Password Generator
ListeningInbox
0Cryptographic password
crypto.getRandomValues(). Nothing is sent to a server or written to disk.Secret key or otpauth URI
Paste a Base32 setup secret (spaces ignored) or an otpauth://totp/... link.
Advanced parameters
Generate a fictional identity
State-synced area codes, deterministic SVG avatar, online credentials, Finance & IDs, and one-click Temp Mail handoff.
Generate test card numbers
Builds structurally valid PANs with genuine issuer prefixes and check digits for payment sandbox testing.
Check Luhn validity
Published gateway test numbers (click to copy)
When testing exact gateway states (3D Secure, declines, international failures), use the provider’s deterministic test PANs rather than random numbers.
Cryptographic password and passphrase generation guide
This generator builds high-entropy passwords and multi-word passphrases locally in your browser using window.crypto.getRandomValues(). Unbiased rejection sampling guarantees every selected character pool appears at least once without sending a single character over the network or writing to disk.
Why rejection sampling beats Math.random() and modulo shortcuts
Standard Math.random() is predictable and unsafe for credentials. This generator reads 32-bit unsigned integers from crypto.getRandomValues() and discards out-of-range draws above the largest multiple of the character set size so modulo bias cannot skew output.
At 16 mixed characters drawn from all 86 characters (~6.43 bits per character), entropy reaches about 103 bits — comfortably beyond offline brute-force capability.
The 86-character pool and look-alike filtering
Four pools combine to 86 characters: 26 lowercase letters, 26 uppercase letters, 10 digits, and 24 symbols. Enabling a pool guarantees at least one character from that pool in the output.
Enabling Avoid look-alikes removes six confusing glyphs (I, l, 1, O, 0, o), trimming the pool from 86 to 80 characters (−2 bits at length 16, fully recovered by adding one character).
Choosing between a random password and a passphrase
- 16–24 character random passwords are best whenever pasting into a password manager.
- 10-word passphrases provide roughly 75 bits from the 185-word pool (7.53 bits per word), matching a 12-character mixed password while remaining readable.
- Enable Avoid look-alikes when typing from a screenshot or reading credentials aloud.
Crack-time model and real-world limits
The crack-time estimate assumes an offline attacker trying 1012 (one trillion) guesses per second against an unsalted fast hash. Salted memory-hard KDFs such as Argon2id run orders of magnitude slower.
Reusing a password across sites defeats any length the moment one site breaches; always generate a unique credential per account.
| Mode | Configuration | Approximate entropy | Half-keyspace guess time | Recommended use |
|---|---|---|---|---|
| Password | 8 chars (mixed 4 sets, 86 pool) | ~51 bits | 25 minutes | Below floor — avoid for primary logins |
| Password | 12 chars (mixed 4 sets) | ~77 bits | ~2,600 years | Minimum baseline for standard web accounts |
| Password | 16 chars (mixed 4 sets) | ~103 bits | 1.4 × 1011 years | Default for primary logins & password vaults |
| Password | 20 chars (mixed 4 sets) | ~129 bits | 7.8 × 1018 years | High-value encryption keys & master secrets |
| Passphrase | 5 words (185-word pool) | ~38 bits | Under 1 second | Short-lived test logins only |
| Passphrase | 8 words + separator | ~60 bits | ~8 days | Local device codes or spoken temporary access |
| Passphrase | 10 words + separator | ~75 bits | ~744 years | Memorable human-typed credential (~12-char equiv.) |
| Mistake | Why it defeats length | What to do instead |
|---|---|---|
| Reusing a password across services | One third-party breach compromises every account sharing that credential | Generate one distinct password per site inside a manager |
| Shortening length to satisfy symbol rules | Removing digits or symbols loses ~0.18 bits/char; shortening length drops >6 bits/char | Keep 16+ characters even when a site restricts symbols |
| Tweaking random output with names or years | Predictable patterns slash effective randomness immediately | Keep pure CSPRNG output untouched |