Convert JPG to PNG

How it works

The JPG is decoded into a bitmap, drawn onto a canvas of the same dimensions, and encoded as PNG. There is no quality setting because PNG has none: the encoder is required to reproduce every pixel exactly, so the only variable is how well the compression algorithm happens to work on your particular image.

That compression is worth understanding, because it explains the wildly different results people get. PNG uses filtering followed by DEFLATE. The filter step predicts each pixel from its neighbors and stores the difference, which produces long runs of zeros wherever an area is flat or changes smoothly in one direction. DEFLATE then compresses those runs very efficiently.

On a diagram or a screenshot of an interface, most of the image is flat color, the differences are almost all zero, and the file shrinks dramatically. On a photograph, adjacent pixels differ slightly and unpredictably everywhere — sensor noise alone guarantees it — so there are few runs to exploit and the file stays large. Same algorithm, opposite outcomes, entirely because of the content.

If a maximum width is set, the image is scaled first, using progressive halving so the reduction stays clean. Scaling before encoding also reduces the PNG's size roughly in proportion to the pixel count, which is the only lever available for making a PNG smaller.

When you'd use this

Editing pipelines are the strongest case. If an image will be opened, altered, and saved several times, each JPEG save compounds the loss — the effect that makes a meme reposted twenty times look like it was photographed off a monitor. Converting to PNG once at the start stops that accumulation.

Software requirements are the second. Some design tools, game engines, embedded systems, and older document formats accept PNG only. In those cases the size increase is simply the cost of the requirement.

The third is misidentified content. Diagrams, charts, logos and UI captures that were exported as JPG by a tool that defaulted to it are being penalised twice: they are larger than they need to be and they have visible artifacts around every edge. Converting them to PNG will not remove the existing artifacts, but it stops the problem growing and often reduces the file size at the same time.

A note on what to expect

If you convert a photograph and the PNG is enormous, nothing has gone wrong. If you convert a diagram and the PNG is smaller, nothing has gone wrong either. The two results come from the same process meeting very different images, and a converter that produced consistently similar sizes for both would be doing something other than what it claimed.