Does this use AI to add detail?
No. It is honest canvas resampling: the browser interpolates between the pixels that are already in your file. AI upscalers work differently, by hallucinating plausible detail from a model trained on millions of images, which is why they can appear to invent texture and lettering. This tool cannot and will not do that, and it says so on the page.
So can I enlarge a small logo to a big one?
You can make it bigger. It will not be sharper. A 200px logo enlarged to 2000px has ninety-nine invented pixels for every one that was captured, and interpolation can only smooth between them. If the logo exists as an SVG or in a vector file, export it at the size you need instead: that is the only way to get a genuinely sharp large version.
When should I use nearest-neighbor instead of smooth?
For pixel art, QR codes, barcodes, and any image whose hard pixel boundaries are the point. Nearest-neighbor duplicates each pixel into an exact block, keeping the edges razor sharp. Use it on a photograph and you get visible blocky stair-stepping instead of softness.
How far can I realistically enlarge a photo?
About 2x before the softness is obvious. At 2x, one real pixel becomes four; at 4x it becomes sixteen, and interpolation is guessing fifteen of them. If you need the image bigger than that and it matters, reshoot it, find the original, or accept that it will look soft.
Why is my output rejected as too large?
The tool caps output at 40 megapixels so a large source at 4x cannot exhaust the browser's canvas memory and crash the tab. Lower the scale factor, or switch to Target dimensions and enter a size under the cap.
Is my image uploaded to enlarge it?
No. The resample runs on an HTML canvas inside the page. The image never leaves your device, which is worth knowing given how many upscaling services require an account and an upload.