About this instrument
This page does one thing to a photograph: it changes how definite the edges look. Not the exposure, not the colour, not the size, not the blemishes. One dimension of picture quality, with every number that governs it shown rather than hidden behind an automatic setting.
Why the scope is this narrow
Sharpening is the operation that is most often done badly, and it is done badly for a specific reason: it is the operation whose result cannot be judged in the view most people look at it in. A tool that puts a sharpening slider beside twenty other sliders, on a canvas fitted to the screen, has arranged for its users to fail. Everything about this page — the four steps, the compulsory 1:1 view, the ring that appears around the worst edge — exists to make the judgement possible rather than to make the operation fast.
The second reason for the narrowness is the deblurring question. A large number of people arrive at any sharpening tool wanting something the physics does not allow, and the ordinary commercial response is to give them a slider marked unblur and let them decide whether they are pleased. This one draws the distinction instead: a frame smeared by even, straight movement has had its information transformed and a transform can be partly inverted; a frame that was out of focus has had its information averaged away and nothing brings it back. Saying that plainly loses a certain number of visitors. It is the only version of the feature worth having.
How it is built
The whole engine is TypeScript running in the browser tab. Separable box cascades stand in for the Gaussians, so a radius-80 pass costs no more than a radius-8 one; a guided filter supplies the base layer for local contrast; a hand-written radix-2 transform carries the deconvolution and the motion estimator. The arithmetic runs in a worker so a slider stays a slider, and the inspection patch is cut at full resolution with an apron of real picture around it so that what is inspected and what is exported are the same computation on the same pixels.
What is not here, on purpose
There is no automatic setting that decides for you, no upscaling, no face-aware retouching, no named looks, and no global tonal correction — even though clarity plainly wants a shadow lift beside it. Each of those is a different problem with its own failure modes, and putting them on one page would mean doing all of them at the level of an afterthought.
Who publishes it
It is published by the team behind PictureEditor.com, credited in plain text at the foot of every page here. Advertising is not yet running: Google has not approved the domain, the switch is off, and no unit renders on any route — what you are reading is the whole page. When approval arrives, three placements are already settled — one on an instrument route, two on a reference page, none of them above the working surface or inside it — and that is what is meant to pay for the hosting. There is no account, no sign-in and nothing to buy, which is a consequence of the architecture rather than a promotion: there is no server behind this site to hold an account on.
Corrections
The print sharpening figures are conventional starting points from press practice rather than measurements from a calibrated machine, and the motion estimator is asserted against synthetic plates with a known kernel rather than against a body of real photographs. If you have evidence that a number here is wrong, it is worth more than an opinion about whether the page feels right, and it will be acted on.