Compress Image

Pick one image, set a quality and an output format, and get a smaller copy of the same picture — with the option to delete the GPS location first. Everything runs in your browser, so the photo is never uploaded.

Drop files here or choose files
Not uploaded · compressed in your browser · one image up to 50 MB

Lower numbers mean a smaller file and more visible quality loss; at the default 75 the difference is usually hard to see.

Output format

For the same photo, WebP is usually smaller than JPG; PNG suits screenshots and line art.

Photo info (EXIF)

Keep everything: the capture time, camera settings and location are written into the compressed image.

How to use

  1. Choose one image: Drop a JPG, PNG, WebP or AVIF file onto the box or click to pick it. One image per run, up to 50 MB, and it stays on your device.
  2. Set the compression strength: Drag the quality slider between 10 and 95. It starts at 75, which suits most photos; lower it if the result is still too heavy.
  3. Pick an output format and decide about EXIF: Leave Output format on Keep original format, or switch to JPG, PNG or WebP. Under Photo info (EXIF), choose Keep everything, Remove location or Remove all.
  4. Compress and save: Click Compress image. You get the size before and after, a line about the metadata in the result, and a preview — then download the file, which is named like photo-compressed.jpg.

How to compress an image in your browser

Compress one image up to 50 MB in your browser without uploading it. Set JPG, WebP or AVIF quality from 10 to 95, with 75 as the default; PNG output is unaffected by the slider. The result may be smaller, unchanged or larger, depending on the settings.

Choose a file, drag the quality slider, optionally switch the output format, click Compress image and download the result. The picture is never resized here, so a 1200 × 800 photo comes back on the same 1200 × 800 grid of pixels — the one exception, the orientation Remove all writes into the picture, is in the limits below.

Nothing is uploaded. The compression engine is libvips compiled to WebAssembly (wasm-vips), downloaded with the page and run by your own browser, and the EXIF reader (exifr) works the same way. The browser isolation that engine needs also blocks third-party scripts, which is why this page carries no ads and no analytics. Close the tab and nothing is left behind, so it is fine for ID photos, screenshots of statements and anything else you would rather not hand to a server.

The other limits are one image per run and about 100 megapixels (roughly 10000 × 10000). Anything libvips can decode will open — JPG, PNG, WebP, AVIF, GIF, TIFF, SVG — and the output format buttons are JPG, PNG, WebP and Keep original format, which is the only way a result comes out as AVIF (an AVIF file left on that setting).

Which quality to choose — and why 95 can make a photo bigger

Test file: photo.jpg, 529 KB, 1200 × 800, already a JPEG
quality 95 → 873 KB (65% BIGGER — original handed back)
quality 90 → 648 KB (23% bigger)
quality 75 → 297 KB (44% smaller)  ← default
quality 50 → 198 KB (63% smaller)
quality 20 → 112 KB (79% smaller)

Start at the default 75. On the test photo above — a 529 KB JPEG that had already been through a JPEG encoder — quality 75 produced a 297 KB file, 44% smaller. Quality 50 brought it to 198 KB and quality 20 to 112 KB. Whether the loss shows depends on the picture, so compare the preview with the original before you download; at low settings, skies, skin and flat color are where JPEG's block edges turn up first.

Pushing the slider up can backfire on a file that was already compressed. The same photo re-encoded at quality 95 came out at 873 KB, 65% bigger than the original, and at quality 90 it was 648 KB. For that JPEG the explanation is straightforward: detail was already discarded when it was first saved, and writing that same degraded picture again with a more generous setting just spends more bytes on it. A high setting recovers nothing — it only costs space.

Note what the page can and cannot tell you afterwards. It reads the sizes, not the quality your file was saved with, so when a same-format result grows it reports both sizes and names the things that can be behind it — the encoder settings, the encoder itself, the metadata written back — rather than declaring one of them the cause.

What it does not do is hand you a bigger file: when the re-encode is not smaller, your original bytes are what you download, under the original file name, and the page tells you why.

That fallback needs three things at once: the input arrives in a format the page can identify from its header (JPG, PNG, WebP or AVIF), the output format matches that one, and Photo info (EXIF) is on Keep everything. Asking for the location or the metadata to go switches it off, because the original is precisely the copy that still contains it — and a GIF or a TIFF never qualifies, since those come back as a JPG, which is a conversion rather than a re-encode.

How far a photo falls is specific to the photo: fine detail, texture and noise cost more bytes than smooth gradients, so treat the table as one measured sample rather than a promise. One habit worth keeping, though, applies to every JPEG — compress from the best copy you have, and compress once. Each round of saving quantizes the picture again, so a file squeezed three times at quality 75 usually looks worse than the same file squeezed once.

