Compression removes or describes information more efficiently. The best strategy depends on whether every original value must survive and how the image will be used.
Lossless compression
Lossless formats reconstruct the encoded pixel information exactly. PNG finds repeated patterns and applies filtering without discarding pixel values. It is useful for diagrams, UI, and transparency, but photographs can remain large.
Lossy compression
Lossy codecs discard information that is expected to be less noticeable. Lower quality usually creates a smaller file, but aggressive settings introduce blockiness, ringing, banding, or smearing. Always inspect important images.
Dimensions matter more than many people expect
Halving both width and height reduces the pixel count to one quarter. When a strict KB target cannot be achieved through encoding quality, reducing dimensions is usually the next responsible step.
Metadata is separate from pixels
EXIF and related blocks may store camera settings, dates, software, and GPS coordinates. Re-encoding rendered pixels without copying those blocks removes common metadata and can slightly reduce size.
Why exact KB is iterative
File size cannot be predicted perfectly from a quality number because image complexity varies. Target-size tools encode several candidates, search for the highest acceptable quality, then scale dimensions if required.
Animated GIF is different
GIF stores indexed-color frames. Its size depends on dimensions, frame count, motion, and palette complexity. Reducing colors and dimensions can help; short video formats are often more efficient when platform support allows them.
Process it privately in your browser.