How the blur is applied
Drop a photo and it is drawn onto a canvas at its native resolution. Click and drag across any part of the picture to mark a rectangle; when you release, that region is redrawn with a Gaussian blur and the rest of the image stays sharp. You can add as many rectangles as you need, undo the last one, or clear them all. The Blur strength slider sets the blur radius in pixels, from 5 to 60 with a default of 22, and moving it re-renders every region at once so you can judge the effect live.
The radius is in image pixels, not screen pixels. On a 1200 pixel wide screenshot a radius of 22 turns a name into an unreadable smear; on a 6000 pixel wide camera photo the same setting looks mild because a face may be 800 pixels across. For large photos push the slider up, or mark the region and check the preview at full zoom.
Auto-detect faces, and why it is optional
The Auto-detect faces button uses the browser's built-in FaceDetector API. Where it exists, mostly Chrome on Android and some Chromium builds on desktop, it returns a bounding box for each face, and the tool enlarges each box by 20 percent so ears, hairline and chin are covered rather than only the eyes and nose. Firefox and Safari do not ship the API, and in that case the page tells you so and you mark the faces by hand. Detection can also miss faces that are turned away, partly hidden or very small, so always look at the result before sharing. Manual rectangles work everywhere and remain the reliable method.
Is a blur really irreversible?
A strong Gaussian blur destroys the high-frequency information that carries letters and facial features; there is no filter that brings it back. What you should avoid is a weak blur on text: at a radius of 5 pixels, large printed characters can sometimes still be guessed, and research has shown that lightly blurred digits are recoverable by statistical methods. For licence plates, document numbers and account details, use a radius of at least 25 on a normal screenshot and more on a large photo, or make the rectangle generous and stack two overlapping rectangles, which compounds the blur. The exported file contains only the blurred pixels; the original file on your device is untouched, so keep or delete it as you see fit.
Output format and what is stripped
The download is always PNG, named after the original with -blurred added. PNG was chosen deliberately: it is lossless, so a screenshot with small UI text stays crisp outside the blurred areas, and it preserves transparency if the source had any. The cost is size. A 1920x1080 screenshot is typically 400 KB to 1.5 MB as PNG, and a 12 megapixel photo can reach 15 to 25 MB. If you need a smaller file for a messenger or an email, run the result through compress and choose WebP or JPEG.
Because the output is built from canvas pixels, all metadata is discarded. EXIF timestamps, camera details and GPS coordinates are not present in the blurred copy; you do not need a separate pass through the EXIF remover, although that page is useful for checking what the original contained.
Limits
Regions are rectangles only; there is no freehand or elliptical selection and no pixelation style. The page handles one image at a time and takes JPG, PNG, WebP, GIF (first frame) and BMP. iPhone HEIC files need a trip through HEIC to JPG first. Everything happens locally in the browser and nothing is uploaded.