Image Upscaler (2x and 4x)

This is not a machine-learning upscaler. The image is drawn onto a canvas two or four times larger using the browser's high-quality smoothing, which blends neighbouring pixels to fill the new ones in between. A gentle 3 x 3 sharpening filter is then run over the result to counter some of the softness that enlarging always adds.

AI upscalers work differently: a trained model predicts plausible texture, such as hair strands or brick edges, that was never in the file. That can look impressive but also invents details. Interpolation never invents anything; what you see is a smoother, larger version of what was there.

The factor applies to both width and height, so the pixel count grows with its square:

Large results take longer, because the sharpening pass visits every pixel, and need a lot of memory. On phones, a 4x enlargement of an already large image may fail or produce a blank result when it exceeds the browser's canvas limits. Upscale only as far as you need.

• A layout or print template requires a minimum pixel size and the source is slightly too small.
• Small graphics or icons need to be bigger and a little softness is acceptable.
• You want to crop into a small area and still have a workable pixel size.

It will not turn a tiny, blurry thumbnail into a sharp photo. If you can find a larger original, that will always look better.

Results are saved as WebP at quality 0.9. They are much larger in bytes than the original; run them through the Image Compressor if file size matters. For an exact target size rather than a fixed factor, the Image Resizer accepts any width and height.

Everything runs on a canvas in your browser. The image is not uploaded and no AI model is involved.