Sharpen Image
Crispen soft photos with a real convolution filter โ honest about limits.
About this tool
A proper sharpening convolution with one slider, full-resolution processing on your device, and a preview that updates as you drag. No upsells, no watermark โ and no pretending sharpening is magic.
What sharpening actually does
Sharpening doesn't add detail โ it amplifies the contrast at edges that already exist, which the eye reads as crispness. That distinction predicts everything about using it well. It rescues: the slightly-soft focus, the photo that lost bite in downscaling (resizing always softens โ a sharpening pass after resizing is a standard finishing move), flat scans. It cannot rescue: real motion blur, deep defocus, or detail the sensor never captured โ cranking the slider on a truly blurry photo produces crunchy blur, not focus. The failure mode to watch is haloing: bright ghost-lines along high-contrast edges, the signature of over-sharpening, which is why this slider tops out at politely-aggressive rather than at fiction. Judge at 100% zoom โ a sharpened image that looks right zoomed out often looks fried up close. Noise sharpens too, so clean first, sharpen last. Photo filters handles the brightness-and-contrast half of apparent sharpness โ often the better first stop.
Frequently asked questions
Can this fix a blurry photo?
A slightly soft one, yes. Genuine motion blur or missed focus, no โ sharpening amplifies edges that exist and can't create ones that don't. Any tool promising otherwise is overpromising.
What amount should I use?
20โ40 for a gentle crisp, 40โ60 after downscaling. Past 60, watch edges for bright halos โ the universal sign of too much.
Why sharpen after resizing?
Downscaling averages pixels together, which softens edges โ a modest sharpening pass restores the bite. It's the standard last step of resize workflows everywhere.
Does the full-size image get processed?
Yes โ the convolution runs on every pixel at native resolution, on your device. Nothing is uploaded, nothing watermarked.