Image Compressor — JPG, PNG, WebP, Free & Offline

Resizing saves far more than the quality slider

A photo 4,000 pixels wide displayed in an 800-pixel column carries 25 times the pixels being used — because (4000 ÷ 800)² = 25. Width scales linearly but pixel count scales with the square of the dimensions, so halving both width and height quarters the file.

No quality setting recovers that, because you would be carefully compressing detail that will never be displayed. Check the dimensions before touching the slider.

The tool resizes to 1,920 pixels wide by default, which suits most web use. An image already narrower than that is left alone; anything wider is scaled proportionally so the aspect ratio is preserved.

Then compression, honestly

Lossy encoders discard detail the eye is least likely to notice, and the quality slider controls how aggressively. It defaults to 80.

The relationship between quality and size is strongly non-linear, which is the useful thing to know: the drop from maximum quality down to the high eighties removes a large share of the file for almost no visible change, while pushing from the middle of the range into the low numbers removes much less and damages the picture visibly. The genuinely useful range is far narrower than the slider implies — most of the benefit is available before any harm is.

What compresses well, and what does not

Photographs and gradients are exactly what lossy codecs were designed for, and they shrink dramatically at good quality settings.

Screenshots, text, line art and flat colour do not. Lossy encoding smears hard boundaries, producing visible fringing around letterforms and UI edges — and it often achieves less size reduction on that content anyway, since flat regions were already cheap to store. For that material either raise the quality substantially or choose PNG.

The three output formats

• JPEG — the default. Lossy, universally supported, right for photographs. No transparency.
• WebP — lossy here, and more efficient than JPEG at comparable quality. Supports transparency.
• PNG — lossless. The quality slider has no effect on it, because PNG does not discard detail. Choose it for screenshots, text and line art, and expect a larger file than the lossy options.

One consequence worth being clear about: selecting PNG does not restore anything. If the source was already a JPEG, its existing artefacts are baked into the pixels and PNG will store them perfectly. Lossless means no further loss, not repair.

Metadata is stripped

Re-encoding through a canvas discards EXIF, GPS coordinates, camera settings and everything else. That trims a little size and removes location data from photos automatically — welcome for anything you are publishing, unwelcome if you needed the capture data. Keep the original if you do.

Finally, re-compressing an already-compressed file compounds the damage, since the new encoder faithfully preserves the previous one's artefacts. Always start from the highest-quality original you have. To judge a result, compare the before and after sizes the tool reports, then view the output at 100% zoom on the most detailed area and on any hard edge — artefacts are invisible in a thumbnail and obvious at full size.

Compressing an image that is larger than it will ever be displayed is solving the smaller problem first — resizing saves far more than re-encoding does. The image resizer handles that.