About this site

FlipPicture reflects a picture across the upright axis, the horizontal axis or both, turns it into any of the eight orientations a rectangle can take, builds a symmetrical frame from one half, and makes three mirror effects from the same engine: a reflection cast off one edge, a kaleidoscope, and the original beside its own mirror. Around that engine sit the same reversal in other forms — a JPEG flipped by moving its compressed blocks, PDF pages mirrored as vectors, typed text set and mirrored, colours inverted exactly, and a live camera shown un-mirrored. Every one of them is a reversal, and that is the entire product. There is no account, no queue, no gallery and no upload; the file is decoded by the browser, redrawn onto a canvas, and handed back as a download.

The narrowness is deliberate rather than a stage on the way to something larger. A reflection has a small number of parameters and a surprising number of ways to be got subtly wrong — orientation flags that are themselves mirrored, a re-encode that costs quality the transform did not, a preview that does not match the export — and a page that only has to get one operation right can afford to get those details right.

It turns pictures only in the quarter steps that combine with a reflection — the eight orientations a rectangle can take — and never by an arbitrary angle. It does not crop, straighten against a reference, or repair an orientation flag as a feature in its own right. Its effects stay inside the mirror: a reflection cast off one edge, symmetry built from a half or a quarter, and the original beside its own mirror in one file. It does not blend layers or reverse a text layer; each of those is a different product.

The rule the whole site is built on is that a feature does not get added because the engine happens to make it cheap. A reflection tool with six unrelated buttons on it is worse at reflecting than one without them.

A statically exported Next.js site on a content delivery network, with no server behind it and no interface that could accept a file if one were offered. The image work is a canvas transform, a small orientation reader, an encoder call and a byte-level metadata copier, all of it running in the tab. Heavier pieces — the HEIC decoder in particular — are fetched only at the moment a file needs them, so opening the page does not download a codec on the chance you might.

The build refuses to ship if any module that touches an image buffer contains a network call, if a colour is written outside the token file, or if the promise about local processing is missing from a page. Those are automated checks rather than intentions, because an intention is not a guarantee.

The site is built with code other people wrote and licensed for anyone to use. Everything below runs in your browser; nothing is fetched from anyone else’s server.

• Next.js and React — MIT licence.
• clsx — MIT licence.
• libheif-js (the HEIC reader, loaded as a separate file only when a HEIC is opened) — LGPL-3.0; its source is published by its authors.
• Anybody and Karla typefaces, served from this site — SIL Open Font License 1.1.

The JPEG block flip, the PDF mirroring, the proof-sheet writer and the metadata copier are this site’s own code, written for it, and use no third-party library.

Nothing to use, and there is no account or paid tier behind it. The site carries no advertising.