Keep the original format, or convert to JPG, PNG or WebP

Same photo.jpg (529 KB) at quality 75:
  JPG   297 KB  (44% smaller)
  WebP  237 KB  (55% smaller)
  PNG   401 KB  (24% smaller)

Test photo.png (210 KB, 800 × 600):
  PNG   177 KB  (15% smaller — identical at any quality)
  JPG   156 KB  (26% smaller, quality 75)
  WebP  127 KB  (39% smaller, quality 75)

Keep original format writes the result in the format the file header says it already is. If that header is not JPEG, PNG, WebP or AVIF, the result is a JPG — which is what happens to GIF and TIFF files unless you choose a format yourself, and the page reports it as the conversion it is.

Converting is often the bigger win. At the same quality 75, the test photo came out at 297 KB as JPG and 237 KB as WebP, so WebP was 20% lighter than JPG on the same file. These figures describe this test image, not every photo. WebP is the right choice for a website or a chat app, while JPG is the safer attachment when something old has to open it, or a print shop asks for one.

The quality slider does nothing when the result is a PNG, and the page says so while you drag it. We wrote the same PNG at quality 20 and at quality 95 and got byte-identical files, 181,714 bytes both times: this build has no palette quantizer, so a PNG result is simply rewritten with maximum lossless compression and lands on one fixed size. The 210 KB test PNG therefore came back at 177 KB whatever the slider said — but at 156 KB as a JPG and 127 KB as a WebP. If a PNG is still too heavy, change the format, not the slider.

Keep PNG when you need crisp text, flat color or transparency, which is the usual case for screenshots and line art. And remember JPG has no transparency at all: a transparent image converted to JPG gets a white background painted in first. PNG is the wrong choice for a photograph, though: it was the weakest of the three on the test photo at 401 KB, and because it cannot discard detail the way JPEG does, converting a photo that was saved at a modest quality to begin with can land you above the original.

Whenever a result does come out larger than the input, the page gives you both sizes and then suggests what is still worth trying from where you are — a lower quality, a different output format, or smaller dimensions in a resizing tool. It won't send you back through the run you just made: a GIF that came out as a JPG, for instance, is pointed at a lower quality or at WebP, not at Keep original format, which would write that same JPG again.

Strip the GPS location before you publish a photo

A photo straight off a phone can carry the moment it was taken, the camera model and the exact coordinates of where you stood. This tool reads those three out of the file as soon as you pick it and shows them under Photo info (EXIF) — before compressing, so you find out while you can still act on it. For JPG, PNG and WebP results it then reads the finished file back and reports what actually survived, instead of repeating what you asked for. An AVIF result — which only happens when an AVIF file is left on Keep original format — cannot be read back in the browser, so the page says it cannot confirm what is left in it.

Keep everything is the default and writes the capture time, camera and exposure settings, GPS position, orientation, XMP and the color profile into the compressed image. Remove location deletes the whole GPS block, plus the XMP block if a location was written there, and leaves the camera model, capture time, exposure settings and orientation in place. Remove all keeps only the color profile — so a wide-gamut photo does not go flat — and writes the orientation into the picture first, which is why a portrait shot cannot come back sideways. That rewrite is also the one case where the numbers change: a picture a camera stored sideways with a rotation tag comes out with its width and height swapped, which is exactly what makes it stand upright.

Metadata has weight of its own. A 2,322-byte test photo carrying GPS, XMP and an ICC profile came out at 2,321 bytes with everything kept, 1,902 bytes with the location removed and 1,462 bytes with everything removed, all at quality 75. On a small image that is a real share of the file; on a 5 MB photo it is noise.

Use Remove location for anything going onto a public profile, a marketplace listing or a rental ad, where a home address in the file is worth more to a stranger than it is to you. Use Remove all to remove the EXIF, XMP and other metadata while retaining the color profile — that clears what is written around the picture, not what is visible in it, so a house number, a face, a name badge or a screenshot of a document is still there and still has to be cropped or covered yourself. Keep everything is for photos going into your own archive, where the date and the camera data are the point.

What this tool cannot do

One image per run — there is no batch mode, so a folder of photos has to go through one by one. It also cannot make a picture sharper or larger: compression only discards information, and nothing here resizes the image. The only change to the pixel grid is the rotation Remove all bakes in: that option writes the EXIF orientation into the picture before deleting the tag, so a file a camera marked as rotated comes back with its width and height swapped (a tagged 64 × 48 test image came out 48 × 64, standing the right way up). An image that is already blurry or too small stays that way.

