Just a minute...(/)
We are currently optimizing your images. Please wait a moment as we ensure the highest quality results for you.
Great, we have successfully compressed your images.
Images Optimized.
We are currently optimizing your images. Please wait a moment as we ensure the highest quality results for you.
Images Optimized.
There are really only two ways to make a JPEG smaller: throw away colour resolution, or re-encode the same detail more efficiently. Most online compressors take the first route because it is faster — they halve the colour data before they even look at quality, which is exactly what makes coloured text, thin gradients and skin tones look smudged afterwards.
This one takes the second route. Full colour resolution is kept, and the savings come from a high-quality re-encode with optimized Huffman tables and progressive scanning — packing the same detail into fewer bytes instead of deleting some of it.
The technical name for the shortcut is 4:2:0 chroma subsampling: half the colour information is discarded before compression even begins. On a landscape you might not notice. On a face, a sunset, or text on a coloured background, you will — it shows up as blotchy patches and soft edges that were sharp in the original. The default pass here uses 4:4:4, which keeps every pixel's colour intact.
Every image is converted to a standard sRGB profile before it is re-encoded. That matters for photos with an unusual or missing colour profile: without this step a viewer simply guesses, and the result is shifted skin tones or blown-out skies. Normalising first means the file looks the same wherever it is opened.
EXIF, camera details and GPS coordinates are stripped in the same pass. That trims a little extra size, and it means a photo you publish is not carrying the address it was taken at.
Compression runs once, on the file you uploaded. That is different from re-saving a JPEG repeatedly in a photo editor, where each save compounds the loss. It also means running an already-compressed JPEG through again is not worth doing — most of the recoverable size was taken the first time, and a second pass costs quality for almost nothing.
On the rare file the high-quality pass cannot handle cleanly, a more aggressive backup pass runs automatically so you get a smaller file instead of an error. It is the exception, not what a normal photo goes through.
Just want a smaller file without the detail? Compress JPG runs the same engine with less to read. Working with other formats? Compress PNG and Compress WEBP handle those, and JPG to PDF turns a set of photos into one shareable document.