Image Compressor

How Image Compression Works

Lossy and lossless: There are two families of image compression. Lossless methods, used by PNG, rearrange data so it takes less room but can be reconstructed exactly, every pixel comes back identical. Lossy methods, used by JPEG, WebP and AVIF, permanently discard information the human eye is poor at noticing, which is why they achieve far smaller files. This tool uses lossy encoding for JPEG, WebP and AVIF, and lossless for PNG.

What actually gets thrown away: Human vision is much more sensitive to brightness than to colour, and much better at seeing large shapes than fine texture. Lossy encoders exploit both facts: they store colour at reduced resolution and round off high-frequency detail. On a photograph of a face or a landscape this is nearly invisible. On a screenshot full of one-pixel-wide text strokes it is very visible indeed, which is why screenshots and diagrams behave so differently from photos.

Why the same setting gives different results: Compression works by finding redundancy. A photograph of a clear blue sky compresses enormously because most of it is similar. A photograph of gravel, foliage or confetti compresses poorly because almost every pixel differs from its neighbour. Two files of identical dimensions and identical quality settings can easily differ tenfold in size.

Compression is not reversible: Once detail is discarded it is gone. Saving a compressed file again at a higher quality does not recover it, it simply stores the already degraded image in a larger file. Always compress from your original.

How to Compress an Image

Add your files: Drag images onto the drop area, click it to browse, or press Ctrl+V to paste directly from the clipboard. Pasting is the quickest route for screenshots, because the image never has to be saved to disk first. You can add as many files as you like and mix formats freely.

Set the three controls: Quality decides how much detail is discarded, the resize option caps the longest edge in pixels, and the format selector decides what the output is encoded as. Leaving format on Keep original means a JPEG stays a JPEG and a PNG stays a PNG.

Compress and review: Press Compress and each image is processed in turn. Every row shows the old size, the new size and the percentage saved, so you can judge whether the trade was worth it before you keep anything. If a result looks too soft, lower the compression by raising the quality and run it again, the originals are still loaded.

Download: Save any single image with the button on its row, or use Download ZIP to get everything in one archive. Copy Summary puts a plain-text before-and-after report on your clipboard, which is handy when you need to show someone what changed.

Choosing the Right Quality Setting

The useful range is narrower than the slider suggests. Values above roughly 90 add file size for detail almost nobody can see. Values below about 50 start producing visible artefacts: blocky patches in smooth gradients, and a halo of noise around sharp edges and lettering. Most real work happens between 65 and 85.

Recommended starting points: Photographs for a website, 75 to 85. Photographs for email or messaging, 65 to 75. Hero and background images that are heavily overlaid with text, 60 to 70, since detail is obscured anyway. Screenshots containing readable text, 90 or higher, or better still, keep them as PNG.

Judge at the size it will be displayed. The most common mistake is inspecting a compressed photo at 400 percent zoom, seeing artefacts and panicking. Nobody will view it that way. Compare at the size it will actually appear on the page.

Why Resizing Beats Quality

Pixel count drives file size more than any other factor. File size scales roughly with the number of pixels, and pixel count scales with the square of the dimensions. Halving the longest edge quarters the pixels, typically cutting the file by seventy percent or more, with no visible loss at all provided the image is still larger than the space it is displayed in.

Most images are far larger than they need to be. A modern phone produces photographs around 4000 pixels wide. A blog article column is usually 700 to 900 pixels wide, and a full-width hero image rarely needs more than 1920. Uploading the original means sending perhaps twenty times more data than any visitor will ever see.

Allow headroom for high-density screens. Retina and similar displays render roughly two physical pixels per CSS pixel, so an image shown in an 800-pixel column looks crisper at 1600 pixels wide. Doubling the display width is a sensible ceiling; going beyond that is waste.

Resize first, then compress. Doing both at once, as this tool does, gives a better result than compressing hard at full resolution. A large image at quality 80 will almost always look better and weigh less than a full-resolution image at quality 40.

JPG vs PNG vs WebP vs AVIF

JPEG is the oldest and most universally supported. It handles photographs well, has no transparency, and shows its age against newer codecs. Choose it when something downstream, an old system, a print workflow, a picky upload form, insists on it.

PNG is lossless and supports transparency. It is the right choice for screenshots, logos, diagrams, icons and anything with sharp edges or text. It is a poor choice for photographs, where it routinely produces files several times larger than JPEG for no visible benefit.

WebP is the pragmatic default today. It typically produces files 25 to 35 percent smaller than JPEG at comparable quality, supports transparency, and is displayed by every current browser. For most web work it is simply a better JPEG.

AVIF compresses harder still, often 20 to 30 percent below WebP, and handles gradients and dark scenes particularly well. Support is now broad but encoding is slower, and not every browser can create AVIF even when it can display it. If your browser cannot encode it, this tool falls back to WebP rather than failing.

Privacy, EXIF and GPS Data

Your files never leave this tab. There is no upload, no queue and no temporary copy on our side, because the work is done by JavaScript running on your own machine. You can confirm this by disconnecting from the network after the page has loaded, the tool keeps working.

Metadata is removed automatically. Photographs carry an EXIF block describing the camera and lens, the exact moment of capture and, on most phones, the GPS coordinates of where you were standing. Re-encoding through a canvas reproduces the pixels alone, so all of that is discarded. Sharing a compressed photo does not disclose where it was taken.

The trade-off worth knowing: If you actually want the metadata, for a photography portfolio, a catalogue or a legal record, compressing here will strip it. Keep an untouched original in that case.

Images and Page Speed

Images are usually the heaviest thing on a page. On a typical site they account for more transferred bytes than scripts, styles and fonts combined. Compressing them is normally the single most effective performance change available, and it requires no code.

Largest Contentful Paint is often an image. Google's LCP metric measures when the biggest visible element finishes rendering, and on most pages that element is the hero image. Shrinking it moves the metric directly, and LCP feeds into search ranking through Core Web Vitals.

Mobile data is not free. A visitor on a slow connection pays for every kilobyte in both time and money. Cutting a 4 MB gallery to 400 KB is the difference between a page that loads and a page that gets abandoned.

Common Mistakes to Avoid

Compressing an already compressed file. Each lossy pass discards more detail, and the losses accumulate. Editing and re-saving the same JPEG repeatedly visibly degrades it. Always start from the original.

Using JPEG for screenshots. Text and sharp edges are exactly what lossy encoders handle worst, you get a fuzzy halo around every letter. Screenshots belong in PNG, or WebP if you need the size.

Using PNG for photographs. The mirror image of the same error. A lossless format cannot exploit the redundancy in a photograph, so you end up with a file several times larger than necessary.

Chasing an arbitrary number. Compressing to hit exactly 100 KB regardless of content is how images end up looking bad. Set a quality you are happy with and accept the size that results.

Practical Size Targets

Website hero image: 1920 pixels wide, WebP at quality 80, aiming for 150 to 300 KB. Above 500 KB you are hurting load time for detail nobody notices.

Article or blog image: 1200 to 1600 pixels wide, WebP at 75 to 80, typically 80 to 200 KB.

Product thumbnail: 400 to 600 pixels, WebP at 75, usually well under 50 KB. With dozens on a listing page, per-image savings multiply fast.

Email attachment: 1280 pixels, JPEG at 70. Many mail servers still reject messages over 20 to 25 MB, and recipients on phones will thank you.

Screenshot for a bug report: Original size, PNG. Legibility matters more than bytes here, and a resized screenshot of an error message is useless.

Frequently Asked Questions