Image resizer
Exact pixel dimensions, a percentage, or a ready-made preset for Instagram, YouTube or a 2×2 inch print. All of it on your own device.
- 0 uploads
- No limits
- No signup
- Works offline
Pixels, percentages and presets
Three ways to say the same thing, useful in different situations.
Pixels are what forms, websites and specifications ask for. Type a width, leave the height empty, and the proportions are preserved automatically.
Percentage is for “make this smaller” without a specific destination — halving a folder of holiday photos before emailing them, for instance.
Presets cover the sizes people look up repeatedly:
| Preset | Pixels | Used for |
|---|---|---|
| Instagram post | 1080 × 1080 | square feed posts |
| Instagram story | 1080 × 1920 | stories and reels |
| YouTube thumbnail | 1280 × 720 | video thumbnails (16:9) |
| LinkedIn banner | 1584 × 396 | profile cover image |
| X / Twitter header | 1500 × 500 | profile header |
| Facebook cover | 820 × 312 | page cover photo |
| 2 × 2 inch @ 300 dpi | 600 × 600 | US passport and visa photos |
| 35 × 45 mm @ 300 dpi | 413 × 531 | Schengen, UK and most Asian passports |
| A4 @ 300 dpi | 2480 × 3508 | full-page print |
When the ratio does not match
If your photo is 4:3 and the target is 16:9, something has to give. There are only three honest options and the tool asks you to pick one:
- Crop to fill keeps the target dimensions exactly and cuts the overflow from the edges, centred. This is what you want for thumbnails and banners.
- Fit inside keeps the whole image visible and lands on the target’s longest edge, so one dimension comes out smaller than requested.
- Stretch forces both dimensions and distorts the picture. It is included because occasionally it is genuinely what someone wants, but it will make faces look wrong.
For precise control over which part gets kept, use the cropper instead — it lets you pan and zoom inside the frame before committing.
Why Lanczos3
Downscaling is a resampling problem. Each output pixel has to represent an area of the source, and how you average that area determines how much detail survives.
Nearest-neighbour picks one source pixel and ignores the rest, which produces jagged edges and aliasing. Bilinear averages four, which is smoother but soft. Lanczos3 uses a windowed sinc function over a 6×6 neighbourhood, which keeps edges crisp without introducing the halos that oversharpening causes. It is slower — noticeably so on very large images — and it is worth it.
The implementation here is the same WebAssembly resizer that Squoosh uses, running in a Web Worker so a batch of fifty images does not freeze the page.
After resizing
Most people resizing for a form need to hit a byte budget too. The result screen offers a one-click hand-off to the compressor, and your resized file travels with it — no second download, no re-selecting files.
How it works
- 01Add your imagesDrop them on the page, click to browse, or paste from the clipboard.
- 02Choose how to size themType exact pixels, drag a percentage, or pick a preset. Leave the height empty to scale proportionally from the width.
- 03Decide what happens to the ratioWhen the target ratio differs from the source, choose to crop to fill, fit inside, or stretch.
- 04DownloadSave one file or the whole batch as a zip.