2026-08-04

How to compress images without losing quality

The real relationship between file size and visible quality — how to hit an exact target file size, when lossless actually matters, and why resizing first helps.

Compressing an image is one decision made four ways, and the difference between them is the difference between a page that loads and a page that does not. This is the whole tool explained, with the numbers we measured rather than the ones we assumed — including the cases where compressing does nothing at all, and why.

Open the Compress Image tool — free, no sign-up, no watermark.

Compress, resize, convert: three different things

Compression changes how the existing pixels are encoded. It leaves the dimensions alone, which is what separates it from resizing, where the pixel count itself comes down, and from cropping, which throws away part of the frame. A compressed photo is the same photo at the same size — it simply weighs less, and the download screen reports the same width and height it went in with.

The saving comes from discarding detail your eye was never going to register: fine gradations of colour, texture in shadow, differences between neighbouring pixels that no screen resolves. How much you can discard before it starts to show is the whole question, and it is what the four levels are for.

One thing this tool deliberately does not do is change the file type. Your compressed image comes back in the format it arrived as — a JPEG stays a JPEG, a PNG stays a PNG. Converting is often the larger saving, but it also changes what the file can do, so it is a decision worth making on its own in Convert Image rather than applied quietly here.

Step 1: add your image

Drag a file in, or click to browse. The tool reads it off disk, measures it, and reports which of those is happening rather than running on a timer.

The progress screen: a spinner between an illustration of a laptop and a browser window, reading Compressing Your Image, with the file name, dimensions and size on a card beneath. The same screen appears again after you press the button, so the wait between the press and the result is never a dead editor.

Accepted formats are JPG, PNG, WebP, AVIF, GIF and TIFF, up to 25 MB. iPhone HEIC files are refused — set Settings → Camera → Formats to Most Compatible, or convert the photo first.

Step 2: compress options

There are only four things to decide, and the screen opens with sensible answers to all of them: Medium level, no size ceiling, lossy, metadata stripped. Change a setting and the estimate underneath updates before you commit to anything.

The compress options screen: the image preview on the left with its measurements and file facts, and on the right the compression level tiles, target file size row, lossy and lossless cards, estimated result panel, more options and the Compress Image button. Everything fits on one screen, so the button is reachable without scrolling on a normal laptop.

Compression level

Four presets that set the encoder quality for you: Low 90, Medium 75, High 55 and Maximum 35.

The four compression level tiles: Low bigger size best quality, Medium selected balanced size good quality, High smaller size good quality, and Maximum smallest size standard quality. Medium is selected on arrival, so pressing the button without touching anything gives you the balanced setting rather than an unstated one.

On a 3000×2000 test image saved as JPEG at quality 92, weighing 1.27 MB, the four levels returned:

Low901.08 MB15% smaller
Medium75310 KB76% smaller
High55126 KB90% smaller
Maximum3555 KB96% smaller

Your own numbers will differ, and the reason matters: how much a picture gives up depends almost entirely on how hard it was squeezed before it reached us. High is where artefacts begin showing on skin and smooth sky while staying invisible on busy detail. Maximum is for thumbnails and previews, where the picture is too small for anyone to look closely.

Low is the one to be careful with. On a file that has already been compressed once, asking for quality 90 usually costs more bytes than it saves.

Target file size

A ceiling instead of a quality guess. The server encodes the picture several times over, narrowing in on the quality setting, and returns the largest result that still comes in beneath the number you named.

The target file size row: Auto selected, then 50 KB, 100 KB, 200 KB, 500 KB and Custom. Auto is the default and means no ceiling — the level you picked is used directly.

On the same test image it landed at 493 KB, 193 KB, 95 KB and 48 KB against requests of 500, 200, 100 and 50 KB, choosing quality 83, 67, 51 and 32 respectively. This is the right control whenever a form states a limit, because it removes the guesswork entirely.

Two things to know about where it runs out of room. It searches only formats with a real quality dial — JPEG, WebP, AVIF and TIFF — so a PNG or GIF paired with a ceiling gets the level you picked and nothing more. And no search can go below what quality 1 produces, so on a very large photograph the floor may sit above a tight target. The answer then is fewer pixels rather than less quality, which means running it through Resize Image first.

Custom selected, revealing a number box reading 300 and a unit dropdown set to KB. Choosing Custom reveals a number box and a KB / MB switch, so a limit stated in megabytes needs no conversion in your head.

Compression type

Lossy discards detail to save bytes. Lossless keeps the pixels exactly as they arrived and saves only what can be saved without altering them.

The lossy and lossless cards, with lossy selected.

Lossy for anything photographic. Lossless for a logo, a screenshot, a chart or a diagram, where a soft edge or a smeared line is obvious in a way it never is on a photograph.

It is not a small difference in weight. On a 1000×1000 crop, WebP went from 54 KB lossy to 967 KB lossless, and PNG from 296 KB to 1.74 MB. We checked that the promise holds rather than assuming it — decoding the output and comparing it against the source sample by sample, lossless WebP, PNG and AVIF all came back identical, every channel of every pixel.

A size ceiling and lossless cannot both apply, since lossless leaves nothing to trade; choosing it greys the target row out and says so.

The lossless card greyed out, reading not available for JPEG, convert to WebP, PNG or AVIF first. JPEG, GIF and TIFF have no lossless mode. Since the output format follows the input, uploading one of those leaves the card disabled with a line explaining why.

Estimated result

A projection of what you are about to get, updating as you change the level or the ceiling.

The estimated result panel showing original size, an arrow, the estimated size and a green percentage smaller chip.

