Enlarging an image sounds trivial, and doing it badly is trivially easy. Drag the corner of a photo box in any editor and you will produce something that looks worse than the original, with soft edges and a faint halo around every line. Enlarging properly is a different operation.
What upscaling actually does
The straightforward way to make an image bigger is to copy pixels. Each new pixel takes the colour of its nearest neighbour, which produces the jagged stair-steps you recognise as a blown-up thumbnail. This tool does something else. It looks at the structure in the image, finds the edges and the shapes inside it, and rebuilds the picture at the larger size with cleaner transitions. The result has the same content but noticeably smoother contours.
What it cannot do is create detail that was never recorded. A 300-pixel image of a face contains a very rough description of that face, and the enlargement is a fuller rendering of that same rough description.
How far to push it
The useful range depends entirely on the source. Doubling a decent photograph is usually invisible as a transformation. Tripling is where you begin to see softness. Quadrupling something tiny produces a large file that is smooth but obviously blurred, and it is rarely better than finding the original.
Before enlarging much, it is always worth spending thirty seconds looking for a better source. A photo embedded in an old email, a version in the original archive, a camera file you forgot about. Any of those beats enlarging a compressed copy, because they contain information this one has already lost.
Working out if it is big enough to print
Printing needs pixels, and the number is worth checking before you spend time enlarging. For a clean A4 print at 300 dots per inch, the long edge wants to be roughly 3500 pixels. At 150 dpi, which is acceptable for a rough draft, half that is usable. A phone photo is usually big enough for A4 without any enlargement at all, which is the honest answer more often than people expect.
Enlarging is also useful for display, where a small image stretched across a large screen is far more obvious than a small image in a printed page.
Smaller sources, larger displays
Print is the obvious case and the one people ask about first. The other is display, where the arithmetic works differently. A small image stretched across a large monitor shows its weaknesses plainly, because there is physical space for the artefacts to appear. On a phone held at a normal distance the same image looks fine. The mismatch between a thumbnail and a widescreen display is where enlarging earns its keep, particularly for an image embedded in a page you cannot go back and rebuild.
It is also worth using when the picture will be shown at a moderate size but was sourced from something slightly too small. A modest jump of around one and a half to two times removes the jagged edges that give a thumbnail away, and at that viewing size the improvement is real rather than cosmetic.
What it will not do is rescue an image that is already blurred or heavily compressed. Those artefacts are in the file, and making it bigger only gives you a larger version of the same problem.
Related reading
Guides that pair with this