in-browser way, so your photos still never leave the tab.

The short version

Use text for a credit line, a transparent PNG for a brand mark. Anchor it bottom-right when you just want a signature, center or tile it when you actually need protection. Keep single marks near 60% opacity and tiled marks near 18%. And when crop-resistance genuinely matters, reach for tile mode — a faint diagonal repeat is the one pattern a quick crop can't defeat. Do the whole thing locally, and your originals never leave your hands.

Made by Toolora · Updated 2026-06-13