What about tools that advertise "AI lossless upscaling"?

Two cases. Some genuinely do have a super-resolution model: having seen many small/large image pairs, it can guess plausible texture for a blurry face — note that it is guessing, the detail it produces was not in your original, and faces in particular can come back looking like somebody else. The rest is resampling plus sharpening, the same as this page, under a louder name. Either way, "lossless" is untrue.

Why is my photo still soft after enlarging?

Because the softness was already baked into the original. Resampling can make the transitions look natural, but a missed focus, camera shake or high-ISO noise are things that were lost when the shutter fired; enlarging only enlarges them too. For pictures like that, denoising and sharpening the original is usually more use than enlarging it.

How far can I push it?

The slider goes to 400%. The number matters less than how much information your source holds: a clean 1200px screenshot at 400% is usually still usable, while a 300px phone screenshot at 400% is just larger mush. We don't set a file-size or pixel cap — what you'll eventually hit is the browser's maximum canvas dimension, and it will say so.

How is this different from the resize page?

That one is a general tool that works both ways (shrinking included), with one scale slider and a smoothing switch. This page only enlarges, adds a sharpening step aimed at the interpolation transition, and its defaults are tuned for enlarging. Use the resize page to make things smaller — sharpening usually hurts when you shrink.

Enlarge an image: first, enlarging is never lossless

2×, 3×, 4×. What it can do is keep the edges clean; what it can't do is invent detail.

or start a blank document

Let's get this out of the way: enlarging cannot be lossless. A 400×300 image holds 120,000 pixels. Blow it up 4× and of the 1.92 million pixels you end up with, 1.8 million were calculated, not photographed. Every tool that advertises "lossless enlargement" is promising the same impossible thing.

So what does this page do? Two specific things. First, it spreads the pixels using the browser's highest-quality resampling (bicubic) rather than nearest-neighbour duplication, which produces stair-steps. Second, it sharpens after enlarging, tightening the soft ring that interpolation creates. The result has clean edges and doesn't look mushy — but texture that was never in the picture will not appear. Genuinely inventing detail takes a super-resolution model; we don't have one, so we don't claim it.

Key features

Sharpening happens after the enlargement

The other way round, the halo that sharpening creates gets enlarged too and turns into thick edges. Sharpening afterwards works on the soft transition interpolation actually produced. The difference is invisible in a thumbnail and obvious at 100%.

Sharpening is adjustable, and can be switched off

Over-sharpened photographs take on an oil-painting look and any noise gets amplified along with the detail. Pull it to 0 for plain resampling. The default of 40% is a safe setting for line art and screenshots.

What it helps most with

Clean sources: line art, icons, screenshots, bitmaps exported from vectors, flat-colour illustration. In images like these the information already lives in the edges, so getting the edges right makes the enlargement nearly invisible.

Nothing is uploaded, and there's no size cap

The whole computation finishes in your browser. There's no limit we impose either — the real ceiling is this computer's memory and the browser's maximum canvas dimension. When you hit it the browser says why, and we pass that sentence on unchanged.

How to use

1
Drag your image in

JPG, PNG and WebP are accepted, and Ctrl+V pastes from the clipboard.
2
Pick a scale

100% to 400% in steps of 25%. The higher you go, the larger the share of pixels that were calculated rather than captured.
3
Set the sharpening

Lower for photographs, higher for line art and screenshots. At 0 it's plain resampling.
4
Look at it at 100%, then download

The two sit side by side. Don't judge from the thumbnail — over-sharpening only shows at actual pixel size.

FAQ

What about tools that advertise "AI lossless upscaling"?: Two cases. Some genuinely do have a super-resolution model: having seen many small/large image pairs, it can guess plausible texture for a blurry face — note that it is guessing, the detail it produces was not in your original, and faces in particular can come back looking like somebody else. The rest is resampling plus sharpening, the same as this page, under a louder name. Either way, "lossless" is untrue.
Why is my photo still soft after enlarging?: Because the softness was already baked into the original. Resampling can make the transitions look natural, but a missed focus, camera shake or high-ISO noise are things that were lost when the shutter fired; enlarging only enlarges them too. For pictures like that, denoising and sharpening the original is usually more use than enlarging it.
How far can I push it?: The slider goes to 400%. The number matters less than how much information your source holds: a clean 1200px screenshot at 400% is usually still usable, while a 300px phone screenshot at 400% is just larger mush. We don't set a file-size or pixel cap — what you'll eventually hit is the browser's maximum canvas dimension, and it will say so.
How is this different from the resize page?: That one is a general tool that works both ways (shrinking included), with one scale slider and a smoothing switch. This page only enlarges, adds a sharpening step aimed at the interpolation transition, and its defaults are tuned for enlarging. Use the resize page to make things smaller — sharpening usually hurts when you shrink.

Try Enlarge an image: first, enlarging is never lossless now

Runs fully in your browser · no install · no signup · images never uploaded