Skip to content
Server-sideDeleted in 30 minutes

Resize Images Online

Drop in images and set a width, a height, or a box to fit them inside. The resampling runs on our server with libvips, which uses a proper Lanczos filter rather than the bilinear scaling a browser canvas does — visible on any reduction of more than about half. Aspect ratio is preserved unless you explicitly ask for a mode that does not.

Use this without the search next time. Prathom Workbench puts Prathom's tools in your toolbar.

Add to Chrome — free

What it does

  • Resize by width, height, percentage, or fit box
  • Batch processing with per-file results
  • Multi-step downscaling for sharper output
  • Optional format change to JPEG, PNG, or WebP
  • Deleted within 30 minutes, no account, no watermark

How to use Image Resizer

  1. 1

    Add your images

    Drop one or more image files onto the box, or click it to browse. PNG, JPEG, WebP, GIF, AVIF, BMP and SVG are all accepted.

  2. 2

    Choose how to size them

    Pick a mode. Width and Height set one dimension and let the other follow. Fit inside box keeps the whole image. Fill box crops to exact dimensions.

  3. 3

    Decide about enlarging

    Leave "Never enlarge" ticked unless you specifically want an image made bigger than it started, which cannot add detail that was not there.

  4. 4

    Save the results

    Each result shows its new dimensions and file size. Save them one at a time, or download the whole batch as a single ZIP file.

How it works

Each file is decoded into a bitmap, drawn into a canvas at the new size, and re-encoded. Three parts of that are worth explaining, because they are where resizers differ.

Choosing the target size. Width and Height modes take one dimension and compute the other from the source ratio, so the result is never distorted. Fit inside box scales until the whole image sits inside the rectangle you gave — usually leaving one dimension short of the number you typed, which is correct rather than a bug. Fill box does the opposite: it scales until the box is covered and crops the overflow from the center, guaranteeing exact dimensions at the cost of some edges. Exact size forces both numbers and is the only mode that can distort; it is last in the list for that reason.

Stepping down. Sampling a 4000-pixel-wide image directly into 200 pixels takes roughly one source pixel in twenty and discards the rest. Thin lines break up and fine textures sparkle. Halving repeatedly instead means every source pixel contributes to an average at each stage, and the final image looks the way you would expect. The tool calculates how many halving passes a given reduction needs and applies them automatically.

Orientation. Photos taken with a phone held sideways are stored the right way up with a rotation flag attached, not rotated on disk. Decoding without honoring that flag produces sideways output — and only for the pictures that happen to have been shot in portrait, which is why the bug survives so long in so many tools. Images here are decoded with orientation applied.

When you'd use this

The everyday case is a content management system or forum with an upload limit, where a 4 MB phone photo is refused and a 1200-pixel version sails through.

Batch work is the other. Product photos, a folder of screenshots for documentation, or event pictures that all need to be the same width are tedious one at a time and trivial in a batch. Because each result is listed separately with its own dimensions and size, you can see immediately if one file behaved differently from the rest — usually the one that was portrait when the others were landscape.

There is also the deliberately awkward case: an avatar or thumbnail slot that demands an exact square from images that are not square. Fill box handles this without you having to crop each one by hand first, and it crops from the center, which is right far more often than not.

What happens to your photo

It is uploaded, resampled here, and deleted. Being specific about that is worth more than a reassuring sentence.

The resampling runs on our server because it produces a better result than a browser canvas does — libvips applies a proper Lanczos filter, where a canvas does bilinear scaling that visibly softens any reduction of more than about half. That is the reason for the trade, and the trade is real: the file is on our machine for the length of the operation.

What can be said precisely is how long. The upload is deleted as soon as the resize finishes; the result is deleted on a thirty-minute timer swept every five minutes, so nothing survives past about thirty-five. There is no account, no history, and no copy kept afterwards.

For a landscape wallpaper none of that matters. For a photograph of a document, a whiteboard after a meeting, a passport page, or a picture of your children it might, and the honest answer is that a tool which never receives the file at all is safer than any promise about how long we keep it. ImageMagick and sharp are both free and run on your own machine.

Examples

Fitting a phone photo to a 1200px blog width

IMG_4821.jpg — 4032 x 3024, 3.8 MB Mode: Width, 1200 px
IMG_4821.jpg — 1200 x 900, 412 KB 89% smaller

The height was never typed in. Width mode derives it from the original ratio, which is why 4032x3024 becomes 1200x900 rather than something slightly off. Note that the file shrank far more than the pixel count did, because JPEG size falls faster than area when fine detail is averaged away.

Making a square thumbnail from a landscape shot

hero-banner.png — 2400 x 1000 Mode: Fill box (crop), 600 x 600
hero-banner.png — 600 x 600 cropped to fill the box

A 2.4:1 image cannot become a square without losing something. Fill box takes the largest centerd square the source allows and scales that, and the result panel says "cropped" so you know pixels were discarded rather than squashed. Fit inside box on the same file would have given 600x250 instead.

Frequently asked questions

Does resizing here lose quality compared with Photoshop?

For downscaling, the difference is very hard to see. This tool halves the image repeatedly before the final draw rather than jumping straight to the target size, which is what stops the jagged, sparkly edges you get from a single large reduction. Photoshop uses more sophisticated filters, but on photographs the gap is marginal.

Why is "Never enlarge" on by default?

Because enlarging cannot invent detail. Scaling a 400px logo up to 2000px gives you a 2000px file that looks softer than the original at every size, and people asking for a big number almost always want the image to fit that slot rather than to be stretched to it. Untick it if you genuinely need the larger canvas.

Can I resize to an exact file size, like under 100 KB?

Not with this tool — it targets dimensions. Use the image compressor instead, which searches for the highest quality setting that fits a byte budget you specify, and reduces dimensions automatically if quality alone cannot get there. Resizing first and then compressing is often the best combination.

What happens to a transparent PNG if I output JPEG?

Transparent areas are filled with white before encoding. JPEG has no transparency channel at all, so something has to go there, and unfilled pixels would come out black. If the transparency matters, choose PNG or WebP as the output format instead.

Is there a limit on how many images I can do at once?

There is no fixed cap on how many files you add at once, but each one is uploaded and processed in turn, so very large batches of very large photos can exhaust it. In practice a few dozen ordinary camera images at a time is comfortable on a laptop. The tool processes them one at a time and keeps going if any single file fails.