Enlarge a low-resolution photo

A low-resolution photo usually has three separate problems wearing one coat. It is small, which is the one you noticed. It has been saved as a JPEG at a low quality setting, which leaves faint steps on an eight-by-eight grid across every flat area. And its colour channels were stored at half resolution, which leaves blotches of the wrong hue along any strong boundary — the reddish fringe on a dark branch against a bright sky.

Only the first is fixed by making the picture bigger. The other two get multiplied. That is the entire argument for running a clean-up pass first, and the reason this page opens with it switched on while the rest of the site does not.

It looks along the eight-pixel grid a JPEG encoder works on and softens a step only when that step is small enough to be an artefact rather than an edge somebody photographed. Then it blurs the two colour-difference channels and leaves brightness completely alone, which is why the picture does not go soft when you raise the strength — it goes less colourful in the blotchy places.

What it does not do is run a general noise reduction over the frame. That is the pass that makes skin look like wax and grass look like a carpet, and doing it before an enlargement bakes the wax in at four times the size.

• A picture that was re-encoded eight times is not coming back.Each round of a chat app throws away a little more and adds a little more grid. At some point the grid and the picture are the same signal and no pass can tell them apart. If the original still exists on somebody’s phone, an hour spent asking for it beats any setting on this page.
• Cleaning first costs you texture, always.Every strength above Off removes some real detail along with the artefacts, because the tool cannot know which faint variation was fabric and which was the encoder. Light is a small price; Strong is a real one, and it is worth checking the result at 1:1 before you accept it.

Why does the photo my friend sent look so much worse than theirs?

Because most chat apps re-encode what you send. The picture is shrunk to fit a width limit and saved again at a low quality setting, often more than once if it has been forwarded. What reaches you is a smaller picture carrying the encoder’s own grid pattern, and that grid is what the clean-up pass is aimed at.

How strong should the clean-up be?

Start at Light. Move to Medium if you can still see a faint square grid in flat areas like sky or a wall when you look at 1:1. Strong is for a picture that has visibly been through several rounds — and it is a real trade, because every step up also softens the texture you were hoping to keep.

Can this get back the detail the app threw away?

No. The clean-up pass removes damage that was added; it cannot return information that was discarded. What it buys you is that the enlargement afterwards spends its effort on the picture rather than on multiplying the damage by four.

• Double the sizeThe factor with the least to apologise for. Opens at 2× on a sharp source.
• Four times overThe usual answer when a picture is close but not close enough. Opens at 4×.
• Enlarged and still softSix reasons a bigger picture looks worse, and which one you actually have.
• What enlarging can and cannot get backWorked through at 2×, 4× and 8×, with the part that is a guess named as one.