PNG to JPG: What You Gain, What You Lose, and When to Keep the PNG

PNG is lossless. It compresses by finding repeated byte patterns, the same way a ZIP does, so every pixel you saved is still there when the file is reopened. That is excellent for screenshots, diagrams, interface captures and anything with large areas of identical colour, because those areas compress extremely well without any quality cost. It is far less effective on photographs, where neighbouring pixels almost never repeat exactly: the compressor keeps hunting for matches, finds few, and the file stays large.

The other size driver is the alpha channel. A PNG can store per-pixel transparency, and transparency does not compress as neatly as solid colour. A 4000-pixel-wide screenshot of a transparent logo can easily weigh several megabytes, most of it spent describing which pixels are see-through. JPG has no alpha channel at all, which is simultaneously why it is smaller and why the conversion changes how the image looks.

A JPG cannot store a transparent pixel. Every pixel must have a colour, so anything transparent has to become something. Most converters fill it with white; some fill with black; a few let you choose. If you convert a transparent logo to JPG and view it on a dark page, you will see the rectangle you never knew was there — a white box around the mark. That is not a bug in the converter, it is the format honestly reporting that it has nowhere to put the information.

The right move is to flatten deliberately before converting. Compose the image over the background it will actually sit on — white for print, the page colour for a website, a brand colour for a slide — and export that flattened result as JPG. If the transparent version still has a use, keep it as PNG and treat the JPG as a delivery copy. Never overwrite the only copy of a transparent asset with a flattened one.

JPG is a photographic format. It discards detail the eye is unlikely to notice, concentrating on brightness and giving colour less precision, which is a very good model of human vision for photographs, skin tones and landscapes. At a quality setting around 80 to 85, a photo is typically a tenth of the PNG size with no visible difference at normal viewing distance. Lower settings start to show blocking in smooth gradients such as skies, and ringing around high-contrast edges.

Those edges are the problem for anything that is not a photograph. Text, logos, line art and screenshots are made of hard transitions between flat colours, exactly the content JPG's maths handles worst. You get coloured fringes and grey mush around every letter, and thin strokes can disappear entirely at small sizes. If the image contains words, a chart with labels, or a logo with fine detail, converting it to JPG for a smaller file is usually a false economy: the result looks broken and often needs redoing.

Start from the original PNG, not from a JPG you already exported. Each JPG pass is a fresh generation of loss, and re-saving a JPG compounds the artefacts of the previous one rather than starting clean. If the image has transparency, decide the background colour first and produce the flattened version as its own file so the decision is explicit and reversible.

Then check the result at the size it will actually be seen, not at 400 percent zoom. A quality setting that looks fine on a phone thumbnail may fall apart on a poster, and the reverse is also true: there is no point paying for 95 percent quality on an image that will render 300 pixels wide. Finally, compare the two file sizes before committing. If the PNG is only a little larger, keeping it avoids the whole set of trade-offs and preserves the alpha channel for future edits.