Screenshot Annotator, Explained

After reading this you will know how to crop, mark up and redact a screenshot so it says exactly one thing, and you will know why baked-in pixelation hides secrets that a light blur leaves readable.

What the tool does and one quick example

A screenshot annotator turns a raw capture into a picture with a point of view. You start with a rectangle of pixels: maybe 2560 by 1440 from a monitor grab. You crop out everything that does not matter, drop an arrow on the button you mean, blur the account number, and add a "1" badge so a reader knows where to look first. The output is a single flattened image in PNG, JPEG or on the clipboard.

Here is the hook. A support ticket with a screenshot of the whole desktop makes the reader hunt. The same screenshot cropped to a 600 by 300 region, with one red arrow and one blurred email address, is understood in about two seconds. The information did not change. The signal-to-noise ratio did.

Two ideas separate this tool from a paint program. First, annotations are objects, not paint, until you export. You can move the arrow, recolor the box, or delete the third badge without touching the pixels underneath. Second, redaction (blur and pixelation) is painted into the exported pixels, on purpose, so nobody can peel it off later.

When to use it, and when not

Reach for annotation when the goal is to direct attention or to hide something before you share. Bug reports, how-to steps, design feedback, and support replies all fit. A numbered sequence of three or four badges beats a paragraph of "click the gear, then the third tab".

Do not use it as a general image editor. If you need to fix exposure on a photo, use the Brightness & Contrast Editor. If you only want to hide regions and never draw an arrow, the focused Blur / Pixelate Redactor is simpler. If your job is stamping ownership across an image, the Watermark Tool tiles text or a logo for you.

Redaction is a security operation, not decoration. If you blur a password so lightly that a reader can still guess the characters, you have leaked it. Treat every blur as if an attacker will zoom in at 400% and run recovery filters.

The one number that decides safe redaction

Blur and pixelation both reduce detail, but they fail in different ways. Gaussian blur is a weighted average of nearby pixels. A common recovery attack sharpens or deconvolves the blur, because the original edges are still faintly encoded in the averaged values. Pixelation replaces a whole block with one flat color, which destroys the sub-block detail completely. Nothing inside a solid block can be recovered because nothing is stored.

The control that matters is the pixel block size relative to the size of the smallest thing you must hide. Call the stroke width of the text you are covering w in pixels, and the block size b. A rough safety rule:

b \ge 3w

Here b is the side length of one pixelation block in device pixels, and w is the width of the thinnest visible stroke in the content you are hiding. If a digit is drawn with strokes about 4 pixels wide, choose a block of at least 12 pixels. That guarantees no single block contains a clean piece of one stroke, so the shape of the character cannot survive.

Blur has an analogous parameter, the radius r, but it is weaker for the same visual coarseness because it preserves low-frequency structure. As a working equivalence, a blur radius r hides roughly what a pixel block of b \approx 1.5r hides, so you need a much larger radius to match the same block. Prefer coarse pixelation when the content is a real secret.

A worked example using the defaults

Marking up a login screenshot

Start from a 1280 by 720 capture of a settings page. The demo uses the field defaults: red stroke, medium line width, and default badge sizing. Suppose you want to show a reader where to change their email, while hiding the current address.

Crop to the region that matters. The form sits in a box from pixel (320, 180) to (920, 480). That crop is 600 wide by 300 tall, a 5.33:1 reduction in area from the original 1280 × 720 = 921600 pixels down to 600 × 300 = 180000 pixels, about 19.5% of the original.
Add a step badge "1" next to the email field label, and a badge "2" next to the Save button. The badges are objects, so you can nudge either one later.
Draw an arrow from badge "1" to the input, using the default red stroke.
Pixelate the email address. The address text has strokes about 3 pixels wide, so by b \ge 3w you set a block of at least 9 pixels. Round up to 12 for margin.
Export to PNG. The annotations flatten onto the pixels, and the pixelated block becomes permanent.

The finished image is 180000 pixels carrying one instruction. Compare that with the original 921600-pixel capture, where the same instruction was buried in menus, a sidebar and a clock.

Cropping to the form drops the pixel count to about 19.5% of the original, which is most of the noise removed before you draw a single arrow.

Reading the result: does the markup carry the message?

Judge an annotated screenshot by a simple test: if you covered the surrounding text, would the arrows and badges alone tell the story? Three signs that they do.

One primary color: Use a single accent, usually red, for the marks that lead the eye. When every arrow, box and badge shares a color, the reader treats them as one instruction set.
Numbered order where order matters: Badges "1", "2", "3" turn a static picture into a sequence. Without them a reader guesses the reading order and often guesses wrong.
Contrast against the background: A red box on a red error banner disappears. Check that each mark has clear luminance contrast with what sits under it, not just a different hue.

Line width scales with image size. On a 600-pixel-wide crop a 3-pixel line reads as a firm stroke, about 0.5% of the width. On a 2560-pixel capture that same 3-pixel line looks like a hairline, so scale up to roughly 10 to 12 pixels to keep the same visual weight.

Common mistakes

Crop first, annotate second. If you annotate on the full frame and then crop, your carefully placed arrow can land half outside the new edge. Cropping first also fixes the coordinate space so line widths and badge sizes look right the first time.

The mistakes that spoil most annotated screenshots are predictable:

Light blur on real secrets. A radius-2 blur on a 12-pixel-tall token is often reversible. Match the guidance above: pixelate with a block of three times the stroke width.
Too many marks. Five arrows compete and the reader sees none of them. If you need five points, you probably need five numbered steps instead.
Exporting JPEG for text-heavy shots. JPEG's block compression fuzzes crisp UI edges and colored text. For screenshots, PNG keeps every arrow and label sharp. Use JPEG only when file size forces it.
Forgetting the redaction is permanent. Once exported, the blurred region cannot be un-blurred, which is the point. Keep the original file if you might need the hidden content again.

Related tools

Pick the narrowest tool for the job. To hide regions and nothing else, the Blur / Pixelate Redactor is direct. For top and bottom caption text, use the Meme Generator. To fit a shot to a platform's frame, the Social Media Image Cropper has the exact sizes. Straighten a sideways capture with Rotate & Flip Image. Cut a subject out of its background with the Background Remover, and clean a grainy scan with the Image Noise Remover.

Frequently asked questions

Can I paste a screenshot straight from the clipboard?

Yes. Copy with your system screenshot shortcut, then press Ctrl+V in the tool. There is no need to save a file first.

Are my images uploaded anywhere?

No. Everything runs locally in your browser. The pixels never leave your machine, which matters when the screenshot contains private data.

Can I edit an annotation after I placed it?

Yes, until you export. Arrows, boxes, badges and text are objects: select any one to move, restyle or delete it. Blur and pixelation are the exception, because export bakes them into the pixels so they cannot be reversed.

Which format should I export?

Use PNG for anything with text, thin lines or flat UI colors, because it stays sharp. Choose JPEG only when you need a smaller file and can accept softened edges. The clipboard option skips saving entirely.

How big should the pixelation block be?

At least three times the width of the thinnest stroke you are hiding. For 4-pixel strokes use a 12-pixel block or larger. When in doubt, go coarser, since an over-blocked region is still readable as "redacted" while an under-blocked one leaks the secret.

Try it yourself in the tool