Define the fields you need and generate as many rows of plausible test data as you want, as JSON, CSV or a ready-to-run SQL INSERT. The part that makes it actually usable is the seed: generation is deterministic, so the same seed always produces the same rows. Without that, regenerating to add one field reshuffles everything else, and the dataset you were half-way through checking changes underneath you. Emails carry the row number too, because fixtures often exist precisely to exercise a unique constraint and duplicate addresses defeat that without any obvious sign. Twenty-odd field types cover names, contact details, places, numbers, dates and text. Nothing is uploaded, and nothing is generated on a server — which also means there is no row limit beyond what your browser will hold.
| Format | Shape | Use it for |
|---|---|---|
| JSON | Array of objects with real types | Fixtures, API mocks, seed scripts |
| CSV | Header row, RFC 4180 quoting | Spreadsheets, import testing |
| SQL | One multi-row INSERT | Seeding a database directly |
It makes generation reproducible. The same seed always produces the same rows, so you can regenerate after adding a field without every other value changing. Anything works as a seed — a word, a ticket number, a date — and changing it gives you a completely different dataset.
Yes. Each carries its row number, so a thousand rows give a thousand distinct addresses. Fixtures frequently exist to test a unique constraint, and a generator that quietly repeats addresses makes that test pass for the wrong reason.
It is entirely invented — names drawn from word lists, addresses assembled from generic streets and cities, and emails on reserved example domains. None of it refers to a real person. It is test data, though, not anonymised production data; the two are not interchangeable and only one of them is safe to share.
A single multi-row INSERT rather than one statement per row — far quicker to execute and much shorter to paste. String values escape quotes by doubling them, which every major engine accepts. Set the table name before copying.
Up to 5,000 at a time. The limit exists so the page stays responsive rather than because of any server cost; for more than that, generate in batches with different seeds.
No. Generation happens entirely in your browser. There is no API call, no account and no quota — the data is created on your machine and stays there.
Values come from a seeded mulberry32 generator rather than Math.random, which cannot be seeded at all. That matters for a fixture tool: without reproducibility, regenerating to adjust one field reshuffles every other value, so a dataset you were half-way through checking changes underneath you. Any text works as a seed — it is hashed with FNV-1a into a 32-bit value — so a memorable word gives the same rows every time. Emails carry the row index, because a fixture's whole purpose is often to exercise a unique constraint and duplicate addresses defeat that silently. The spec suggested Faker and flagged its bundle size in the same breath; it was skipped. Faker ships hundreds of kilobytes even tree-shaken, and twenty-odd field types need a handful of word lists and a PRNG — which also makes the generation testable rather than a wrapper around someone else's randomness. Output is emitted as JSON, CSV with RFC 4180 quoting, or a single multi-row SQL INSERT, which executes far faster than one statement per row and is much shorter to paste; SQL strings escape by doubling quotes, the portable form across engines.