Aspect ratio explained

A profile picture that looked fine turns into a close-up of someone's forehead. Nothing went wrong — the software did exactly what it was designed to do.

Updated 28 August 2026 · 11 min read

You uploaded a perfectly good photograph. The site turned it into a picture of your forehead, or a strip of sky with a sliver of the actual subject in it. Nothing broke. A rule ran, it did exactly what it was written to do, and the result was terrible.

The rule is short enough to state in one sentence: fill the space from the centre and throw away whatever hangs over the edge. What follows is why that rule exists, when it hurts, and how to make the decision yourself before something else makes it for you.

The shape of a picture, separate from its size

Every image is a rectangle with a width and a height in pixels, and those two numbers do two different jobs. Together they say how big the picture is. Divided into each other, they say what shape it is — and the shape is the part that causes trouble.

An image 1600 pixels wide and 900 tall has the same shape as one that is 1280 × 720: divide width by height in each case and you get the same number, 1.78. The second is smaller in every dimension but not a different shape — scale one onto the other and they match exactly. That shared shape is written 16:9, those numbers reduced to the smallest whole pair.

So a ratio has no size in it. It is only a statement about how the two sides relate: wider than tall, taller than wide, or square. A 16:9 image can be a postage stamp or a cinema screen. Shape and size are separate decisions and go wrong independently — a photo can be the right shape and far too small, or enormous and the wrong shape.

The handful of shapes you actually meet

Only a few ratios are in common circulation, and each is a fossil of the equipment that produced it.

4:3 is the shape of old television sets and of many phone cameras left on their default setting: wider than tall, but not dramatically so. 3:2 comes from 35 mm film, whose frame measured 36 × 24 mm, and because most camera systems grew out of that format, interchangeable-lens cameras still shoot it. 16:9 is the shape of modern screens and nearly all video, which is why thumbnails and banners are almost always this ratio.

1:1, the square, has no photographic ancestry — it became common because grid layouts want every tile identical regardless of what was photographed. 4:5 and 9:16 are the vertical formats, and they exist because phones are held upright: 4:5 is a mildly tall rectangle that fills more of a scrolling feed than a square does, and 9:16 is 16:9 on its side, a full phone screen edge to edge.

The awkward part is that these shapes are mutually incompatible. A camera hands you 4:3 or 3:2, a video platform wants 16:9, a profile field wants 1:1. Nothing in that chain agrees with anything else, so a conversion happens at every step — and the only question is who decides how.

Three ways to force one shape into another

When an image of one shape has to occupy a container of another shape, there are exactly three things software can do about it. Every image tool, website and app you have ever used picks one of them.

Fit: keep everything, add bars

Scale the image until it fits entirely inside the container, then fill the leftover space with a solid colour. Nothing is lost from the picture; what you get instead are bars, down the sides or along the top and bottom depending on which way the mismatch runs. This is what televisions do with films wider than the screen. Fit is honest and rarely used for layout, because those bars look wrong in a page of neat tiles and the image ends up smaller than the space allocated to it.

Fill: crop away the overflow

Scale the image until it covers the container completely, then discard whatever sticks out. The container is full, the image is as large as it can be, and there are no bars. The cost is that some of your picture no longer exists in the displayed version.

This is the default nearly everywhere, for a defensible reason: it is the only option that makes a clean, consistent grid out of images that arrived in unpredictable shapes. The price is paid quietly, by individual images, in ways the layout never has to acknowledge.

Stretch: distort until it fits

Scale width and height by different amounts so the image matches the container exactly. Nothing is cropped, no bars appear, and everything in the picture is now the wrong shape. Faces widen, circles become ovals, buildings lean. This is where people looking mysteriously broad in a header image comes from. No competent tool does it on purpose — it happens when someone types a new width and height by hand without keeping the two linked.

Why it is always the head

Filling a container requires deciding which part of the overflow to discard, and the near-universal answer is: keep the middle, drop the rest equally from both sides. Centre-crop is simple, symmetrical, and needs to know nothing about the picture.

It is also, for photographs of people, close to the worst possible choice. Portraits are framed with the face somewhere in the upper portion of the frame, because a face pinned to the exact geometric centre looks wrong — the body gets cramped and the head sits low. The convention is so consistent that it is invisible until something exploits it badly.

So force an upright portrait into a square container. Fill scales the image until the square is covered, which means the top and bottom have to go, and it removes an equal slice from each end. The slice it takes off the top is the one containing the top of the head. The picture was composed well, the rule was applied correctly, and the output is a decapitation.

The same logic explains the useless landscape thumbnail. A wide photo squeezed into a square keeps the middle column and discards both sides, so a scene with the subject off to the left becomes a photograph of some background. A 16:9 header made from a vertical phone photo keeps a horizontal band from the middle — for a full-length portrait, somewhere around the waist.

