Test Data Generator
Generate names, emails, usernames, UUIDs, IP and MAC addresses, Luhn-valid card numbers and more — seeded so the same seed always yields the same rows, entirely in your browser.
About this tool
Test data has a property almost every random generator throws away: it has to be reproducible. A bug that only shows up with a specific row of the 20 you pasted yesterday is worthless if those 20 rows are gone forever. So this tool is built around a seed: the same seed with the same fields produces exactly the same rows — today, next month, on your machine and on your colleague’s. Write the seed into the bug report ("checkout-bug-47") and anyone can regenerate the evidence.
Every format is locked to its public standard rather than to "looks about right". UUIDs are version 4 with the correct version and variant bits (RFC 9562). Card numbers pass the Luhn checksum (ISO/IEC 7812-1) so they survive form validation, yet correspond to no real issuer range. Emails use the domains RFC 2606 reserves for documentation, so test mail can never be delivered to anyone. Country codes are ISO 3166-1 alpha-2, IP addresses avoid the network and broadcast addresses of each /24, and MAC addresses carry the unicast bit set correctly.
Everything runs in your browser. There is no server, no request and no logging — the rows are produced by plain arithmetic on the seed, which is also why it works with the network switched off and why the same seed cannot help but give the same answer.
Why the order you pick fields does not change the values
Fields are sorted into a fixed catalogue order before anything is generated, so asking for name, email, uuid gives the same values as email, name, uuid. Your selection order only reorders the columns, never the data.
If generation followed the order in the request, the random stream would be consumed in a different sequence and every value would change — same seed, same fields, different table. That breaks the seed promise in the one case where it matters most: a colleague rebuilding your data from a bug report and ending up with something else to look at.
Sorting into a canonical order before the loop removes that entire class of surprise. The field list you see is in catalogue order, the columns come out in the same order, and each value is a pure function of the seed and the set of fields selected.
It also means the seed alone is not enough to reproduce a table; the field set is part of the input. Two runs with the same seed and different fields share a starting point and diverge as soon as the first extra column is drawn.
Where the email and username values come from
Both are derived from the name in the same row, so [email protected] belongs to Ada Smith. When name is not selected, a name is still drawn for that row and simply left out of the output.
A generator that invents a name and an address independently produces rows that are obviously fake to anyone reading them — a name with no relationship to the address above it. Deriving the address from the row keeps the table coherent, which matters as soon as the data is put in front of a person.
The internal name is drawn whether or not you asked for it, which keeps the random stream aligned: a table of names and emails is the same table you would get from emails alone, minus the name column. That follows from drawing from one sequence rather than re-seeding per column.
The hidden value lives in a different object from the row that gets printed. An earlier version shared a single object and an internal field leaked into the output; the rows you see now contain only the fields you selected.
How a text seed becomes a number the generator can use
The seed is hashed with FNV-1a into a 32-bit integer, so any text works — order-service-v2 is as valid a seed as 42. A seed written purely in digits is used as that number directly rather than hashed.
The point of accepting text is that seeds end up in bug reports, and "checkout-bug-47" tells a reader far more than a ten-digit number does. Hashing is deterministic and platform-independent: the same string produces the same integer wherever the page runs, which is what the reproducibility promise rests on.
Numerals are treated as numerals rather than hashed, so 42 means 42 and not the digest of the two characters "4" and "2". Values inside the 32-bit range are used as they stand; anything longer, or containing a non-digit, falls back to hashing.
The randomness is the exception rather than the default. Left empty, the seed becomes a random value and the tool shows you which number it chose, so the run can be reproduced later by typing that number back in. Randomness is disclosed; it is not hidden inside the generation.
Why fake card numbers have to pass Luhn
A fake card number that fails the Luhn check is rejected by a payment system at the first gate. So your test exercises the rejection path, not the happy path — and it goes green while the code that actually charges a card never runs.
Luhn is simple: from the rightmost digit, double every second digit (subtracting 9 when it exceeds 9), then sum everything; the number passes if that sum divides by 10. The last digit of a real card exists precisely to satisfy it. Generate 16 digits at random and roughly one in ten passes — meaning nine in ten "card numbers" this tool could produce naively would be format-invalid while looking perfectly plausible.
That difference is fatal in a test suite. An invalid check digit makes the downstream validator return early, so the branch that actually processes a payment is never executed; the suite is green and production breaks. This is what "realistic test data" actually means — not that it looks like a card number, but that it survives the checks a real system performs.
The same reasoning applies to the other fields here: ISBN-10 and EAN-13 carry their own check digits; IBAN has its MOD-97 check; a UUIDv4 must set the 13th hex digit to 4 and the 17th to 8, 9, a or b (the version and variant fields); and the first octet of an EUI-48 MAC address has meaning in its lowest two bits, multicast and locally-administered. Each of these is pinned to a published standard rather than invented here.
Frequently asked questions
What exactly does the seed control?
Everything. Each generated value is a pure function of the seed and its position in the sequence. Change the seed and every row changes; keep it and you can regenerate the identical table forever. An empty seed picks a fresh random one each session and shows you which number it used, so even "I did not care about the seed" is reproducible.
Are the card numbers dangerous?
No. They satisfy the Luhn checksum, which is a 1954 arithmetic test over the digits themselves — it says nothing about whether an account exists. The generated numbers start with 4 but are not inside any real issuer’s allocated range and cannot be charged. They exist so a payment form’s format validation can be exercised end to end.
Why do generated emails use reserved example domains?
RFC 2606 sets aside a block of example domains for documentation and testing, so mail sent to them can never reach a real inbox. A generator that emits addresses at a live consumer mail provider is not producing test data — it is producing misdirected mail and the occasional very confused recipient.
How random are the passwords?
Each password is 14 characters and guaranteed to contain at least one lowercase letter, one uppercase letter, one digit and one symbol, with the rest drawn uniformly from the combined pool and the whole string shuffled. They are generated from the seed, which means they are only as secret as the seed — fine for filling a signup form in a test run, never for real accounts.
Does this send my data anywhere?
No. Generation is plain arithmetic in the page — there is no server, no request and no telemetry, and the tool keeps working with the network switched off. The clipboard buttons write directly from the page to your clipboard.
Related tools
Paste JSON, YAML or TOML and get the other two back, with syntax errors pinned to the exact line and column. Runs entirely in your browser — nothing you paste is uploaded.
Test a regular expression against your own text, see every match and capture group highlighted, get each piece of the pattern explained, and check what JavaScript, PCRE and RE2 actually support. Everything runs in your browser.
Work out the network address, broadcast address, host range and mask of any IPv4 or IPv6 CIDR — and check whether an IP belongs to it. Nothing leaves your browser.
Paste two name lists from Excel, Sheets or CSV and see what matches, what is unique to each side, and which entries are duplicates or near-misses. Nothing leaves your browser.