Test files that tell you what they break.

Free, ready-made files and browser tools for testing uploads, parsers, APIs and the edge cases that break them. Every file has a stable URL, a published SHA-256, and a note on the exact bug it is built to catch.

Browse all test files 220 files, every one verified Generate a file → Any size, any format, on demand

Free, no sign-up. Generated from source, no third-party copyright. Every file carries a published SHA-256.

Start here

Three things people usually need. Pick the one that matches yours.

A clean sample file

A valid, boring file of a given format, to prove the happy path works. Straight to the download.

Sample CSV100 clean rows, 4 KB Sample JSONEvery scalar type, 382 bytes Sample PDFMinimal valid one-pager, 750 bytes Sample XLSXReal workbook with a formula Sample DOCXReal Word document Sample PNG256x256, valid and unremarkable Sample ZIPFlat archive, three entries

A file of an exact size

For upload limits, timeouts, progress bars and chunking. Straight to the download page.

0 bytesThe empty-file edge case 1 MBSmall enough for a quick loop 10 MBThe usual default upload cap 100 MBWhere progress bars start to matter 250 MBThe largest stored file The full size ladder0 bytes to 250 MB, twelve steps Bigger, or any sizeStreamed on demand, nothing stored

A file that breaks something

Deliberately malformed, hostile or awkward, to prove your error handling is real.

Corrupt PDFBroken xref table Truncated PNGValid header, the bytes stop Zip bombSmall file, enormous when opened Zip slipEntry that escapes the target folder Spoofed typeExecutable bytes wearing a .jpg name Reserved filenamesCON, NUL and friends, which Windows refuses Unicode entriesNon-ASCII names inside a ZIP

Need a file that is not here?

The generator streams a file of any size (binary, CSV, JSON, NDJSON, XML, SQL, log, UUIDs or ZIP) straight from a Worker, so nothing is stored and nothing is capped. It also serves slow, truncated and error responses that a static file cannot reproduce.

Open the generator Or build files with CLI commands

Browse by format

One page per format, with every variant of it that exists here and what each one is good for.

CSVDelimiters, quoting, BOMs, ragged rows JSONValid, deep, huge and deliberately broken XMLNamespaces, entities, billion laughs YAMLAnchors, tabs, the Norway problem PDFOne page, many pages, corrupt xref ExcelXLSX and legacy binary XLS WordDOCX, ODT and RTF TextLine endings, encodings, very long lines PNGTransparency, interlaced, truncated ImagesJPEG, GIF, WEBP, BMP, ICO, TIFF ZIPFlat, nested, empty, Unicode entries 7z, TAR, GZIPThe other archive formats Audio and videoMP3, WAV, MP4, and a spoofed MP4 EmailEML, EMLX, MBOX and Outlook MSG SQLiteReal database files with rows in them HTML and webPlus TSV, iCalendar, vCard, RTF GeoJSONAnd KML and GPX FontsTTF, WOFF and WOFF2 EPUBValid, and a broken container Design, CAD and 3DAI, PSD, EPS, DXF, glTF, OBJ, PLY, STL Large files1 KB to 250 MB, and bigger on demand

Test packs

Curated bundles that each answer one question, with a README naming every file and what it should do. All eight packs.

Upload tortureEverything an upload form should survive Filename compatibilityNames that break on one OS but not another Encoding and UnicodeBOMs, UTF-16, normalization, emoji Archive and ZIP securityZip slip, bombs, traversal Document parserPDF, Word, Excel, valid and broken Image pipelineFormats, truncation, spoofed types Structured dataCSV, JSON, XML, NDJSON edge cases Email parserEML, MBOX, MSG, attachments RAG test corpusExtraction traps, with an answer key

AI & LLM evals

Test data for LLM apps and AI agents: what you can download, and what only you can build.

Synthetic data for AI evalsWhat to reuse, what to build yourself RAG test corpusDocuments with extraction traps JSONL for evalsThe format, and ten ways it breaks Mock API for agentsRecords, pagination, failures on demand

Learning

Reference pages for test data and the things that quietly break file handling. The full hub.

Filename compatibilityWhat breaks across Windows, Mac and Linux File systemsPOSIX, NTFS, APFS, ext4, SMB, NFS, NAS File signaturesThe magic numbers that identify a format Character encodingsUTF-8, UTF-16, BOMs, normalization Upload validationThe checklist, with a file for each check MIME typesThe right Content-Type for every extension HTTP status codesEvery code and when you actually see it Dates and timesISO 8601, epochs, Excel serials, DST HTTP headersContent-Type, Disposition, Range, CORS Safe test dataReserved domains, IPs, numbers and cards Regex cheat sheetSyntax, flags and ready-made patterns Crypto and certificatesSSL, PEM, hashing, encryption QA test stringsUnicode, emoji, injection and edge cases Generate files (CLI)Make test files on any OS

Tools

Nineteen small tools, all running in your browser. Nothing you paste or drop ever leaves your machine. All of them on one page.

File signature inspectorWhat a file really is, from its magic bytes File hash and checksumSHA-256, SHA-1, SHA-512, MD5, CRC32 Hash generatorSHA-1, SHA-256, SHA-384 and SHA-512 of text Base64Encode and decode, including base64url URL parserScheme, host, path, query and fragment JSON formatterFormat, validate and locate the parse error CSV to JSONBoth ways, with RFC 4180 quoting CSV inspectorEncoding, delimiter, ragged rows, formula cells JSON Schema test casesValid, boundary and invalid data from a schema Text and JSON diffTwo blocks side by side, line by line Regex testerLive matches and capture groups Cron explainerPlain English, and when it next runs UUID generatorRandom v4 UUIDs in bulk Password generatorStrong random passwords and secrets Fake test dataNames, addresses and cards, reserved ranges only Test image generatorPNG or JPEG at any exact pixel size QR code generatorFrom text or a URL, downloadable Unix timestamp converterEpoch seconds or milliseconds, both ways JWT decoderHeader and claims, decoded in the page

Why this site exists

I am a developer, so I know the drill. You need a 10 MB PDF, a CSV that actually breaks your importer, or a filename Windows refuses to create. Instead you lose twenty minutes to sketchy download sites, sign-up walls, and files you cannot verify. Or you burn AI tokens generating something approximate, and still have to check it yourself.

So I built the thing I wanted and gave it away. Every file is generated from source, carries a published SHA-256, and says exactly which bug it catches.

HEXA is for hexadecimal. That is the whole idea: a QA site that works at the byte level, not the file level. What a file contains is rarely the interesting part. What its bytes say about it, and whether your code agrees, is where the bugs are.

Free, no account. Ads cover the hosting, and that is all they are for. Every tool runs in your browser, so nothing you drop in or paste ever leaves your machine.

Frequently asked

Where can I download a sample CSV file with 1000 rows?

Use the ready-made CSV files or the on-demand generator: /gen?type=csv&rows=1000&cols=5 streams a CSV of any size, free.

What is the EICAR test file?

A harmless 68-byte string every antivirus must detect as if it were malware, so you can safely test that scanning works. It is on the security page (served from a separate host).

How do I get a large test file for upload testing?

Download from the size ladder (0 bytes to 250 MB) or stream any size on demand: /gen?type=bin&size=2gb. Bandwidth is free.

Are these files safe and free to use?

Yes. Every file is generated from source (no third-party copyright), free to download and redistribute, with a published SHA-256 so you can verify it.