Some platforms use face detection or let you choose a focal point. Neither is reliable enough to plan around. Crop to the required shape before you upload, and there is no overflow left for anything to make a decision about.

Cropping to a ratio versus resizing to one

This is the distinction worth carrying away, because getting it backwards is what produces distorted images.

Cropping changes the shape by removing pixels. You choose a rectangle of the required proportions, keep what is inside, discard what is outside. Everything that survives is untouched — the same pixels, at the same scale, in the same relationship to each other. A face keeps its proportions.

Resizing changes the size by scaling the pixels. Scale width and height by the same factor and the shape is preserved; the picture simply gets larger or smaller. That is the safe kind, covered in resizing without losing quality.

The failure is using a resize to change the shape. Take a 4:3 photo, type in 16:9 dimensions, and the software has no choice but to scale the width more than the height. Nothing is discarded, which sounds appealing, but every object is now wider than it was. Faces are the giveaway; you notice a stretched face long before a stretched wall.

So: to change shape, crop. To change size, resize. If a dialogue offers separate width and height fields, find the control that links them — usually a padlock or a “maintain proportions” checkbox — and leave it on.

Working out the largest crop that fits

When you crop to a fixed ratio, one dimension always runs out before the other. That dimension sets the limit, and the biggest crop available is the one where the limiting side is used completely.

Take a photo of 4000 × 3000 pixels, a standard 4:3 shot, and suppose you need it at 16:9.

Start by assuming the width limits you, and use all 4000 px of it. A 16:9 rectangle 4000 px wide needs a height of 4000 × 9 ÷ 16, which is 2250 px. You have 3000 px of height available, so that fits. The width really was the limit, and the largest possible crop is 4000 × 2250. Check it: 4000 divided by 2250 is 1.78, the same as 16 divided by 9. That leaves 3000 − 2250 = 750 px of height to discard, and a centre-crop takes 375 px off the top and 375 px off the bottom.

Now the same photo as a square. A square using the full 4000 px width would need 4000 px of height, and you only have 3000. So the width was not the limit this time — the height was. The largest square is 3000 × 3000, which discards 4000 − 3000 = 1000 px of width: 500 px from each side under a centre-crop.

And for a 4:5 vertical from the same original, the height is again the limit at 3000 px, so the width becomes 3000 × 4 ÷ 5 = 2400 px — a 2400 × 3000 crop that throws away 1600 px of width. The more the target shape differs from the original, the more of the original disappears.

The general form: multiply the full width by the target height and divide by the target width. If the result fits inside the image, the width was the limit. If it does not, flip it round and work from the height instead.

Circular avatars and the margin you should leave

Round profile pictures are not round files. An image file is always a rectangle; the circle is a mask applied at display time. The container is square and the visible region is the largest circle that fits inside it, so the four corners are hidden every time — a little over a fifth of the square's area, permanently out of view.

So if you crop your face to fill the square tightly, the circle clips the sides of your head and the top of your hair, because those are the parts nearest the edges. Leave breathing room: crop looser than feels right, with visible space on all four sides, and let the circle trim padding rather than person.

The wider version of that advice applies to anything you upload. Keep the subject away from the edges, because the edges are the first thing any crop takes. And if you already know an image is destined for a square — a profile field, a product grid — compose it as a square while you shoot: imagine the square in the middle of the viewfinder and keep everything important inside it.

Where a platform publishes exact pixel dimensions rather than just a shape, matching them is better still; the common ones are collected in social media image sizes. Where a shape is fixed by regulation rather than fashion, the framing rules are far stricter than any feed — see passport photo size requirements.

Common questions

Does cropping to a ratio lower the quality? No. Cropping removes pixels from the edges and leaves the survivors exactly as they were. The image has fewer pixels overall, so it can be shown at a smaller maximum size before it looks soft, but nothing remaining is degraded. Saving as JPG afterwards re-encodes it, which costs a small amount; PNG does not.

Why does my photo look fine in my gallery and wrong on the website? Your gallery is fitting the image, showing all of it with empty space around it; the website is filling a fixed container and cropping the excess. Same file, two rules, two results.

What ratio should I use if I do not know where the image will end up? Keep the original at whatever your camera produces and crop copies for each destination. Cropping from a large original is free; recovering the edges of something already cropped is impossible.

Can I add bars to make a photo the right shape instead of cropping it? Yes, and it is reasonable when the whole picture matters more than filling the frame. Note that some destinations then apply their own fill to your padded image, in which case you added bars for a crop that removes them again.

My image is the right ratio but still looks blurry. That is a size problem, not a shape problem. An image can be exactly 16:9 and only 400 px wide, at which point the correct shape does not help. Check the pixel dimensions the destination wants and start from an original at least that large.

Is 1600 × 900 really the same as 1280 × 720? The same shape, yes — both divide out to 1.78. Not the same size: the first has about 1.4 million pixels, the second about 0.9 million. Interchangeable to a layout, not to the eye.