Convert PNG to JPG

How to convert PNG to JPG

1. Choose one or more PNG images — drag them in or use the file picker.
2. Set the JPG quality if you want a smaller file or sharper output.
3. Download the JPG — a single file, or a zip if you converted several at once.

A screenshot converted to JPG can come out larger than the PNG it replaced

Before anything is compressed, PNG filters each row against its neighbours. Five filter types — None, Sub, Up, Average and Paeth — let the encoder pick a predictor per scanline and store only the gap from it: Sub looks left, Up to the row above, Paeth weighs left, upper and upper-left together. A row of flat grey, or one identical to the row above, collapses to a run of zeros that DEFLATE codes for almost nothing. Screenshots, diagrams and logos are built from exactly that redundancy.

Baseline JPEG, standardised as ITU-T T.81 in 1992, does the opposite. It splits the image into 8×8 blocks, rewrites each as a sum of cosine waves and quantises the higher-frequency terms hardest, betting that tone varies smoothly and the eye forgives fine detail. A photograph pays off that bet: nearly all its energy sits in the low frequencies, so at a mid-range quality setting JPEG shrinks it roughly six- to nine-fold — more on a large frame, less on a web-sized one — with loss most viewers never notice. That is the intended conversion.

A hard edge is the one thing that assumption forbids, because a step holds energy at every frequency. Throw the high terms away and the reconstruction overshoots and oscillates — the Gibbs phenomenon — so black text on white gains the grey shimmer that betrays a JPEG screenshot at a glance. The encoder also spends bits on every block an edge crosses, and a full-screen capture is nearly all edges, so a screenshot re-encoded as JPEG can finish larger than the PNG it replaced, artefacts included.

Under one bit per pixel, JPEG only beats the PNG on size by adding visible artefacts

The file hints at the answer. Multiply its size in bytes by eight, divide by width times height — dimensions a viewer or the operating system reports, since this tool shows only the byte count — and read the bits per pixel. A 1920×1080 screenshot in 200 kB sits near 0.8; a real photograph, once downscaled or saved through JPEG even once, lands around two to six, and only an untouched camera original still carrying sensor noise climbs into double digits. The screenshot is where PNG has already won.

A faster check is provenance: ask what made the file. A camera or a scanner yields a photograph, and PNG was always a poor container for one. A screen capture, or an export from a spreadsheet, a slide or a vector editor, yields line art with the redundancy PNG feeds on. The clean 3-D render is the odd one out — no sensor noise, broad flat fields, geometrically hard silhouettes — so it behaves like line art, not like the camera shots it sits beside on that list.

So the sub-one-bit PNG is not a dead end for JPEG. Drop the quality setting far enough and the file comes out smaller than the PNG — but only by coarsening the text edges PNG stored cleanly. The hybrid is where this bites, and it reaches a converter as a PNG because slide exporters and design tools emit it: a chart beside a photo, a caption over a product shot. JPEG has no notion of regions and quantises the whole frame uniformly, so the ringing the lettering sets off spills into the smooth photo beside it. Judge a hybrid by its text; it fails first.

Questions