Image Resizer
Resize to an exact pixel size, a percentage, or a known preset, with high-quality stepped downscaling.
Pixels. Leave blank to derive it.
Pixels. Leave blank to derive it.
Relative to the original size.
Caps the long side, whichever it is.
Enlarging cannot add detail, only file size.
Drop an image here
or paste a screenshot with Ctrl+V. Nothing is uploaded.
Everything here runs on your device. Nothing you enter is uploaded or stored.
Resizing sounds like the most trivial operation in image editing, and the difference between doing it well and doing it badly is entirely in how the pixels get sampled. A 4000 pixel photograph taken down to 400 in one step looks visibly worse than the same reduction done in stages, and most online resizers do it in one step.
How to use it
- Drop an image onto the panel, press Choose images, or paste from your clipboard. The dimension fields fill in with the real size.
- Pick how you want to express the new size — exact pixels, a percentage, a preset, or a cap on the longest edge.
- Leave the aspect ratio locked unless you want to stretch the image on purpose.
- Press Download, or select several files first and use the batch button for a ZIP.
Why stepped downscaling matters
Browsers offer exactly one image scaling algorithm, and at best it is bilinear filtering: each output pixel is blended from a small neighbourhood of input pixels. That works well when the reduction is modest.
It falls apart on large reductions. Shrinking by a factor of ten means each output pixel should represent about a hundred input pixels, but bilinear filtering only looks at four of them. The other ninety-six are discarded entirely. Whether a fine detail survives becomes a matter of whether it happened to land on a sampled pixel, and the result is aliasing — shimmering moiré across brickwork, hair and fabric, and text that breaks into fragments.
Repeated halving fixes it. Each halving averages a small neighbourhood correctly, and doing it several times means every source pixel contributes to the result. Halve until the image is within twice the target, then resample once more to land exactly on it. This is what the tool does, and it reports the stages it used underneath the preview so you can see it happening.
Four ways to ask for a size
People arrive with the size in different forms, and forcing everyone through a width field is needless friction.
Exact pixels is for a known requirement — an avatar that must be 256 by 256, a banner that must be 1500 wide. Fill in one field and the other follows the ratio.
Percentage is for “half the size”, which is how people usually think about a photo that is simply too big.
Presets cover known destinations, each with the reason attached. A list of round numbers is a worse interface than a plain field, but “1920 — a desktop wallpaper or presentation slide” tells you whether it is the one you want.
Longest edge is the most useful and least offered. Cap the long side at 1600 and a landscape photo becomes 1600 wide while a portrait becomes 1600 tall. It is the only mode that handles a mixed folder of both orientations sensibly in one pass.
What it does not do
It will not target a file size in kilobytes — that is the compressor’s job, because bytes depend on content and not just dimensions. It does not crop, so asking for a square from a rectangle fits the whole image inside rather than filling the square; use the cropper for that. It does not upscale intelligently, because no browser API does. Very large outputs are capped at the canvas limit and you are told when that happens.
Nothing is uploaded. Everything runs on your device.
Preset sizes
| Preset | Size | Use |
|---|---|---|
| 4K | 3840 × 2160 | High-resolution display or print proof |
| Full HD | 1920 × 1080 | Wallpaper or presentation slide |
| HD | 1280 × 720 | Video thumbnails, older displays |
| Web content | 1200 wide | In-article image at full column width |
| Blog body | 800 wide | Standard content-width image |
| Thumbnail | 400 wide | List or grid thumbnail |
| Avatar | 256 × 256 | Square profile picture |
Questions
Why does my resized image look soft or shimmery?
Because it was probably reduced in one step. When a browser shrinks an image by more than about half in a single pass, it samples too few source pixels for each destination pixel, so fine detail turns into aliasing and shimmer. The fix is to halve repeatedly until you are within twice the target size, then do a final resample. This tool does that automatically and tells you how many stages it used.
Can I resize an image to a larger size without losing quality?
No, and no tool can. Enlarging cannot add detail that was never captured — it can only interpolate between the pixels you already have, which produces a softer, larger file. That is why never-enlarge is on by default. If you genuinely need a bigger canvas, turn it off, but expect softness rather than detail.
How do I resize an image to an exact file size like 200KB?
That is a different job, and the compressor does it. File size depends on the encoder and on the image content, not just the dimensions, so there is no pixel size that reliably produces 200KB. The compressor searches for the quality setting that fits a byte target. Reducing dimensions here first is still the most effective way to make a file smaller.
What happens if I only fill in the width?
The height is worked out from the original aspect ratio, so the image keeps its proportions. Fill in both and turn off the ratio lock to stretch it deliberately. Fill in both with the lock on and the image is fitted inside that box instead, so it never distorts.
Last updated