Aspect Ratio Crop Calculator

About this tool

Every editor asks the same two questions when you drop a clip into a different shape: do you want the frame filled, or the whole picture kept? Cover fills the frame and throws away whatever falls outside; contain keeps every pixel and pays for it with black bars. The two are easy to confuse because both produce a correctly shaped video — the difference only shows up in what is missing from it.

This page answers both at once, with the geometry made explicit. The important number is not the output size, which is usually what you asked for, but the crop rectangle: the region of the source that survives. A 1920×1080 clip fitted to a 9:16 frame keeps a 607.5-wide strip out of 1920 — roughly a third of the picture. Knowing that a strip, not the frame, is the real subject is the difference between a usable crop and a ruined one.

The anchor decides which part survives. Because cover always leaves slack on one axis only, three positions are enough: centre, and the two ends. That covers the two habits that matter — hide the middle of a wide shot, or protect the top of a tall one — without the nine-cell grid that makes people pick the wrong corner.

The comparison block reports both modes for the same input, including the pixels lost or added. It is deliberately not the answer to "which is better": cover keeps the frame and loses content, contain keeps content and loses the frame. The tool makes the trade visible rather than making it for you.

Why the fitted numbers come out exact, not 0.9999

The geometry is computed from the ratio as a fraction rather than as a decimal like 16/9. Working with the two integers avoids the float error that turns a clean 607.5 into 607.4999999924.

A ratio such as 16:9 is not exactly representable as a decimal. If the code turned it into 1.7777… first and then divided, a 1080-tall frame would yield a width of 607.4999999924 — a number that looks right on screen but fails any exact comparison. The tool keeps the two sides separate and multiplies instead: height times the numerator, divided by the denominator.

The output height is handled the same way, for the same reason. It is recomputed from the ratio rather than by scaling the canvas height, because multiplying the canvas by the scale factor carries its own floating-point tail — one measured case produced 1920.000000024 instead of 1920.

The remaining noise, where it cannot be avoided at all, is trimmed at the exit: values are rounded to ten decimal places on the way out, so the question of whether the fitted rectangle exactly matches the target ratio can be answered with a tolerance instead of an equality that would always be false.

What the rectangle used means in each mode

Under cover it is the crop box — the slice of the source that survives. Under contain it is the entire source, because contain keeps every pixel and puts the cost into the bars instead.

The same output field reports a different thing depending on the mode, and that is deliberate. Cover throws pixels away, so the region actually used is smaller than the source; contain throws none away, so the region used is the full frame, and the surplus shows up as padding around it.

Getting this wrong is easy and visible: if contain reused the crop box from cover, the used region would come out taller than the source itself — an area of 1920 by 3413 taken from a 1920 by 1080 frame. The two are therefore computed separately, and the picture rectangle in the output is measured from the used region, not from the whole source.

Why the trim and pad percentages share no denominator

Trimming is reported as a share of the source pixels; padding as a share of the canvas pixels. Cover measures loss against what you started with, contain measures added bars against what you end up with.

The two percentages answer two different questions, so they are measured against two different wholes. A trim tells you how much of the original frame is gone — the removed pixels over the source pixels. Padding tells you how much of the finished frame is bars — the added pixels over the canvas pixels.

Sharing a single denominator would be tempting and wrong, because in cover the canvas is smaller than the source and in contain it is larger. Each percentage is measured against the frame that actually contains the change — the source for what was taken away, the canvas for what was added.

Why only one axis ever changes

Cropping 1920×1080 to 1:1 with cover moves width only: the result is 1080×1080, 420 pixels dropped from each side and no height cropped at all. With contain, height is the one that moves: the canvas becomes 1920×1920, 420 pixels added top and bottom. Both axes never change at once.

Cover means scaling until the frame covers the target ratio. When the source is wider than the target, height is the constraint and width is trimmed; when the source is taller, it reverses. Covering needs only one direction to be filled, so the other necessarily has surplus.

That fact is why the anchor needs only three values. With a single axis carrying the surplus, center, start and end describe where that surplus goes. If both axes could be trimmed, the anchor would need nine combinations.

Contain moves one axis too, in the opposite direction: padding 1920×1080 out to 1:1 fills height to 1920 and leaves width untouched. So cover and contain change the same number of pixels in opposite directions, which is exactly why both modes have to be offered.

Only when the source ratio equals the target ratio does neither axis move and the change come out as zero. Switching modes then shows no difference, and that is not a failure to compute: this input simply has nothing to trim or pad.