Image Compressor

Compress an image and see what each quality setting actually costs, side by side, before you commit to one. Most compressors make you guess a number and resubmit; this one shows you the whole curve.

Metadata is stripped in the process, which usually saves a few kilobytes on its own and removes the GPS coordinates phone cameras embed.

How it works

Lossy compression works by discarding information the eye is bad at noticing. JPEG and WebP both split the image into blocks, convert each to frequencies, and round away the high-frequency detail — fine texture and sharp edges — while keeping the broad shapes. The quality number controls how aggressively that rounding happens.

The relationship between quality and file size is not linear, and that is the single most useful thing to understand here. Going from 95 to 85 typically saves half the bytes for a difference almost nobody can see. Going from 85 to 75 saves much less and starts to show on flat areas like skies. Below about 60, blocking artefacts appear around edges and text becomes mushy. The sweet spot for web photographs is usually 75–85, which is why the table below brackets it.

PNG is a different animal. It is lossless, so there is no quality dial at all — its size comes from how well the pixel data predicts itself. Re-encoding a PNG saves whatever metadata it carried and not much else. If a PNG is too big, the real answers are to reduce its dimensions, cut the colour depth if it is a simple graphic, or convert it to WebP, which has a lossless mode that beats PNG outright.

Which brings up the thing compression cannot fix. If a 4000-pixel photograph is displayed in an 800-pixel slot, no quality setting solves that — you are sending twenty-five times the pixels needed and then throwing them away in the browser. Resize first, compress second. This page flags it when the image looks oversized for any plausible display size.

A caution on repeat compression: every lossy pass discards more, and it never comes back. Compressing an already-compressed JPEG rounds away detail that was already an approximation, which is how images acquire that over-processed look after being saved and re-saved a few times. Always work from the largest, least compressed original you still have.

Frequently asked questions