Aspect-Ratio Crop, Explained

After reading this you will know how aspect ratio and pixel geometry combine when you crop, how to hit a target ratio without stretching an image, and how to avoid the small mistakes that soften or re-encode your output.

A crop removes pixels from the edges of an image and keeps a rectangular region. Nothing inside that region changes: a pixel at the center keeps its exact color and position relative to its neighbors. What changes is the frame around it.

The interesting constraint is aspect ratio. A profile picture must be square. A YouTube thumbnail must be 16:9. A story must be 9:16. If you crop by eye you will land near those ratios but not on them, and the platform will then crop again or add bars. Locking the ratio while you drag solves this: you choose where the rectangle sits, and the tool guarantees the shape.

What aspect ratio actually means

Aspect ratio is the ratio of width to height, written W:H. It says nothing about size. A 1920 x 1080 image and a 1280 x 720 image share the ratio 16:9 because both reduce to the same fraction.

R = \frac{W}{H}

Here R is the aspect ratio as a single number, W is the width in pixels, and H is the height in pixels. For 16:9, R = 16/9 \approx 1.778. For a square, R = 1. For a 9:16 story, R = 0.5625. Reduce any ratio by dividing both numbers by their greatest common divisor: 1920:1080 divided by 120 gives 16:9.

Landscape
Wider than tall, so R \gt 1. Example: 16:9 at 1.778.
Portrait
Taller than wide, so R \lt 1. Example: 9:16 at 0.5625.
Square
Equal sides, R = 1. Example: 1:1 at 1080 x 1080.

When to crop and when to resize instead

Crop when the frame is wrong: too much sky, an off-center subject, a ratio that does not match the target. Resize when the composition is already right but the pixel count is wrong.

The difference matters because they are not interchangeable. Cropping a 1.778 landscape to a 1.0 square throws away real content on the sides. Resizing the same image to a square instead stretches it, so faces widen and circles turn into ellipses. If your ratio must change, crop. If only the size must change, use the Image Resizer.

To change ratio without discarding content, add padding rather than cropping. A 16:9 image placed on a square canvas keeps every pixel and fills the top and bottom with a background color. That is a composition choice, not a crop.

The geometry of a locked crop

When the ratio is locked, the crop rectangle has one degree of freedom in shape: pick a width and the height follows, or pick a height and the width follows.

H_{crop} = \frac{W_{crop}}{R} \qquad W_{crop} = H_{crop} \times R

Here W_{crop} and H_{crop} are the crop dimensions in pixels and R is the locked ratio. The rectangle must also fit inside the source. If the source is 1920 x 1080 and you want a 1:1 crop, the largest square that fits has side equal to the shorter dimension, 1080, because a side of 1920 would run off the top and bottom.

S_{max} = \min\left(W_{src},\; H_{src} \times R,\; \frac{W_{src}}{R} \cdot R\right)

In plain terms, the maximum crop width is capped by both the source width and by H_{src} \times R. For a square in a 1920 x 1080 frame that cap is 1080 \times 1 = 1080, so the largest square is 1080 on a side. You then slide that square left or right to choose which slice of the width to keep.

A worked example with the demo image

Cropping a 1920 x 1080 frame to a centered square

Start from the demo defaults: a source of 1920 x 1080, ratio locked to 1:1, centered.

  1. Find the ratio number: R = 1/1 = 1.
  2. Find the largest fitting side: S_{max} = \min(1920,\; 1080 \times 1) = 1080. The crop is 1080 x 1080.
  3. Center it horizontally. The leftover width is 1920 - 1080 = 840 pixels, split evenly, so the left edge sits at 840 / 2 = 420.
  4. The crop rectangle is therefore x from 420 to 1500, y from 0 to 1080.
  5. Pixels kept: 1080 \times 1080 = 1{,}166{,}400. Pixels discarded: 1920 \times 1080 - 1{,}166{,}400 = 907{,}200, which is 43.75 percent of the original.

