Files, Binary & Compression
Choose or drop one or more files; each is identified from its first bytes the moment it is added.
Drop files here, or use Choose files. Up to 20 files per batch; any size.
Only the start of each file (and the end of a ZIP) is read, locally in your browser. Nothing is uploaded.
Choose or drop files to identify them.
Only changes how much of the header is displayed; nothing is read again. Signatures found further in (tar, ISO, PE header) are shown separately.
Choose or drop files to identify them.
The report is plain text: one block per file with its size, verdict, MIME, extension check and header bytes.
You need a file type checker when a name stops being trustworthy: an attachment with no extension, a download that will not open, an upload form that must reject a script renamed to photo.jpg. The extension is only a label someone typed; the first few bytes of a file — its magic number or signature — are written by the program that created it.
This file type checker reads those bytes locally, names the format and its MIME candidates, and compares the result with the extension. It also says what the verdict rests on — an exact signature, a checked structure such as the file names inside a ZIP, or a text guess — and highlights the bytes that matched, so you can verify the answer instead of taking it on faith.
A verdict you can verify byte by byte, with the extension checked against it.
The header is shown in hex and ASCII with the matched signature highlighted — 89 50 4E 47 for PNG, 50 4B 03 04 for ZIP — and signatures deeper in the file, like tar's ustar at offset 257, are shown with their offset.
A PNG renamed to .jpg is flagged as a mismatch with the expected format for .jpg spelled out, while a .docx that is correctly a ZIP container is marked compatible instead of wrong.
Choosing or dropping files identifies them immediately, and new files are added to the list rather than replacing it. Only the first 64 KiB of each file is read, plus the last 64 KiB when it starts like a ZIP, so a 20 GB disk image is as quick as a 20 KB icon.
Exact signature means fixed bytes at a fixed offset matched. Signature plus checked structure means an inner field was also read — the file names inside a ZIP, the brand in an MP4 ftyp box, the RIFF subtype. Text or pattern guess means no fixed magic exists, as with JSON, XML or plain UTF-8.
Compatible is not an error: .docx, .jar and .apk are all ZIP files and .m4a is an MP4 container. Mismatch is the one to look at. The browser MIME row shows what File.type reports — it comes from the extension, which is exactly why it cannot be trusted for validation.
Images, PDF and office documents, archives and compression formats, executables (ELF, PE, Mach-O, WebAssembly), audio and video containers, fonts, databases and capture files, plus text formats with a clear opening marker such as XML, SVG, HTML and PEM.
Reading only the file names in the ZIP central directory, it separates DOCX, XLSX, PPTX, ODT, EPUB, JAR, APK and IPA from a plain ZIP — without decompressing a single entry.
Match, mismatch, compatible, no extension and cannot compare — so a correct .docx is not reported as a wrong .zip, and a missing extension gets a suggested one.
Each file shows its exact byte count next to KiB, MiB or GiB. There is no size limit, because the cost of identification does not grow with the file.
Drop it into the file type checker above: identification never uses the name, only the first bytes, so a missing extension changes nothing except that the extension check reads "no extension" and a suggested extension is shown.
Because a .docx really is a ZIP archive. Here the file names in its central directory are read, so a normal .docx is reported as a Word document; if those names cannot be found, it stays ZIP and the .docx extension is marked compatible, not wrong.
Do not rely on File.type: browsers fill it in from the extension, which the browser MIME row here shows side by side with the detected type. Read the first bytes with file.slice(0, n).arrayBuffer() and compare them with a signature, which is what this page does.
It is a good first filter against renamed files, not a guarantee. Magic bytes only cover the start of a file, so a crafted polyglot can pass. Servers should still re-encode or fully parse images and documents, limit size, and scan when it matters.
file somefile, which uses the libmagic signature database, and file --mime-type for just the MIME type. This page is a file type detector built on the same idea, with a smaller, documented table, and adds the extension comparison and highlighted bytes.
Many do. This one does not: files are read with the browser's File API in this tab, only the header (and the end of a ZIP) is read, and the report you copy contains only names, sizes, verdicts and header bytes.
Still have a question?
Load the examples to see a match, a mismatch, a ZIP subtype and an unrecognised file side by side, then point the file type checker at the files you actually doubt.
Everything runs locally in your browser.