Compress an image

Reduce the file size of photographs, screenshots, and website assets without sending images to a server. Choose AVIF for the smallest file, usually half the size of WebP for the same look but a few seconds to encode, WebP for fast and broadly supported web delivery, JPG where only JPG is accepted, or PNG when the file has to stay PNG. Smaller, balanced, and higher-quality presets make the tradeoff understandable without exposing complicated encoder settings.

The presets keep the original pixel dimensions while re-encoding image data. For a form or a listing with a limit, choose Under a size you set and enter the number of KB: JPG and WebP get the highest quality that fits, and PNG the largest palette that fits. Several images compress in one go and download as one ZIP. The preview and exact output size appear before download, making it easy to decide whether the result suits your website, email, listing, or document. PNG transparency cannot be preserved in JPG output; use WebP or PNG when transparency matters.

Decoding, canvas drawing, and encoding run in a dedicated browser worker. The source image stays on your device and the page remains responsive. For JPG output, choose whether to keep the photo date, camera details, and ICC colour profile; location metadata is removed from that copy. Extremely large photographs can require substantial device memory, particularly on mobile browsers. Use the resize tool first when reducing dimensions will provide a larger and more predictable saving than quality compression alone.

Format choice affects both compatibility and size. JPG is dependable for photographs and older publishing systems but has no transparency. WebP often stores a similar-looking photo in fewer bytes and can retain transparent areas, making it useful for modern websites and product assets. Logos, diagrams, and screenshots with sharp text show ringing in both at aggressive settings; PNG output suits them better.

PNG output reduces the image to a palette of up to 256 colours, with transparency and gentle dithering. Screenshots and logos often shrink by 40 to 60 percent; photographs stay far smaller as WebP or JPG, and a plain PNG is kept whenever it is smaller.

The quality control changes how much visual information the encoder discards. A lower setting can produce a much smaller file, but smooth gradients, hair, foliage, and text edges reveal damage first. Start with the balanced preset, compare the output, and move toward smaller only when the intended use tolerates it. Social thumbnails and email previews can often accept more compression than a portfolio image or marketplace product photograph.

File size alone is not the same as page performance. A 4000-pixel photograph displayed at 600 pixels still makes the browser decode unnecessary pixels even after quality compression. Resize oversized assets to their largest real display dimension, then compress the resized copy. Keep the original, because repeated lossy saves compound their damage.

Check transparent results against both light and dark backgrounds. Fine semi-transparent shadows and antialiased edges are easy to miss on a single checkerboard.

Up to twenty images can be handed over at once, and the settings above are applied to each in turn. They run one at a time, because decoding twenty photographs together is how a phone browser runs out of memory. The result is a row per file with its own download, and a file that cannot be read loses only itself: the row says why and the files after it are still processed.

Download the compressed copy and test it in the destination that matters. Confirm transparency, orientation, color, and legibility on both a large screen and the smallest expected display. Compare bytes rather than filename extensions alone, and keep the master image outside the delivery folder so a later edit does not begin from an already lossy derivative.