How to Upscale an Image Without It Turning to Mush

Enlarging an image does not create detail — it spreads the detail you already have across more pixels, and fills the gaps with a guess. Understanding that is the difference between a usable 2× enlargement and a blurry mess.

What upscaling really does

A 400 × 300 image enlarged to 800 × 600 has four times as many pixels, but only a quarter of them contain real information. The other three quarters are interpolated by an algorithm. Better algorithms make smarter guesses, which is why the result can look quite good — but no algorithm can recover detail that was never captured.

Interpolation methods, in plain terms

Nearest neighbour — copies pixels. Fast, blocky; only useful for pixel art where you want hard edges.
Bilinear / bicubic — averages neighbouring pixels. Smooth and slightly soft; the default in most editors.
Lanczos — a wider, weighted average that preserves edges better. This is the best general-purpose choice, and what quality resamplers use.
AI / neural upscaling — a model trained to hallucinate plausible detail. Impressive on faces and textures, but it invents content, which makes it unsuitable for evidence, documents, or anything where accuracy matters.

What upscales well

Screenshots and UI captures — text and interface elements have crisp edges that resampling handles well, especially with a sharpening pass.
Logos and flat graphics — few colours, hard edges, no noise.
Slightly soft photos — a sharp photo enlarged often looks worse than a slightly soft one, because artefacts get magnified too.

What does not

Heavily compressed JPGs — you will be enlarging the compression blocks as well as the picture.
Very small thumbnails (under ~200 px) — there is simply not enough information.
Noisy low-light photos — grain gets scaled along with everything else.
Fine text below ~8 px — it turns to soup regardless of method.

The workflow that works

Start from the best source available. The original file, not a resized copy or a screenshot of a screenshot.
Use PNG as the intermediate format so you do not stack compression artefacts on top of interpolation.
Go in one step. Scaling 400 → 1600 directly is better than 400 → 800 → 1600.
Apply light sharpening after scaling. Interpolation softens edges; a gentle unsharp pass restores apparent crispness. Overdo it and you get halos.
Export to PNG to preserve the result, then convert to JPG at high quality if you need a smaller file.

How much can you enlarge?

For clean sources, 2× is usually indistinguishable from the original at normal viewing distance. 3× is acceptable for screen use. 4× and beyond is where artefacts become obvious — fine for a background or a mock-up, not for something that will be examined closely. Going from a 200 px thumbnail to a 2000 px print is not realistic with any method.

Set expectations for print

Print needs roughly 300 pixels per inch at final size. A 1000 px wide image prints sharply at about 3.3 inches; at A4 width (8.3 inches) it will look soft no matter what you do. Upscaling can make it acceptable at a distance, but it cannot make it sharp.

Try it now: Upscale Image · Resize Image