That 43.75 percent is not damage, it is deliberate removal. The kept 1080 x 1080 region is pixel-for-pixel identical to the same region in the source, assuming lossless output.

The square is limited by the 1080 height, and 840 pixels of width fall outside it.

Explore how ratio reshapes the crop

On a fixed 1920 x 1080 source, a 1:1 lock gives a 1080 x 1080 crop keeping 56.25 percent of pixels; 4:3 gives 1440 x 1080 keeping 75 percent; 16:9 keeps the whole frame at 1920 x 1080; and 9:16 gives 608 x 1080 keeping 31.6 percent.

Reading the output dimensions

The crop produces an image whose pixel dimensions are exactly the crop rectangle. A 1080 x 1080 crop is a 1080 x 1080 file. There is no resampling, so no blur is introduced by the crop itself.

Check the ratio of the result against the target. Divide output width by output height and compare to your intended R. For the worked example, 1080 / 1080 = 1.000, which matches 1:1 exactly. If you had cropped by eye to 1075 x 1082, the ratio would be 0.994, close but not square, and a platform expecting 1:1 would nudge it.

Largest crop from a 1920 x 1080 source at common ratios
TargetRCrop WCrop HPixels keptPercent kept
1:11.00010801080116640056.25
4:31.33314401080155520075.00
3:21.50016201080174960084.38
16:91.778192010802073600100.0
9:160.5625608108065664031.67

Common mistakes

The most frequent error is confusing crop with resize, covered above. Three others come up often.

Re-encoding a JPEG needlessly. A crop keeps pixels, but saving the result re-encodes the file. A JPEG saved from a JPEG loses a little more each time. If the source is a photo, one careful re-encode is fine. If the source is a screenshot, logo, or diagram, save as PNG so the crop stays exact.

Rounding the ratio wrong. A 9:16 crop of a 1080-tall source is 1080 \times 0.5625 = 607.5 pixels wide. That is not an integer. The tool must round to 608 or 607, which shifts the true ratio to 0.5630 or 0.5620. The error is tiny but real, so trust the reported dimensions rather than assuming perfect fractions.

Cropping first, then upscaling to recover size. If you crop to 1080 x 1080 and then enlarge to 2160 x 2160, you cannot recover detail that the source never had. Upscaling interpolates. Plan the crop so the region you keep already has the pixels you need.

Watch out when a source has an odd starting ratio. A phone screenshot at 1179 x 2556 is roughly 0.461, narrower than a 9:16 story at 0.5625. Cropping it to 9:16 removes height, not width, and you may clip content near the top or bottom that you meant to keep.

Related tools

After cropping, you often want to change the file size or the pixel count. Shrink the encoded bytes with the Image Compressor, which shows the output size as you adjust quality. Enlarge a small crop with the Image Upscaler when you need more pixels than the crop produced. If you are preparing a site icon, crop to a square first, then feed it to the Favicon Generator for every required size.

Frequently asked questions

Does cropping reduce image quality?

The crop itself does not: kept pixels are unchanged. Quality loss comes only from re-encoding on save. Choose PNG for graphics and screenshots to keep the crop exact, and accept one light JPEG re-encode for photos.

What aspect ratio should a profile picture be?

1:1. Most platforms display avatars in a circle or square, so a 1080 x 1080 square crop centered on the face works everywhere.

How do I crop for Instagram stories or Reels?

Use 9:16, a ratio of 0.5625. A common target size is 1080 x 1920. Starting from a landscape source you will keep only a narrow vertical strip, so center the crop on your subject.

Why is my cropped width not an exact number?

Because pixels are integers and most ratios are not. A 9:16 crop of a 1080-pixel height needs a width of 607.5, which rounds to 608. The resulting ratio, 0.5630, differs from 0.5625 by less than one tenth of one percent.

Can I crop to exact pixel coordinates instead of dragging?

Yes. Setting explicit left, top, width, and height gives a repeatable crop, which is useful when you process several images that share the same layout, such as screenshots from the same device.