Upscale Image

What enlarging actually does

Worth thirty seconds, because it explains every result you will get.

An image is a grid of pixels. A 500-pixel-wide picture holds 500 pixels of information across, and that is all the information there is. Enlarging it to 2000 pixels means the computer has to produce 1500 pixels that were never recorded, and it does that by looking at the neighbours and blending between them.

That blending is smart — it is a smooth, weighted average rather than a crude copy, which is why the result looks better than simply blowing the pixels up. But it is still an educated guess between known points. No new detail appears, because there is none to find.

This is the honest difference from the tools advertising "AI upscaling". Those run a trained model that invents plausible detail — it will put in eyelashes, brick texture or fabric weave that was not in your file, based on what such things usually look like. Sometimes that is magic. Sometimes it invents the wrong thing, and on faces and text it can be unsettling. Either way, it is a different operation from this one, and it is worth knowing which you are getting.

Where this genuinely excels

There is a whole category of image where interpolation is not a compromise at all — it is the correct method, and the result is effectively perfect.

Flat colour and clean edges. A logo, an icon, a line drawing, a diagram, a chart, a cartoon, a UI screenshot, a signature, a simple graphic. These images are large areas of uniform colour with defined boundaries between them, and blending between neighbours reproduces that faithfully. There is no texture to lose and no detail to invent, so nothing is lost.

Enlarge a logo 4x and it will look clean. That is the case to use this for, with confidence.

Meeting a minimum size. A print shop that wants 1000 pixels, a marketplace that rejects images under a certain width, a template with a fixed slot. The requirement is dimensional, and this satisfies it in seconds.

Anything viewed at a distance or at a glance — a banner, a background, a slide image, something on a wall. The softness that would bother you at arm's length is invisible at three metres.

Where it will disappoint you

Equally specific, so you do not waste time finding out.

A blurry photo stays blurry, only bigger. Enlarging cannot undo camera shake, a missed focus or motion blur — the sharp version was never recorded. If your photo is out of focus, no amount of enlargement changes that, and Sharpen Image cannot either. That page says the same thing about the same problem.

A low-resolution photo stays low-resolution. A face that is 40 pixels across has about 40 pixels of facial information. Make it 160 pixels and you have a smoother version of the same 40 pixels of information. This is why "enhance" in television programmes is fiction.

Enlarging magnifies every flaw. This is the one people do not expect. A heavily compressed JPEG has blocky squares and halos around edges; enlarge it 4x and those artefacts become four times bigger and much more obvious. Sensor noise, banding in a sky, a watermark someone tried to blur — all magnified. If your source is a small compressed JPEG saved from a chat app or a website, enlarging usually makes it look worse rather than better. Start from the best copy you can find.

Pixel art comes out blurred. Retro game sprites, 8-bit icons, QR codes and anything whose charm is the hard square edges need the opposite of smoothing — each pixel copied as a crisp block. Smoothing is always on here, so those come out soft and wrong. For pixel art you want a tool with a nearest-neighbour or "sharp edges" option.

Practical settings

The scale. 2x is the default and where quality holds up best. 3x is a reasonable stretch. 4x is fine for flat artwork and noticeably soft on photographs. Only go as large as you actually need — there is no benefit in overshooting, and the file grows with the square of the scale, so 4x is four times the file size of 2x.

Quality applies to JPEG output and defaults to 95%, which is high. Leave it there. This is the wrong moment to save file size — you have just made an image larger in order to look better, and compressing it hard undoes the point. Shrink the file afterwards with Compress Image if you need to, where you can see the trade-off directly.

The format is chosen for you, correctly. A PNG comes back as a PNG with its transparency intact — which matters, because logos and icons are the best use of this tool and they usually have transparent backgrounds. Everything else comes back as a JPEG.

One habit worth keeping: look at the result at 100%, not fitted to the screen. That is the size someone will see it at, and it is the only honest test of whether the enlargement worked.

What to do when it is not enough

If your enlarged photo is not good enough, the order of things to try is: find a bigger original — a photo app, an email attachment, the camera roll, the original file rather than a messaging-app copy, which are almost always heavily reduced. Then retake or rescan it if that is possible, at a higher resolution. Then use it smaller, which is often the right answer. And only then reach for an AI upscaler, accepting that it will invent detail.

Everything runs on your own device. No upload, no watermark, no account and no queue, and there is no limit on how many images you enlarge.

Related tools

Resize Image is what you want for making an image smaller, or for hitting exact dimensions rather than a multiple. Sharpen Image adds apparent crispness to an enlarged image — use it lightly and afterwards. Compress Image brings the file size down once you are happy with the result.