Resize image dimensions

Two steps, one pass — why that matters

Getting an image to a particular shape usually takes two moves: scale it down to roughly the right size, then trim it to exactly the right frame. Done in two separate tools, or in two separate saves, the picture is resampled twice. The first pass throws away detail to reach the smaller size, and the second pass then interpolates what is left a second time. The damage compounds. It shows up as soft edges on text and logos, and as shimmering moiré in fine patterns like fabric, brickwork or hair.

This page composes both steps into a single mapping before drawing anything. The intermediate size exists as a number and never as pixels: when you ask for 4000 by 3000 scaled to 400 by 300 and then cropped to 300 by 300, the tool works out which rectangle of the original that final square corresponds to, and reads the original once. The geometry is identical to doing it in two moves. The quality is not.

How to resize image px values exactly

Both tabs show the same thing: the pixels you will get. They differ in how they get there. Size resizes the whole picture, and any crop keeps its share of it. Crop or pad leaves the picture at its current size and moves the frame around it — smaller numbers crop, larger numbers pad. Because neither tab stores a step that a later edit can quietly redefine, it makes no difference which one you use first.

That matters more than it sounds. A tool that stores “scale to 200” and “canvas 1000” as two separate instructions will silently turn your crop into padding the moment the scale drops below it: the numbers stay where you typed them and the meaning changes underneath. Here the crop is held as its share of the picture, so shrinking the picture shrinks the crop with it and a crop stays a crop.

The row above the preview shows the chain as it stands, and says what each step did rather than only what it produced — original, resized, cropped, padded. Three numbers in a row look identical whether the last one cropped or padded, and that difference is exactly the one worth seeing before you download.

Enlarging, and what it can and cannot do

The output is allowed to be larger than the original in either step. Enlarging cannot recover detail that was never captured, so the result is softer than the source, and the page says so as soon as the plan involves any enlargement rather than letting you find out after downloading. That trade is often still worth taking: a slightly soft image at the dimensions a platform insists on is more useful than a crisp one it refuses to accept. Knowing which dimensions a platform actually insists on is the harder half of that question; for the two LinkedIn publishes, and the difference between them, see LinkedIn banner size. Some platforms insist on less than people assume: Instagram names a width and a range of heights around it, so several shapes are equally correct, and Instagram post size quotes the sentence that says so.

The new area, and which formats can carry it

Padding to a larger canvas creates area that no original pixel reaches. By default that area is left genuinely transparent, and the checkerboard in the preview is what a PNG will contain. Choose a solid color instead when the destination will sit the image on a known background and you would rather pick the surround yourself. JPEG has no alpha channel at all, so selecting JPEG while the new area is transparent means it becomes white in the file — the page states that before the download rather than after, and the preview switches to match so that what you see is what you get.

Which part survives the crop

Once the canvas step is on, the chain has to decide what falls outside it. Cropping around the center is a reasonable default and a poor rule — the thing worth keeping is rarely dead center. Drag the subject inside the frame, or use the sliders, and the crop happens around what you chose. There is deliberately no zoom control here: the subject's size is already the scale step above, with its own exact width and height, and a second control for the same thing would either do nothing or argue with the first.

The PNG this page writes

Browsers are poor PNG encoders. Measured here on September 6, 2026: an 800 × 500 chart that is 4,938 bytes on disk came back from the browser's own toBlob at 21,180 bytes — more than four times the original, at identical dimensions and with identical pixels. Nothing was lost; the encoder simply does not choose a color type from the image or pick a compression filter per row, and every PNG that has been through an optimizer already has both.

So this page writes the PNG itself. It counts the distinct colors in your actual output and writes a palette when there are 256 or fewer, packing to four bits per pixel when there are 16 or fewer; it writes grayscale as one channel instead of four; and it chooses a filter for every row with the same heuristic libpng uses. That measured 74% smaller on the chart and 31% smaller on a photograph, where there is no palette to find and the filtering does all the work. Not one pixel changes: the palette holds exact values and the filters are reversible, which is what makes this different from the quantizers that shrink PNGs by discarding color.

Both encoders run, and the smaller file wins — so the download can never be larger for having tried. The work happens in a background thread with the row count shown as it goes, because it is real work on a large image and a frozen tab is indistinguishable from a broken tool.

When another page is the better tool

If you only need to know what you are starting from, check the image dimensions instead — it reads the file and changes nothing. If the target is simply a square, making the image square is one decision rather than two, and it shows what cropping and padding each cost before you choose. For a profile picture that a platform will mask, cropping to a circle shows exactly which corners disappear. When the same dimensions have to reach a whole folder of images, the bulk image resizer applies one setting across every file in a single pass. And if the picture is already the right shape but too heavy, compressing it reduces the bytes at a quality you choose and leaves the dimensions alone — which is the lossy answer to the size question, where the paragraph above is the lossless one.

This page is deliberately not the biggest tool in that list. It is the one that gives you exact control over one image and the two moves that most often have to happen together.