HEIC photos from an iPhone cannot be opened here; decoding them needs a library this build does not include, and choosing one gets you the "isn't an image our tools can read" message. Export or share the photo as a JPG first — an iPhone normally converts it for you when you attach a photo to an email.

Animation does not survive a re-encode. The file this tool writes holds the first frame only, and there is no setting that compresses an animation and keeps it moving: a 381-byte, 50 × 50 animated GIF converted to WebP came back as a 310-byte still image. A GIF left on Keep original format also comes back as a JPG, because the format cannot be identified from the header. There is one way an animation reaches you intact, and it is the safety net above rather than a feature: an animated WebP left on Keep original format and Keep everything whose single frame does not come out smaller is handed back as your original file, animation and all.

There is no field for an exact target size here — the slider is the only control, so hitting "under 200 KB" means trying a value, reading the result and trying again from the original. That is not guaranteed to get there: quality 10 is as low as the slider goes, and since the picture is never resized, a big photo can stay above a tight limit. Reducing the dimensions first, with a resizing tool, is what fixes that. Resizing, watermarking, password-protecting and converting to PDF are all other tools.

Need to shrink a scanned PDF instead? Use Compress PDF.

Frequently asked questions

How much smaller will my image get?
It depends on the photo and the quality you pick. On our 529 KB, 1200 × 800 test JPEG, quality 75 gave 44% smaller, quality 50 gave 63% and quality 20 gave 79%. A 210 KB test PNG only dropped 15% when kept as a PNG, but 39% converted to WebP. Detailed, noisy pictures usually shrink less than smooth ones.
How can I compress an image under 200 KB?
You can try manually, but this page cannot guarantee a result under 200 KB. Choose JPG or WebP, start at quality 75, then click Start over and retry from the original at a lower quality. The minimum is 10; if the file is still too large, reduce its pixel dimensions elsewhere.
Why did my image get bigger instead of smaller?
The size line tells you which of two things happened: the file was converted to another format, or it was re-encoded in the same one. A conversion can simply cost more for that picture, which is the usual story when a photo becomes a PNG. A same-format re-encode that grows may involve different encoder settings, the encoder itself or metadata — our test JPEG grew 65% at quality 95 — and the page cannot read the setting your file was saved with, so it reports both sizes and leaves the cause open. Either way it suggests the next thing to try: click Start over, pick the image again, and lower the quality or switch the output format. For a JPG, PNG, WebP or AVIF input identified from its header, Keep original format plus Keep everything returns the original if re-encoding is not smaller. GIF and TIFF inputs become JPG and do not qualify for this fallback.
Can I compress a JPEG without losing quality?
Not literally — JPEG is lossy, so every re-encode throws something away. Quality 75 is the default starting point rather than a promise that the difference is invisible: our 529 KB test photo dropped 44% there, and the preview is what tells you whether that is acceptable for your picture. Compress from your best original and only once, since each extra round quantizes the picture again.
Why doesn't the quality slider change my PNG?
PNG compression is lossless, so there is no quality dial to turn. We measured it: the same PNG written at quality 20 and at quality 95 produced byte-identical 181,714-byte files. A PNG result is always rewritten with maximum compression and one fixed size. To make it smaller, switch the output format to JPG or WebP.
Are my photos uploaded to a server?
No. The compression engine (libvips as WebAssembly) and the EXIF reader run inside your browser, and no image bytes are sent anywhere. The page carries no ads or analytics either, because the browser isolation the engine requires blocks third-party scripts. Close the tab and nothing remains.
Does compressing remove the GPS location from my photo?
Only if you ask it to. Keep everything, the default, writes the location into the compressed image. Remove location deletes the GPS block and any location in XMP while keeping the camera model and capture time; Remove all leaves just the color profile. For JPG, PNG and WebP results, the page reads back the capture time, device and GPS fields it can detect. AVIF results cannot be checked this way, so the page reports that it cannot confirm what remains.
Can I compress a HEIC photo from my iPhone?
No. HEIC decoding needs a library this build does not include, so the file is rejected as unreadable. Share or export the photo as a JPG first — an iPhone normally converts it for you when you attach a photo to an email — then compress that copy. JPG, PNG, WebP and AVIF all open fine.
Can I compress several images at once?
No, one image per run. Pick a file, compress it, then click Start over for the next one — your quality, output format and EXIF choices stay exactly as you set them, so each repeat is just drop and click. For a whole folder in one go you need a desktop or command-line batch tool.