You want to post a photo but there is a child in the background, a licence plate, a name badge, an address on a parcel, or an account number on a screen. The picture is fine; one region of it is not. This is one of the most common editing jobs there is, and it is done badly surprisingly often.
Blur and pixelate are not the same protection
A blur spreads each pixel's value across its neighbours. Detail becomes soft, but information about it is still present in the smeared result — which is why heavily blurred text has occasionally been recovered, and why blur is the weaker of the two choices for anything that must genuinely be unreadable.
Pixelation replaces a block of pixels with a single averaged colour. Everything inside the block that was not the average is gone; there is nothing left to reconstruct from. For a number, a name, or anything textual you truly need destroyed, coarse pixelation is the safer choice.
The practical rule: blur for things you want de-emphasised, pixelate for things you want unreadable. Filum offers both, each in light, medium, and strong.
Face detection is not the same as choosing what to hide
Many blur tools are specifically face blurrers: they find faces automatically and blur them. That is convenient and it solves the narrow version of the problem. It also leaves everything else in the frame untouched — the plate, the ID number, the home address on the parcel, the name badge, the customer record on the screen behind.
Automatic detection also fails in exactly the situations that matter: a face turned away, partially covered, at an angle, small in the frame, or belonging to someone the model was not trained well on. A tool that silently misses a face gives you false confidence, which is worse than no tool at all.
Drawing the regions yourself is slower and complete. You decide what is sensitive, including the things no detector would look for, and you can cover as many areas as one image needs. Filum works that way: draw a box around anything, add as many as you need.
The details that give away what you hid
Two things commonly leak. The first is a hard rectangular edge around the blurred region, which advertises both that something was hidden and exactly what shape it was. Softening the edges of each region removes that tell. Filum feathers them.
The second is the rest of the file. Every pixel outside the regions you drew should be left exactly as it was — for PNG, identical to the last bit. JPG and WebP have to be re-encoded once, because that is how those formats work, so the honest thing is to re-encode at high quality and show the measured quality of the result rather than degrading the whole picture silently in order to obscure one corner of it.
There is a third leak that has nothing to do with pixels: metadata. A photo can carry GPS coordinates, a device serial, and a timestamp, none of which a blur touches. If the reason you are blurring is that the image should not identify a place or a person, the metadata deserves the same attention as the pixels.
The obvious problem with uploading it
A tool that hides a sensitive part of a photo should not require you to send the un-hidden photo to someone else's server first. That is precisely backwards, and it is how nearly every online blur tool works.
Filum's Blur Image runs entirely inside your browser on your own device. The picture never leaves your computer and never reaches a server, blurred or otherwise. It supports JPG, PNG, and WebP.