Resize Image

Pick a target size from where the image will actually live

The right dimensions are dictated by the destination, not by a round number that feels tidy. A full-width hero on a content site rarely needs more than 1920px across; most article body images are displayed at 700–1000px and anything wider is bytes the visitor downloads and never sees. An email header tops out around 600px because that's the historic content width Outlook and Gmail render. A marketplace or forum avatar is usually stored at 400px square even if it displays at 96px, so it stays crisp on high-DPI screens. Resize to roughly 2× the largest display size and you cover Retina without paying for a 6000px camera original.

Sending an unresized phone photo is the most common waste. A modern 12-megapixel capture is 4032×3024 — fine for printing at A3, absurd as a 280px thumbnail in a chat. Downscaling it first cuts the file from several megabytes to tens of kilobytes before you even reach for Compress Image, and the two steps compound: resize sets the pixel count, compression tunes the bytes per pixel.

Common target widths, and what they're for

These are display-side widths. Double them for crisp rendering on high-density screens, then let compression decide the final file size.

Social platforms re-encode whatever you upload, so feed your largest sensible source and let them downscale — see the per-network presets in the [Aspect Ratio tool](/creator/aspect-ratio/).

Why upscaling can't add what isn't there

Resampling is interpolation: to enlarge an image the resampler invents intermediate pixels by averaging the ones it has. Bicubic and Lanczos do this gracefully, which is why downscaling stays sharp — you're discarding data and choosing the cleanest survivors. Upscaling is the reverse and has no new information to draw on, so past roughly 2× you get softness, haloing on hard edges, and visible smearing on text. No browser-side resizer recovers detail that the sensor never recorded; that needs a trained super-resolution model guessing plausible pixels, which is a different category of tool with its own artefacts. If you need a bigger usable image, the fix is almost always to re-export from a higher-resolution source, not to upscale the small one.

Resize, crop, or compress — they solve different problems

• Resize changes the pixel dimensions while keeping the whole frame — every part of the picture survives, just smaller or larger.
• Crop changes which part of the frame you keep and almost always changes the aspect ratio — use it when the composition or shape is wrong, not the size.
• Compress keeps the dimensions identical and only reduces bytes-per-pixel — use it when the picture is right but the file is too heavy.
• Common mistake: resizing to hit a file-size limit. Dimensions and file size are only loosely linked — a 1500px JPEG at high quality can be heavier than a 2500px one at moderate quality. Set the size you need for display, then compress to the byte budget.