Enlarge a photo without it going blurry

Yes. Resizing stretches the pixels you already have, so a blurry photo becomes a larger blurry photo. This uses a model trained on millions of image pairs to work out what detail plausibly belongs in the gaps — edges stay sharp and text stays readable at sizes where a plain resize turns to mush.

The image is cut into sections and several are worked on at once, so the estimate shown next to each option is what you should expect — it is measured, not guessed. 4× takes roughly four times as long as 2×, because each section produces four times the pixels. If several people are enlarging photos at the same moment, yours waits its turn first.

Enlarging a 4000x3000 photo by 4× produces around 192 million pixels, which is more than is sensible to hold in one piece. It is done in overlapping sections and reassembled, keeping memory flat regardless of size. The overlap gives the model context at each edge, which is why the joins are invisible rather than showing as a faint grid.

The limit is time rather than pixels. The image is cut into sections and each takes about ten seconds, so a job is refused when it would not finish inside the fifteen-minute deadline — around 4 megapixels at 2x, and a quarter of that at 4x, since each section then produces four times the pixels. If yours is refused, bring it down with the image resizer and enlarge that. Refusing up front is deliberate: the alternative is a long wait that ends in failure.

It invents detail that was never recorded, so it is a plausible reconstruction rather than a recovery of the truth. Faces can end up subtly wrong, and it will not read a licence plate that was never legible. Treat the output as a better-looking picture, not as new evidence.