Most watermarks fail in one of two ways. Some are so small and faint that a repost crops them off without anyone noticing. Others are so large and dark that the photo is ruined for the people it was meant for. Getting it right is less about software than about three decisions: where the mark goes, how big it is relative to the picture, and how visible it is.
This guide walks through those decisions with real numbers, explains the difference between a corner mark and a tiled one, covers file formats and quality, and is honest about what a visible watermark can and cannot protect. The examples use the image watermarker on Toolz, which composites your logo onto a photo in the browser.
TL;DR: A visible watermark is a logo, signature or handle composited onto an image so credit travels with it. Use a transparent PNG for the mark. For attribution, put a small mark in a corner at 50 to 80% opacity; to make removal harder, tile it across the frame at 10 to 25%. Size the mark against the image's shorter side so it looks consistent on portraits and panoramas. A watermark deters casual reuse but does not prevent removal. The Toolz watermarker does corner, edge, center and tiled placement with size, opacity, margin and rotation, keeps the original resolution, and runs client-side, free, no signup.
What is an image watermark?
An image watermark is a visible mark placed over a photograph to identify its owner or source. It is usually a logo, a signature, a website address or a social handle, drawn semi-transparently so the picture remains visible underneath.
The point of a visible mark is that it stays attached to the pixels. File metadata such as EXIF author fields is stripped by most social platforms on upload and lost entirely in a screenshot. A mark drawn into the image survives both, and it survives resizing and recompression as well. That makes it the most durable form of attribution for images that will be shared.
There is a second kind of watermark that is invisible: data hidden in the pixels, the file or the text that only software can detect. AI systems increasingly add these to what they generate, and they answer a different question, about where content came from rather than who owns it. For the text version of that idea, hidden Unicode characters in AI output, see the guide to invisible watermarks in AI text. This guide is about the visible kind you add to an image yourself.
Should you use a corner watermark or a tiled one?
Use a corner mark when the goal is credit, and a tiled mark when the goal is to make the image costly to steal. The two serve different purposes and trade off in opposite directions.
| Factor | Corner (single) mark | Tiled mark |
|---|---|---|
| Main purpose | Attribution | Deterrence |
| Typical opacity | 50% to 80% | 10% to 25% |
| Typical size | 10% to 20% of the short side | 5% to 15% of the short side |
| Visual intrusion | Low | High |
| Removal by cropping | Often possible | Not possible |
| Removal by cloning or inpainting | Easy, one small area | Slow, many areas |
| Good for | Portfolios, social posts, blog images | Client proofs, stock previews, product shots at risk of scraping |
A corner mark is polite. It sits out of the way, credits the photographer, and does not compete with the subject. Its weakness is that a corner is exactly what gets cropped when an image is reposted in a different aspect ratio.
A tiled mark repeats across the whole frame. It cannot be cropped away, only painted out, and doing that across twenty or more instances takes real effort. The cost is that everyone who views the image sees it through the pattern, which is why it suits previews and proofs rather than finished work.
Where should a single watermark go?
Put it where it does not cover the subject and where a crop is least likely to remove it. The watermarker offers nine anchor positions: the four corners, the middle of each edge, and the center.
| Position | Crop risk | Intrusion | Common use |
|---|---|---|---|
| Bottom right | High | Low | Default for most photographers |
| Bottom left | High | Low | Images with the subject on the right |
| Bottom center | Medium | Medium | Wide landscape shots |
| Center right or center left | Medium | Medium | Portraits with space beside the subject |
| Center | Low | High | Proofs where credit matters more than looks |
| Top positions | High | Low | Images with busy lower halves |
Two practical rules help. Look at where the subject is before choosing, and avoid placing the mark over a face, a product label, or text in the photo. And check the tone of the area behind the mark: a white logo over a bright sky disappears, while the same logo over foliage reads clearly.
The tool opens with the mark at bottom right, which is the conventional choice.
How big should a watermark be?
Big enough to read at the size people will see the image, and no bigger. A useful starting range is 10 to 20% of the image's shorter side for a corner mark. The watermarker's Size slider runs from 2% to 60% and starts at 18%.
The important detail is what the percentage is measured against. Many tools size the mark as a percentage of image width. That makes the same setting look tiny on a tall portrait and oversized on a wide panorama. The Toolz watermarker measures size and margin against the shorter side of the image, so one setting behaves consistently across orientations.
