Make a WhatsApp Profile Picture

How it works

Your reference image goes to an image-editing model with an instruction describing a centerd square profile image: a clear face or subject, strong contrast at small size, safe cropping, identity preserved, no text. A second instruction rules out distorted subjects, unreadable text, watermarks, bad crops, and clutter.

A new picture is generated at exactly 1080 by 1080 pixels. Because it is generated rather than cropped, a vertical or horizontal source becomes a square composition instead of a square cut out of a rectangle, and a second run gives a different picture rather than the same one again.

A fifth of the file is decoration

WhatsApp stores a square and draws a circle.

Draw a circle inside a square and it covers about seventy eight per cent of the area. The rest is four corner wedges, and in this app they are not merely de-emphasized — they are never rendered, in any view, at any size. If something sits in a corner it does not exist.

That has a consequence people rarely plan for. The safe region is not a smaller square in the middle; it is round. Content near the middle of an edge survives, because the circle touches the edge at its midpoint. Content the same distance from center but on a diagonal does not.

Square logos are the usual casualty. A mark designed with a border, a boxed background, or elements pushed into the corners loses precisely those elements and keeps only its middle.

The number was already legal

Most social formats on this site need their dimensions corrected before generation. This one does not, and it is worth saying why.

Images are generated in a latent space where both dimensions must be multiples of eight. 1080 divided by eight is 135 exactly. So the generator is asked for 1080 by 1080 and returns 1080 by 1080, with no correction step, no scaling pass, and nothing to go quietly wrong.

Square social formats are generally the well behaved ones for this reason. Trouble comes from formats invented for a layout rather than for a renderer.

Your upload is not resized first

This graph is unusual among the ones on this site: it hands your file to the model at whatever size you uploaded it, with no downscaling step in between.

Everywhere else, a resize normalizes the input to about one megapixel before anything reads it. Here the reference goes in whole. A twelve megapixel phone photograph and a one megapixel screenshot are therefore not treated equivalently, and the same subject uploaded at two different sizes can produce noticeably different compositions.

The output is 1080 by 1080 either way. It is the interpretation that shifts, not the result size. If a generation comes back oddly framed, resizing your source down and running it again is a real thing to try, and it is not obvious from the outside that it would make any difference.

Contrast at fingernail size

Almost nobody looks at your profile picture.

They see it at roughly forty to fifty pixels across, next to a message, in a list of thirty other people, while scrolling. At that size a face is about a dozen pixels wide and detail is meaningless. What survives is the difference between light and dark, and the outline of one shape against its background.

This is why a photograph that is beautiful full screen can be useless as a profile picture. Soft light, a mid-toned background behind a mid-toned subject, a wide shot where the face is small — all fine at full size, all indistinct at fifty pixels.

The instruction asks for strong contrast at small size for exactly this reason, and it is the one thing worth checking yourself. Shrink the download and look at it in a row of other pictures.

Publication gate

This page ships once the workflow has been run against a vertical portrait, a horizontal photograph, and a square logo, with each output measured at 1080 by 1080, masked to a circle to confirm what survives, and viewed at fifty pixels against both a light and a dark chat background.