Compress Image

Most people need the first mode. Upload forms rarely care about quality numbers; they check bytes, and they reject anything over the line. A target lets you type that line once and get a batch of files that all pass, each at the best quality it can afford. The search runs on every file separately, so a detailed landscape and a plain document scan in the same batch each get their own setting.

The other two modes give you direct control. Quality mode is predictable in how the photos look and unpredictable in how much they weigh. Percentage mode sits between the two: you decide how much lighter the files become relative to where they started, and the tool works out the quality needed for each one.

Quality is a scale inside the encoder, not a percentage of the picture that survives. A setting of 75 does not throw away a quarter of the photo; it tells the encoder how much fine detail it may discard in exchange for a smaller file. The visible effect depends on the content. A portrait with a blurred background holds up well at low settings, while a page of small text or a screenshot with thin lines shows damage much sooner.

The two ends of the scale behave differently. Between 60 and 85, each step down saves a fair amount of space for a small visible change, which is why most web photos live there. Above 90, the file grows quickly while the picture hardly changes, and below 40 the savings get small while the damage becomes obvious. When a target is tight, it is often better to accept a slightly smaller pixel size at a decent quality than a full-size picture at a very low one; that is the trade the downscaling switch makes for you.

Every lossy save adds a little loss on top of the last one. If you plan to edit a picture again, keep the original and compress a copy at the end, rather than compressing the same file several times.

Same as input keeps a JPG a JPG and a PNG a PNG, which is the safe choice when the form or person on the other side expects a particular type. Changing the format is often the bigger win, though. A photo stored as PNG is lossless and heavy by design; converting it to JPG or WebP usually brings a much larger drop than any amount of PNG tuning. If you only need a different format and no size limit, the image converter page is the more direct tool.

WebP and AVIF are newer formats that reach the same visual quality as JPG in fewer bytes, and both keep transparency. Every current major browser displays them, so they are a good fit for pictures on a website. Older desktop software and some upload forms still refuse them, and in those cases JPG remains the format that opens almost everywhere. MDN's guide to image file types compares the formats and their browser support in detail.

PNG output has its own switch. Lossy reduces the picture to a limited palette of colors, which works very well for logos, icons, charts and screenshots and keeps the alpha channel. Lossless keeps every pixel exactly and only reorganizes how the data is stored, so the savings are smaller but nothing changes visually.

Allow downscaling is on by default. It lets the tool reduce the width and height when even the lowest allowed quality is still over your target, which is common with large camera photos and small limits. Set a Minimum width if the form also names a pixel size you must not go under. When a form gives exact dimensions as well as a KB limit, use Resize first to set the width, height or longest side before compression begins, or prepare the size with the Resize Image tool.

Chroma subsampling controls how much color detail is stored. 4:2:0 records color at half the resolution of brightness, which the eye rarely notices in photos and which saves a lot of space. 4:4:4 keeps full color resolution, so red text and thin colored lines on a screenshot stay crisp. Auto picks for you based on the image.

Progressive JPG stores the picture in passes, so a slow connection shows a rough version first and then sharpens it; such files are often slightly smaller too. Strip metadata removes GPS position, camera data and comments, and Keep color profile holds on to the embedded ICC profile so colors do not shift. Both are on by default, and Auto orient turns phone photos upright using the orientation the camera recorded.