About Quinvert

Quinvert converts files between 96 formats. 495 conversions work today, plus 218 unit calculators. It is free, it needs no account, and 423 of those 495 never send your file anywhere.

This page explains how that works, who is behind it, and where the claims can be checked. Everything below is verifiable against the site itself or against the software it ships.

Most conversions run inside your browser

Of the 495 conversions that work today, 423 run entirely on your own device. The decoder and encoder are WebAssembly modules the page downloads once; your file is read by JavaScript, converted in memory, and handed back as a download. It is never uploaded, because there is no upload — the network request that would carry it does not exist.

You do not have to take that on faith. Open your browser's developer tools, switch to the Network tab, and run a conversion. For any of those 423 pairs you will see the WebAssembly module come down and nothing go up. That is a stronger guarantee than a privacy policy, because it is one you can check yourself in about fifteen seconds.

The 72 that do not, and why

Seventy-two conversions run on a server, and every page that uses one says so before you choose a file rather than after. They divide three ways: 36 office-document pairs that need LibreOffice, 26 archive pairs that need 7-Zip, and 10 font conversions that need fonttools. None of the three has a WebAssembly build that would fit in a web page, so for those the file goes to a server, is converted, and the working directory is deleted as soon as the result has been read.

We would rather have 423 honest browser conversions and 72 clearly-labelled server ones than 495 that all quietly upload. The distinction is marked on each conversion page, in the Privacy notice, and in what the software actually does. If a page does not say a conversion runs on our server, it does not — though on seven pages it does not run anywhere yet, which is the next section.

Seven pairs are published and do not run yet

EPUB, MOBI, AZW3, FB2, CBZ, LIT and PDB into PDF have pages on this site and no working converter behind them. They need Calibre, which reads the Amazon and Palm formats LibreOffice cannot, and it has not shipped. Each of those seven pages says so at the top, is excluded from search engines, and offers no drop zone — it is a description of a format and a conversion, not a promise of a tool.

They are named here rather than quietly counted in with the rest because the alternative is a headline figure that includes conversions nobody can run. The honest number is 495 working today out of 502 published, and the seven are the difference.

Who runs Quinvert

Quinvert is run by Valentin Grube, from Germany. The full legal details — name, postal address, telephone number and VAT identification number — are published on the Imprint page, as German law requires of a commercial online service.

That matters more than it sounds. A great many converter sites publish no identifiable operator at all, which leaves a visitor with no one to write to when something goes wrong and no jurisdiction under which to complain. There is a real name and a real address behind this one, and the address on the Contact page reaches a person rather than a ticket queue.

How the pages are written

Every technical claim on this site comes from one place: a registry that records, per format, when it was published, by whom, under which specification, its bit depth, its colour models, whether it supports transparency or animation, which browsers decode it, which software writes it, and what is known to go wrong with it. The conversion pages are written against that registry, and a page cannot state a fact the registry does not hold.

The consequence is that pages will talk you out of a conversion. Converting an MP3 to FLAC restores nothing and makes the file several times larger — the MP3 page says so. Converting a 45-megapixel raw photograph to PNG produces a file few people actually want — that page says so too. A site that recommends every conversion equally is not being helpful; it is selling page views.

What it costs, and how it is paid for

Nothing, and advertising. There is no paid tier, no subscription, no account and no credit system. Larger files on the server-side conversions have a size limit, which each page states, and that limit exists because server time costs money rather than as a lever to sell an upgrade.

What Quinvert does not do is monetise your files. There is no storage, no conversion history, no account to attach anything to, and — for the 423 browser conversions — nothing that ever reaches us in the first place. The Privacy notice sets out exactly what is collected, which is very little, and why.

What Quinvert deliberately does not do

It has no accounts, keeps no history, and stores no files. It does not remove DRM, so a protected ebook or a rented video will not open here whatever the extension says. It does not have a contact form, because a form needs a backend and a backend is the thing this site is built to avoid — there is an email address instead. And it does not claim to be the right tool for professional work that a dedicated application does better: raw photo development, video editing and typesetting all belong somewhere else.

Each of those is a decision with a cost, and each is stated on the page where it bites rather than buried here. A converter that admits what it is not for is more useful than one that claims everything.

The software it is built on

The conversions are done by open source projects — image codecs, a PDF library, a spreadsheet reader, a media toolkit, an OCR engine — and every one of them is named, with its licence, on the Open source page. Nothing here is a proprietary black box that happens to be free.

That list is generated from what actually ships rather than maintained by hand, so it cannot fall out of date with the software. The same is true of the list of conversion formats list: it is built from the registry, which means a format appears there when the code that handles it exists and not before.

Mistakes, and how to report one

Some of what this site says will be wrong. Formats are old, specifications are ambiguous, and encoders behave differently from what their documentation claims. Three published statements were corrected in a single week recently: BMP output was described as keeping transparency when the encoder writes 24-bit colour and cannot; KML to KMZ was called lossless when styles and overlays do not survive it; and a subtitle page was out by a factor of eight on a drift calculation.

Those were found by reading the code against the prose, and they are the reason this page exists in the shape it does. If you find another, the address on the Contact page is the one to use. Corrections get made rather than argued with.