EnlargePictureA PictureEditor.com tool

The reasoning behind the buttons

The tool takes about four seconds to use. These take rather longer, and they exist because the four-second version cannot tell you whether to believe the picture it just handed you.

Why we wrote any of this down

Almost everything published about making pictures bigger is written by somebody with a subscription to sell, which is why it all says the same two things: that quality is preserved, and that the results are stunning. Neither claim survives five minutes with a loupe and a picture you care about. So the guides below are the arguments we would make in person, with the losing cases left in.

They share a subject with the tool and they do not share its job. The pages with a factor in the title are for getting something done. These are for the half hour before or after, when the question is not how but whether — whether 8× is worth attempting on this file, whether the sharper result from another site is sharper or merely more confident, whether the softness you are looking at is even a size problem.

Nothing here is a case study with a cherry-picked crop. What these three say about the four methods in the tool you can settle in half a minute with your own picture and the magnifier on the front page. What they say about trained models you cannot, because this build does not have one — that half is an argument rather than a demonstration, and it is written to be argued with.

The three of them

  1. What enlarging can and cannot get backThe one to read if you are about to rely on a result. It works through 2×, 4× and 8× and names, at each step, which part of what you are looking at was recorded and which part was worked out.You have a small photo and a decision resting on it.
  2. Where a trained model beats arithmetic, and where it doesn’tA comparison written by somebody who had to implement the unfashionable side properly, including the sources where a network wins outright and the ones where it quietly ruins your picture.You want to know whether the AI button elsewhere would have done better.
  3. Enlarged and still softSix things that produce a soft result, in the order worth checking, ending with the one nobody wants to hear: sometimes the problem was never the size.You already enlarged it and the result disappointed you.

Two more are written and not published

One on running the clean-up pass before the enlargement rather than after, and one on how big a photograph can be printed before the viewing distance stops rescuing it. Both are drafted; neither is here, because a guide that repeats what the tool already says on screen is furniture rather than depth. They land when they have something the interface does not.

Or skip the reading

If you already know what you need, the factor pages open straight at it and every one of them carries the short version of the argument at the bottom.