Image Compressor & Converter

Who it is for

It suits anyone who needs a photograph to fit an upload limit, a set of product shots trimmed to one dimension, or a screenshot in a different format. It is a batch re-encoder, not an editor: no cropping, rotating, retouching or colour correction.

What each setting does

Output format

"Keep original format" re-encodes each file as its own type. Choosing JPG, PNG or WebP re-encodes to that type instead: the picture is drawn onto an HTML canvas and written out again, and the file extension is renamed to match, so photo.png becomes photo.jpg. That canvas step writes at a fixed 92 per cent quality, so the slider has less influence when you also change format; use the target size to push a converted file down.

Requested image quality

The slider runs from 10 to 100 per cent and starts at 80. It is passed to the encoder as a request, not a guarantee, and it only means something for lossy formats: PNG is compressed without discarding detail, so the slider barely changes a PNG-to-PNG result. Shrinking the dimensions does far more.

Maximum width or height

Zero keeps the original pixel dimensions. Any other number is an upper bound on the longer side with the aspect ratio kept, so 1600 turns a 4000 × 3000 photo into 1600 × 1200. It is a ceiling, not a target: a smaller picture is not enlarged.

Target file size

A figure between 1 and 100,000 KB adds a second pass that tries to reach it, and the smaller of the two attempts is kept. It is best effort: the result line reports either "Target met." or "Target not met; try a smaller dimension or a different format." A target far below what the current dimensions can hold will not be reached by quality alone.

Download option

Individual downloads give each result its own button. The ZIP option also builds images-optimized.zip in the browser and reveals one download button when processing finishes.

What it can read and write

Input is whatever your browser can decode as an image; anything it does not report as an image type is refused with "Please select image files." Output is the original type, JPG, PNG or WebP — HEIC, AVIF, TIFF, SVG and PDF are not offered as outputs even when the browser can open them.

Writing depends on the browser too: where it cannot encode the type you asked for, the canvas standard has it fall back to PNG, which can be larger than the file you started with. An animated GIF or WebP loses its animation, because the canvas holds one frame and the output is a still picture of the first.

Transparency, colour and metadata

JPG has no alpha channel. Converting a PNG or WebP logo with a transparent background to JPG therefore flattens it, and in current browsers the transparent areas usually come out black rather than white. Keep PNG or WebP when transparency matters.

Because every output is written from decoded pixels, the surrounding metadata is not carried across: EXIF fields such as camera model, capture date and GPS coordinates, plus embedded ICC colour profiles and XMP data, are absent from the new file. That helps if you would rather not publish where a photo was taken, and hurts if you relied on the capture date or a wide gamut.

Lossy compression is not reversible, so re-processing an already-compressed JPG adds a second round of artefacts around edges and text. Work from the original file where you can.

Reading the result card

1. The heading line gives the output file name and its size to the nearest kilobyte.
2. Below it, both sizes appear to two decimals as "Original: … KB. Output: … KB.", ending with whether a target you set was met.
3. If a file could not be processed the card says "Processing failed; this download is the original file." — so that download is your untouched original.

Common mistakes

• Dragging the quality slider down to shrink a PNG screenshot. Reduce the dimensions, or convert to WebP, instead.
• Converting a transparent logo to JPG and only noticing the black background after it is published.
• Setting a 20 KB target on a full-resolution camera photo. Pair a target with a maximum dimension so the second pass has room to work.
• Deleting the original. Nothing here can restore the detail a lossy encode removed, and each further pass adds more.

For a longer explanation of what you trade away at each quality setting, see image quality versus file size.

Limits

• One set of settings for the whole batch.
• No cropping, rotating, watermarking or colour editing.
• Animation, EXIF, ICC profiles and XMP data are not preserved.
• The target size is best effort and reports a miss rather than trying indefinitely.
• Processing runs on the page's main thread, so long batches depend on the memory your browser gives the tab.

Questions