Generate random integers, floats or sequences in a range.
Updated
With a seed, the same parameters always produce the same sequence (uses mulberry32 PRNG). Leave unchecked for cryptographic randomness.
crypto.getRandomValues. With a seed, a seeded PRNG (mulberry32) is used so the same seed reproduces the same sequence — useful for tests and teaching.Quick start
Pick a distribution, set its parameters, generate. Everything runs locally — toggle a seed for reproducible results.
Uniform integers, decimals with precision, Gaussian samples around a mean, shuffled 1..N sequence, unique-set lottery, dice or coin flips.
Min / max for ranges, μ + σ for Gaussian, sides for dice. Bulk count up to 100 000.
Inspect the live histogram + summary stats, then copy or download as CSV. Optional seed makes the run reproducible.
In-depth guide
This tool gives you one screen for every common randomness task — uniform integers, decimals, Gaussian samples, shuffled sequences, dice and coin flips — backed by a cryptographic RNG by default and a seeded PRNG when you need reproducibility.
Pick the distribution that matches your use case:
Tick Reproducible mode and type any string. We hash it (FNV-1a) into a 32-bit seed, then drive a small mulberry32 PRNG. Same seed + same parameters = same sequence, every time.
This is invaluable for unit tests, teaching probability, and any workflow where a reviewer needs to re-run your "random" experiment.
For batches ≥ 20 numeric values, we draw a 10-40 bin histogram of the result. Uniform data should produce roughly equal bars (with sampling noise); Gaussian data shows a bell curve centered on μ.
If your uniform histogram has a visible bias, your sample size is just too small — the underlying CSPRNG / mulberry32 are statistically uniform; bin counts converge as you generate more.
Three quality-of-life sub-modes:
crypto.getRandomValues; with a seed, a deterministic PRNG in the same tab. No network requests fire when you click Generate.This makes the tool safe for sensitive uses: jury allocation, internal raffles, code-review pull-request sampling — none of it touches a server.
Yes, by default. Without a seed, every value comes from crypto.getRandomValues — the browser's CSPRNG. Only the optional 'Reproducible mode' uses a seeded PRNG (mulberry32).
Turn on Reproducible mode and provide any string. The same seed + parameters always produces the same sequence, which is handy for tests, debugging and teaching — at the cost of not being unpredictable.
Up to 100 000 values at once. Everything runs locally so there is no API rate limit; the only cap is your browser's memory for the result array and download.
Box-Muller transform on two uniform-(0,1) samples. The mean (μ) and standard deviation (σ) controls let you target any normal distribution.
That is genuine sampling noise. Even truly random samples cluster a little — bin counts only smooth out as the batch grows past a few thousand values.
Yes — pick 'Unique set' mode. The result contains 'Count' distinct integers in [min, max] (count must not exceed the range size).
For everyday lotteries, raffles and games — yes. For published scientific studies, prefer documented seeded runs (so reviewers can reproduce) or hardware RNGs (random.org, /dev/random).
Keep exploring
Hand-picked utilities that pair well with the one you're on — all free, client-side, and zero-signup.
Shrink PDF file size without uploading to a server.
Combine multiple PDFs into one in your browser.
Split a PDF by pages or page ranges, download as zip.
Convert JPG, PNG and WebP into a single PDF, locally.