It is arithmetic, not a trial encode — running the real thing on every click would mean a round trip to the server each time — so treat it as the shape of the answer rather than the answer. The figure on the download screen afterwards is the real one.

It is deliberately cautious about files that have already been compressed. Re-saving at the same format cannot repeat a saving that has already been taken, and the estimate accounts for that instead of promising a reduction the encoder cannot deliver.

When there is nothing left to save

The estimate panel showing no reduction at this level, with a note suggesting High or Maximum or a target file size. If the chosen level would not actually make the file smaller, the panel says so in words instead of showing a cheerful percentage.

This is the normal state of affairs for a file that has been compressed once already, and it is worth understanding rather than working around. Taking our test image down to quality 55 and then feeding it back in:

Low48% largeryour original, unchanged
Medium21% largeryour original, unchanged
Highidenticalyour original, unchanged
Maximum26% smallerthe compressed file

The server will not hand back a file bigger than the one you gave it — if the re-encode comes out heavier, you get your original bytes back. But a level that produces no saving is still a wasted press, and the panel is there to spare you it.

More options

The more options checkbox: remove EXIF metadata, ticked.

Remove EXIF metadata strips the camera, date and location tags your photo arrived with. It is ticked by default, which matches what the file actually gets. Leave it ticked unless you specifically need that information to travel with the picture — and certainly leave it ticked for anything going somewhere public, because a photograph straight off a phone carries the coordinates it was taken at. Losing the block also trims the file a little: roughly 800 bytes on a test shot carrying make, model and four location tags.

Step 3: download

The download screen: a large download button showing the dimensions, format and file size, then Copy Image, Change Settings and Compress Another, then a row of tiles for the other image tools. The row underneath hands the compressed image straight to another tool without re-uploading it.

Change Settings goes back to the options with the same picture still loaded, so trying a harder level costs nothing. Compress Another clears it for a new file. The tiles below pass the finished image to Resize, Crop, Convert, Upscale or Watermark.

Where the honest edges are

Compression cannot make a picture better. Every lossy setting throws something away, and a file already squeezed once has less left to give, so repeatedly compressing the same image degrades it a step at a time. Take each new version from the original instead.

The output format always matches the input, so this tool cannot turn a JPEG into a WebP — use Convert Image, and expect a larger saving from the format change than from the quality setting.

PNG responds to the quality setting very unevenly, because what it spends its bytes on is the number of colours it keeps rather than a fidelity dial; on some images the level buttons barely separate at all.

Dimensions are never changed here. An animated GIF is flattened to one frame. Pictures go through one at a time. Uploads are capped at 25 MB. Compression happens on a server, so the file does travel — it is processed and returned straight away, never written to disk and never shared.

Common questions

How do I compress an image without losing quality?

For a photograph, aim for invisible loss rather than none: Medium is where most people stop being able to tell. If you truly need every pixel intact, and your file is a WebP, PNG or AVIF, switch Compression type to Lossless — we decoded the output and compared it against the source sample by sample, and it came back identical. It costs a lot of weight, though: a 54 KB lossy WebP became 967 KB lossless on our test crop. A JPEG cannot be compressed losslessly, because the format has no lossless mode.

How much smaller will my image get?

It depends far more on how compressed the file already is than on the picture. On our 1.27 MB test image the four levels returned 1.08 MB, 310 KB, 126 KB and 55 KB. Feed a file that has already been squeezed hard back in and the numbers collapse — the same image taken down to quality 55 first gave nothing back at Low, Medium or High, and only 26% at Maximum. The estimated result panel tells you which situation you are in before you press anything.

How do I compress an image to a specific size like 100 KB?

Use Target file size rather than guessing at a level. Pick 50, 100, 200 or 500 KB, or choose Custom and type your own figure in KB or MB. On our test image a 100 KB request came back at 95 KB, at quality 51. Whether it can reach a tight number depends on the pixel count, so when the floor sits above your target, reduce the dimensions first with Resize Image and then compress.

What is the difference between lossy and lossless compression?

Lossy compression permanently discards detail it judges you will not miss, which is how a photograph can lose four-fifths of its weight and still look right. Lossless compression only reorganises the data, so the picture decodes back to exactly the same pixels — nothing is discarded, and the saving is far smaller or even negative. Photographs suit lossy; logos, screenshots, charts and line art suit lossless.

Can I change the file format while compressing?

Not here — the compressed file comes back as whatever it arrived as. Convert Image handles format changes, and the finished screen hands your compressed image straight to it if you want both.

Does compressing an image remove EXIF and GPS data?

Yes, by default. A test shot that arrived with its camera make, its model and four location tags came back holding none of them. Removing them also trims the file a little, about 800 bytes on that test. Untick the box if you need the camera settings or the date to travel with the picture.

Does compressing reduce the image dimensions?

No. The picture comes back at exactly the width and height it went in at, and the download screen shows those numbers. If you want smaller dimensions as well, use Resize Image — that is usually the bigger saving of the two when an image is far larger than the space it is displayed in.

Why did my image not get any smaller?

Almost always because it had already been compressed before you uploaded it, and the easy savings were taken the first time. Asking for a gentle level on such a file can even cost bytes rather than save them. Two things protect you: the estimate panel says No reduction at this level before you press, and the server refuses to hand back a file bigger than the one you gave it. The fix is a harder level, a target file size, or fewer pixels.

Does it work on a phone?

Yes. It runs in any modern browser on Android and iOS with nothing to install, and you can compress straight from your camera roll.

Can I compress several images at once?

Not in one pass — the tool takes one image at a time. Change Settings on the download screen sends you back to the options with the same picture loaded, so trying several levels on one image does not mean re-uploading it.

Try the Compress Image tool, or browse the rest of the Image Toolkit.