Here is how that works with real numbers. Take a logo file that is 1000 × 400 pixels (an aspect ratio of 2.5 to 1), with Size at 18% and Margin at 4%, anchored bottom right.
Landscape photo, 6000 × 4000 pixels:
- Shorter side: 4000 px
- Mark width: 4000 × 0.18 = 720 px
- Mark height: 720 × (400 ÷ 1000) = 288 px
- Margin: 4000 × 0.04 = 160 px
- Left edge of mark: 6000 − 720 − 160 = 5120 px
- Top edge of mark: 4000 − 288 − 160 = 3552 px
Portrait photo, 4000 × 6000 pixels:
- Shorter side: 4000 px
- Mark width: 720 px, height 288 px, margin 160 px
- Left edge: 4000 − 720 − 160 = 3120 px
- Top edge: 6000 − 288 − 160 = 5552 px
The mark is the same physical size on both, and it sits the same distance from the corner. Had the size been 18% of the width instead, the landscape mark would be 1080 px wide and the portrait mark 720 px, which looks inconsistent side by side in a gallery.
What opacity should a watermark be?
For a single corner mark, 50 to 80%. For a tiled mark, 10 to 25%. The Opacity slider in the tool runs from 5% to 100% and starts at 70%.
The reason tiled marks need far lower opacity is accumulation. One mark at 60% is readable and unobtrusive. Twenty marks at 60% turn the photo into a pattern with a picture behind it. At 15%, each instance is faint, but together they cover the frame enough that removal is tedious.
Judge opacity on the darkest and the lightest image you plan to mark, not only the one in front of you. A white mark at 50% can vanish on a snow scene and glare on a night shot. If your images vary widely, keep two versions of the logo, one light and one dark, and pick per image.
| Situation | Suggested opacity |
|---|---|
| Portfolio image, corner signature | 50% to 70% |
| Social post with a handle | 60% to 80% |
| Client proof, centered logo | 30% to 50% |
| Tiled mark on stock preview | 10% to 20% |
| Tiled mark on product photos | 15% to 25% |
How does tiling work, and how dense should it be?
When "Tile across the whole image" is checked, the mark is repeated in a grid, and the position and margin controls are replaced by a Spacing slider. Spacing runs from 0% to 200% in 5% steps and starts at 60%. Like size, it is measured against the image's shorter side: it is the gap added between marks, horizontally and vertically.
A worked example on the same 6000 × 4000 photo, with Size at 10% and Spacing at 20%:
- Shorter side: 4000 px
- Mark width: 400 px; with the 1000 × 400 logo, height 160 px
- Gap: 4000 × 0.20 = 800 px
- Horizontal step between marks: 400 + 800 = 1200 px
- Vertical step between marks: 160 + 800 = 960 px
That gives about five columns and four rows across the frame, roughly 20 marks with some cut by the edges. Lower the spacing for a denser pattern, or raise it if the image needs to stay readable.
Rotation matters most for tiling. A diagonal pattern, around 30 to 45 degrees, is harder to remove than a straight grid because it does not line up with the edges of objects in the photo. The Rotation slider runs from −90 to 90 degrees. When tiling is rotated, the tool draws the grid around the center of the image and well past its edges, so the corners are covered instead of being left bare, a common flaw in simpler tools.
For a single mark, rotation turns the logo around its own center. That is occasionally useful for a signature that should follow the angle of a horizon.
What file format should the watermark and the output be?
The watermark itself should be a PNG with a transparent background. Transparency is stored in PNG's alpha channel, defined in the W3C PNG specification, and it lets the logo sit directly on the photo. A JPEG has no alpha channel, so a JPEG logo brings a solid rectangle, usually white, with it. If your logo exists only as an SVG, export it to PNG at several times the size you plan to use so the edges stay crisp when drawn large.
The tool accepts any image your browser can decode for both inputs: PNG, JPEG, WebP, GIF and AVIF in all current browsers. HEIC photos from iPhones do not decode in most browsers and need converting first.
The output format depends on where the image is going. The watermarker exports PNG, JPEG or WebP, and the file is named after the original with -watermarked added.
| Output | Compression | File size | Keeps transparency | Best for |
|---|---|---|---|---|
| PNG | Lossless | Largest | Yes | Archiving, further editing, graphics |
| JPEG | Lossy, 92% quality | Small | No | Photos for the web and social media |
| WebP | Lossy, 92% quality | Smallest | Yes | Websites that serve WebP |
Browsers encode the file with the canvas toBlob method described in the WHATWG HTML standard. JPEG and WebP are written at a quality of 0.92. If your source photo was already a JPEG, saving the watermarked copy as JPEG adds a second round of lossy compression. At 92% that is rarely visible, but choose PNG when the result will be edited again. For choosing between formats more generally, the image format converter guide compares them in detail.
