How to Resize an Image to Exact Dimensions or by Percentage Without Stretching It

Resizing an image sounds like a one-field operation: type a number, hit go. Then the upload form rejects your file for being a few pixels off, or the photo comes back with stretched faces, or you blow a 200×200 logo up to 4K and it looks like wet cardboard. None of those are bugs in the tool. They are the difference between knowing what resizing actually does to your pixels and guessing.

This guide covers the three things that trip people up the most: hitting an exact width and height, scaling by percentage, and keeping the aspect ratio locked so the image never distorts. It also covers the one rule that quietly governs all of it — downscaling keeps quality, upscaling cannot invent detail.

Exact dimensions vs. percentage: pick the mode first

There are two honest ways to ask for a new size, and they answer different questions.

By percentage answers "make this smaller (or bigger) by a ratio I trust." Set it to 50% and a 4000×3000 photo becomes 2000×1500. The aspect ratio is preserved for free because both dimensions scale by the same factor. This is the mode you want when you do not care about a specific target number — you just want the file lighter, or you want a thumbnail that is "about half size."

By exact width or height answers "I have a hard target." Most upload slots are like this: an avatar that must be 200×200, a WordPress featured image that wants 1200×675, a WeChat cover that is rigidly 900×383. Here you type the number you were given and let the tool do the arithmetic for the other side.

The Image Resizer splits this into separate modes precisely so you do not have to think about it as one ambiguous box. Resize by width, by height, or by percent and the aspect ratio is held automatically. Choose Custom width × height and you decide whether both numbers are obeyed exactly.

Lock the aspect ratio so the image does not stretch

This is the single most common way to ruin a resize: entering a width and a height that do not match the original proportions, with the lock off.

When the aspect-ratio lock is on, you give one dimension and the other is computed to keep the shape. A 4000×3000 photo (a 4:3 shape) asked for 1200 wide becomes 1200×900 — still 4:3, nothing distorted. When the lock is off, both numbers are obeyed literally. Ask that same 4:3 photo for 1200×675 and it gets squashed vertically to fit a 16:9 box. Faces go long, circles become ovals, a square logo turns into a rectangle.

So the rule is simple: keep the lock on unless you are filling a pixel-exact slot that genuinely needs both dimensions. Ad units, email-template hero images, and some CMS featured-image slots really do demand exact pixels and will crop or break the layout otherwise — those are the legitimate cases for turning the lock off. For social posts, blog images, and almost everything else, leave it on and you will never accidentally stretch anything.

Downscaling keeps quality; upscaling cannot add what isn't there

Here is the rule that explains half the bad outcomes people blame on "low quality" resizers.

Downscaling — going smaller — is the safe direction. You start with more pixels than you need and the resampler averages them down. A good resampler (high-quality smoothing, which a serious tool turns on by default) produces a clean, sharp result even at large reductions. Shrinking a 4000-pixel photo to 1200 throws away pixels you were never going to see anyway.

Upscaling — going bigger — is where physics stops you. A resizer can only redistribute the pixels that already exist; it cannot manufacture detail that was never captured. Blow a 200×200 logo up to 1920×1920 and you get a soft, blocky mess, because there simply is no extra information to draw on. No in-browser resizer can fix this. If you genuinely need a small image at a large size, that is a job for an AI super-resolution model, which hallucinates plausible detail — a different tool, a different trade-off.

The practical takeaway: resize down from the largest original you have, never from a small copy. And avoid the re-encode trap — resizing a JPEG, saving it, then resizing that result again stacks compression artifacts every pass. Always start from the original file.

A worked example: a 4000px photo down to 1200px wide

Let's do the most common real resize end to end.

You have a 4000×3000 phone photo (4:3, around 4 MB) and you want it 1200 pixels wide for a blog post.

Drop the file in and choose Resize by width.
Type 1200 in the width field. Because by-width mode keeps the ratio, the height fills in as 900 automatically — 1200×900 is still 4:3, so nothing stretches.
Pick an output format. WebP at quality 82 is my default for web: it lands around 80–150 KB, roughly 25–35% smaller than the equivalent JPEG, with no visible artifacts.
Check the per-row readout. It shows the original dimensions (4000×3000), the new dimensions (1200×900), and the final byte count before you commit.
Download.

That is the whole operation. The numbers add up because you let one mode hold the ratio instead of typing two values and hoping they matched.

When I am batching covers for several posts in one sitting, I drop all of them in at once, set by-width 1200, and export the batch as a single zip. The first time I did this I caught myself about to type both width and height by hand for each file — the by-width mode quietly saved me from a dozen chances to fat-finger a wrong height and stretch every image. Now it is muscle memory: one dimension, lock on, done.

A note on privacy: resizing happens on your own device

If the photo is personal — a passport scan, a kid's picture, a medical document — you should care where the bytes go. With the Image Resizer they do not go anywhere. Every file is read, decoded into an image, repainted at the new size onto an off-screen canvas, and re-encoded entirely inside your browser tab. Open the DevTools Network panel while you work and the request count stays at zero. No account, no quota, no analytics on the image bytes.

There is a useful side effect: re-encoding through the canvas strips every EXIF tag — GPS coordinates, camera model, timestamps. For anything you are about to post publicly that is exactly what you want, since strangers should not be able to read the exact spot a photo was taken straight out of the file. (If you are archiving and need EXIF preserved, do not run the file through any canvas-based resizer.)

One thing the tool cannot scrub for you is the filename. The download keeps your original name with the new dimensions appended, so if the source was salary-2026.png, rename it before sharing. The pixels are private; the filename is whatever you uploaded.

When you want smaller files, not smaller dimensions

Resizing and compression are different jobs. Resizing changes the pixel grid — fewer pixels, smaller picture. Compression keeps the same dimensions and squeezes the file by tuning quality. If your photo is already the right size but too heavy for an upload cap (the classic "5 MB resume photo, 2 MB limit"), you do not want to drop resolution at all — you want to trim quality. For that, reach for an image compressor and step the quality down a notch at a time until you cross under the cap, watching the byte count as you go.

Resize when you have a dimension target. Compress when you have a file-size target. Most of the time you reach for one or the other, and knowing which is half the battle.

Made by Toolora · Updated 2026-06-13