A file corrupter built by people who needed one
Files Corrupter started as an internal testing utility. We kept needing deliberately broken files, a truncated ZIP to test an unpacker, a PDF with a destroyed xref to check an error message, a video with no index to see what a player would do, and every online tool we tried had the same two problems.
The first was privacy. Almost every corrupter uploads your file to a server. That is a non-starter for anything confidential, and it is an odd thing to ask for when the entire operation is a byte transform a browser can do locally in milliseconds.
The second was that they did not work reliably. Flipping fifty random bytes in a DOCX usually lands inside a compressed stream, and Word recovery opens the document anyway. The same trick on a PDF gets quietly repaired by Acrobat. A tool that produces a file which still opens is not a file corrupter, it is a random number generator with extra steps.
What we built instead
The engine identifies the format from its extension and magic bytes, then applies a profile tuned to that format structure. For a ZIP container it destroys the local file headers, the central directory and the end-of-central-directory record. For a PDF it takes out the header, every xref section, the trailer and the EOF marker. For MP4 it destroys the ftyp brand and the moov atom. For an ISO it clears the entire 32 KB system area before the volume descriptors. Only then does it scatter entropy through the payload.
All of it runs in your browser. There is no backend, no database, no log file with your filename in it, and nothing to breach. You can confirm this yourself by watching the network tab, or by turning off your internet connection and using the tool anyway.
How this stays free
Because it costs us almost nothing to run. There is no per-file compute, no storage and no bandwidth beyond serving a few hundred kilobytes of static assets. There is no paid tier, no feature held back and no plan to add one.
Where we stand on responsible use
The overwhelming majority of use here is ordinary engineering work: testing error paths, rehearsing restores, building fuzzing corpora, teaching. We are aware the tool has an obvious second use, and we are not going to pretend otherwise, but we are not going to encourage it either. Submitting a deliberately damaged file as completed work is dishonest, it is usually detected, and the consequences under an academic integrity policy are considerably worse than asking for an extension. Our disclaimer spells this out.
Get in touch
Bug reports, format requests and questions are all welcome. The contact page is the fastest route.