Does watermarking reduce image quality or change the photo?
The pixels outside the mark are not resized, cropped or shifted. The tool creates a canvas at exactly the source's width and height, draws the photo at full size, and composites the mark on top. The preview shows the dimensions, and the downloaded file matches them.
Two things do change:
- Compression, as described above, if you export as JPEG or WebP.
- Metadata. Drawing an image onto a canvas and exporting it produces a new file with no EXIF data. The camera model, capture date and any GPS location from the original are not carried over. That is often a benefit when publishing, since location data in a shared photo is a privacy risk. If you need the metadata preserved, keep the original file. The EXIF remover guide explains what that data contains.
One edge case: if the photo you are watermarking has transparent areas and you export as JPEG, those areas cannot stay transparent and will usually be filled with black. Export as PNG or WebP to keep them.
Can a watermark stop someone stealing your images?
No. A visible watermark deters and attributes; it does not prevent. Anyone determined can crop a corner mark, clone over it in an image editor, or run an inpainting model that fills the area convincingly in seconds. Tiling slows this down but does not stop it.
What a watermark does is raise the effort and make intent obvious. Casual reposting, which accounts for most unattributed reuse, tends to leave marks in place. And removing a mark is more than a technical act. In the United States, a watermark containing the owner's name or other identifying information can count as copyright management information, and removing or altering it knowingly to conceal infringement is prohibited under 17 U.S.C. § 1202. Other countries have comparable rules. That turns a removed mark into evidence in a takedown or claim.
For images with real commercial value, treat a watermark as one layer alongside other measures:
- Publish only resized copies, never the full-resolution original.
- Register important work where registration strengthens legal remedies.
- Use reverse image search periodically to find reuse.
- Send takedown notices when you find copies.
- License clearly, so honest users know how to use the image properly.
What are common watermarking mistakes?
| Mistake | Result | Fix |
|---|---|---|
| Using a JPEG logo | White box around the mark | Use a transparent PNG |
| Sizing by width | Marks look inconsistent across orientations | Size against the short side |
| Same opacity for tiled and single marks | Tiled images look ruined | Drop tiled opacity to 10% to 25% |
| Mark over the subject's face | Photo unusable, viewers annoyed | Choose a different anchor |
| White mark on a bright image | Mark invisible | Use a dark version of the logo |
| Watermarking the only copy | Clean original lost | Keep the unmarked file |
| Re-saving JPEG repeatedly | Visible artifacts build up | Export PNG for intermediate copies |
| Tiny low-resolution logo scaled up | Blurry, jagged mark | Export the logo larger |
How do you add a watermark with the Toolz watermarker?
The whole process takes under a minute:
- Add the photo. Drop it on "Image to watermark", click to browse, or paste a screenshot from the clipboard. Paste goes to this first box only.
- Add the watermark. Drop or browse for your logo in the Watermark box. A transparent PNG is recommended.
- Choose placement. Click one of the nine position buttons, or check "Tile across the whole image".
- Adjust. Set Size, Opacity, Margin (or Spacing when tiling) and Rotation. The preview updates as you drag.
- Pick a format and download. Choose PNG, JPEG or WebP, then click Download. The preview is exactly what you get.
The watermarker works on one image at a time, on purpose. A fixed position rarely suits a mixed batch: a bottom-right mark that works on a landscape may land on someone's hand in a portrait. Checking each image avoids that.
Everything runs client-side. Both images are decoded and composited on an HTML canvas in your browser; nothing is uploaded or stored, and the page keeps working offline once loaded. The tool is free, no signup, with a daily download allowance like other tools on the site.
Once an image is marked, the other tools in the social image tools hub handle what usually comes next: the social media image cropper for platform sizes, the matte generator for fitting a photo onto a square or story canvas without cutting it, the photo card generator, and the carousel scroll generator for splitting a panorama into swipeable slides. Crop before you watermark, so the mark is positioned against the final frame.
When you are ready, add your logo to a photo and start from the defaults: bottom right, 18% size, 70% opacity.
