Enhance a flat photo or scan

No, and the distinction matters here more than it usually does. This is arithmetic on the pixels you supplied: a histogram is measured, a scale factor is applied, and neighbouring pixels are compared to find edges. Nothing is generated, nothing is predicted, and no model is downloaded or run. The practical difference is that everything in the output was in your image already - which is exactly what you want from a photograph of a document you may have to rely on.

It finds the darkest and lightest tones actually present, ignoring a fraction of a percent at each end so one dead pixel or one specular highlight cannot spoil the measurement, and stretches everything between them to fill the full range. The same scale is applied to red, green and blue, so a neutral grey stays neutral and no colour can appear that was not there. That last point is a deliberate limitation: stretching each channel to its own range would also remove colour casts, but it cannot tell a cast from a subject that is genuinely one colour, and it turns a sunset grey.

No. Sharpening increases the contrast at edges that are already there, which makes an image look crisper and can undo the softening that a resize or a heavy JPEG pass introduced. It cannot recover detail the lens or the sensor never captured - a photograph taken out of focus stays out of focus, with slightly more contrast at the blurry edges. Anything claiming otherwise is generating detail and guessing at what it should be.

Because there was nothing to correct, and the tool prefers saying so to doing something anyway. An image that already spans the full tonal range gets no stretch, and a flat area with no edges gets no sharpening - trying anyway would amplify the paper grain in a blank scan into visible blotches. The result panel tells you which corrections actually ran rather than claiming both every time.

Yes. A batch of photographs from one camera under one set of lights usually needs the same correction, which is exactly the case where doing them individually is tedious. Each image is measured on its own, so a folder of mixed exposures gets the right stretch for each one rather than an average that suits none of them.

No. The corrections run on your device, in your browser, and a content security policy the browser enforces stops this page sending anything to another domain. There is no model to fetch and no service to call, so unlike most of what this site does, this tool would work with the network switched off after the page had loaded.