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.

Around eight seconds per section of the image at 2×, and four times that at 4× because each section produces four times the pixels — the estimate shown next to each option is measured, not guessed. Jobs run one at a time, so a busy moment adds to it. Nothing is downloaded to your device: the model already lives on our machine and stays loaded between jobs, which is why this is far quicker than the same work done in a browser.

Enlarging a 4000×3000 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.

It exceeded the four-megapixel input limit, which exists because each job holds the only slot for as long as it runs and a very large image would block everyone else. Use the image resizer to bring it down first, then enlarge that. Refusing up front is deliberate — the alternative is a 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.