Reduce image file sizes and compare the result before downloading. Choose a quality setting, a target file size or a maximum width. Lossy compression can change visible detail, so keep your original and use the comparison preview. Images are processed in your browser rather than uploaded to this site.
Auto uses WebP where the image has transparency and JPEG where it does not.
Select an image to compare.
This sounds like it should go without saying, and it is the single most common failure in browser-based compressors: you put in a 90 KB photo, and a 140 KB file comes back.
The reason is that re-encoding an already-compressed JPEG at high quality can genuinely produce more data than the original. The compressor is faithfully recording all the compression artefacts that were already there. You end up with a larger file that also looks slightly worse.
So this tool checks. If the result is the same size in pixels and no smaller in bytes, it discards it and gives you back the original, untouched, and tells you why. A file marked already small is not a failure — it is the correct answer.
By quality is the one to reach for most of the time. Around 80 is where most photographs stop looking any different while the file is a fraction of the size. Use the wipe handle on the preview rather than trusting the number: the right quality depends on the picture, and faces, gradients and flat colour all behave differently.
To a target file size is for when something else sets the limit — a 2 MB upload cap, a 500 KB attachment rule, a marketplace that rejects anything larger. It searches for the highest quality that still fits at full resolution, and only reduces the pixel dimensions if the target genuinely cannot be reached any other way. That order matters: a tool that shrinks the image first will hit your 500 KB target with a needlessly small picture.
To fit a maximum width is usually the biggest single win for a website, and the most overlooked. A 4,000 px photo displayed in an 800 px column is carrying twenty-four times more pixels than anyone will ever see. No amount of quality tuning fixes that — only resizing does.
JPEG for photographs. Universally supported, and still excellent at what it does.
WebP for roughly 25–35% smaller files at the same visible quality, with transparency support that JPEG does not have. Every current browser reads it. The caveat is outside the browser: some older desktop software and a few email clients still do not, so it is a better choice for your website than for a file you are emailing to a client.
PNG is for screenshots, logos, line art and anything with hard edges or transparency. It is lossless, which is why a photograph saved as PNG is enormous. If a PNG photograph is slowing your site down, converting it to JPEG or WebP is the fix.
Leave the format on Auto and the tool checks each image for transparency, sending transparent images to WebP and everything else to JPEG.
JPEG and the quality-controlled WebP output used here employ lossy compression. Some image detail can be discarded, especially at low quality settings, and raising the quality again cannot reconstruct what was removed.
WebP also supports a separate lossless mode; it is inaccurate to describe the whole format as always lossy. See Google’s WebP format documentation. This tool exposes JPEG and WebP output choices, not a dedicated lossless WebP encoder setting.
Start from the original file for each new version. Inspect fine text, gradients and edges as well as the overall picture. Keep the original if the smaller version is not acceptable for its intended use.
The compressor processes selected images in the page using your browser. It does not upload those files to this site. Very large files remain limited by your device’s memory and browser capabilities.
If the encoded version is no smaller and has the same pixel dimensions, the tool can return your original file. That preserves the original bytes instead of giving you a larger re-encoded copy.
No. Re-encoding from canvas does not copy the source EXIF metadata, but when the tool returns your original untouched, its metadata remains too. Do not treat an unchanged download as a metadata-sanitised file. Inspect the resulting file before sharing it if location or other metadata matters.
Start near 80 and inspect the comparison. The right setting depends on the image and use. The same slider value can produce different output across formats and browsers. See MDN’s canvas export documentation for the browser encoding controls.
Choose target-size mode, check the completed result and leave some room below the receiving system’s limit. The tool uses 1,024 bytes per displayed KB, while some upload forms use 1,000; for a strict decimal-byte cap, choose a slightly smaller target. Read our file-size guide for the workflow.
This canvas-based tool is for still images. A GIF or animated input can be rendered as a single frame rather than preserving its animation. Keep the original animation or use an animation-aware encoder when motion matters.
Updated 19 October 2026: clarified lossy output, the original-file metadata exception, binary file-size units and still-image limitations. Published by Talentra Solutions · Report a tool issue.