About ImageMunch

Why local-first

Most online image tools upload your pictures, process them remotely, and delete them "within a few hours." We simply skip the part where your photos travel. Your passport scan, your store's product shots, your screenshots with passwords visible — none of it needs a server to be resized or compressed, so none of it touches one here.

Who runs it

ImageMunch is an independent project. It is not a subsidiary or a white-label of any image software vendor, and there is no named editorial board or staff reviewer behind these pages — we would rather say so than invent someone to quote. Questions, corrections, and bug reports go to [email protected] and are read.

How we build and test

Every local tool that runs on your own computer is written for this site rather than wrapped around someone else's service, and every limit a page states is a limit the tool itself enforces — the 100 MB ceiling, the 40 megapixel ceiling, and the 30 files per batch are checked in the code that handles your picture, not only in the copy that describes it. The two AI generator pages are the honest server-side exception, and only the description you type is sent there.

Before a page ships, a script renders it and fails the build if the title duplicates another page's, if the export formats a page advertises are not the ones its own controls offer, if the diagram on the page no longer matches the tool it illustrates, or if the structured data disagrees with the text a person reads. That gate runs over the whole set, so a change to one page is checked against every other page's wording.

Each engine is then run against real files — photographs, screenshots, scans, and animated GIFs — in the browsers we can test here, and the result is compared with what the source application produces. We do not publish benchmark scores, star ratings, or "best tool" rankings: without a comparable test lab behind every product in a category, a number like that is decoration.

The third-party code that does the heavy lifting is open source, version-pinned, and served from this site rather than a CDN, so the version you get is the version we tested:

• libheif 1.23.2 · LGPL-3.0-or-later — Reading HEIC and HEIF files inside the convert and compress tools.
• libde265 1.0.x as bundled by libheif-js 1.23.2 · LGPL-3.0-or-later — The HEVC picture decoding that libheif calls when it opens an iPhone photo.
• gif.js 0.2.0 · MIT — Assembling and quantizing the frames of an animated GIF in the make and compress tools.
• omggif 1.0.10 · MIT — Reading the frames, delays, and disposal rules of an existing GIF before it is rebuilt smaller.
• tesseract.js 7.0.0 · Apache-2.0 — Driving the OCR worker that turns the text in a picture into selectable text.
• tesseract.js-core 7.0.0 · Apache-2.0 — The compiled Tesseract engine itself, running as WebAssembly inside the worker.
• eng.traineddata (tessdata_fast) 4.1.x, main branch · Apache-2.0 — The English language model the OCR page loads alongside the engine.
• MediaPipe Tasks Vision 1.0.1 · Apache-2.0 — Running the face finder that draws a box around each face on the blur and pixelate pages, and the segmentation engine behind the portrait background remover.
• BlazeFace short-range face detector (float16) 1 · Apache-2.0 — The trained weights the face finder loads to locate faces inside your picture.
• MediaPipe Selfie Segmenter (float16) latest, 249,537 bytes, sha256 191ac952 · Apache-2.0 — The trained weights that decide which pixels in your photo are the person and which are the background.

The full licence texts and source links are on the open-source page, and that list is checked against the files actually on disk by the same gate, so it cannot drift into a claim the site does not honour.

What we cannot promise

Browsers change. A WebAssembly build that worked in last month's Safari may misbehave in this month's, and a phone with 3 GB of memory will refuse a picture that a laptop accepts. When something here produces a wrong colour, a broken GIF, or a silent failure, write to the address above with the file type and the browser — those reports are how the pages get fixed, and several of the limits stated on this site exist because a reader hit them first.

The fine print

Tools are provided free and as-is. Browser memory is the practical limit for very large projects; each tool page states its limits plainly. Exports are generated on your device and saved by you — we hold no copies because no copies are sent.