Estimated entropy 0 bit
No password

Exhaustive search at 100 billion guesses/s:

Generator settings
16

Passwords are generated locally in your browser with the Web Crypto API and never leave your computer.

RAW Tools password

Strong Password Generator

Robert Eisele

Password Generator creates strong, random passwords entirely inside your browser. It uses the Web Crypto API (crypto.getRandomValues) instead of Math.random(), so the password never touches the network and is never predictable from timing or seed leaks.

Pick a length and character classes to generate a password, or edit the result and enter one of your own. The entropy scale and crack-time estimate update with every change. Generated passwords use the exact configured search space; manually entered passwords use a clearly marked estimate based on their detected character classes.

What Actually Makes a Password Strong

Most password advice repeats the same handful of rules. Here is the condensed, non-redundant version:

The Math Behind the Crack-Time Estimate

The tool's crack-time indicator is not a lookup table, it is computed from the exact settings you chose. Let b be the size of the character pool (digits contribute 10, uppercase and lowercase 26 each, symbols 32, so the full default pool is b = 94) and n the chosen length.

If repeated characters are allowed, the number of possible passwords is bn, which corresponds to an entropy of

\[ H = n \cdot \log_2 b \quad \text{bits.} \]

With Exclude Duplicate enabled, each character can only be used once, so the count follows a falling factorial instead:

\[ b \cdot (b-1) \cdot (b-2) \cdots (b-n+1), \qquad H = \sum_{i=0}^{n-1} \log_2(b-i). \]

Because every factor in that product is at most b, excluding duplicates can only ever lower the entropy compared to allowing repeats for the same pool and length, never raise it. The difference is small for short passwords and grows for longer ones as the pool gets used up.

For manually entered passwords, the original generation process is unknown. The tool therefore estimates b from the character classes that occur in the text and displays the result as an estimate. This describes the size of a corresponding random search space, not the unpredictability of human choices: dictionary words, dates, repeated fragments, and keyboard patterns can be cracked much faster than that number suggests.

The tool converts the resulting bit count into a crack time by assuming an attacker capable of \(10^{11}\) guesses per second (100 billion per second, a realistic upper bound for offline brute-forcing weakly-hashed data) and dividing the number of possible passwords by that rate. A few concrete results from that formula:

Exclude Similar removes the visually ambiguous characters 0/O, 1/l/|, and 2/Z from the pool, shrinking it from 94 to 87 characters. That is a small entropy cost, useful when a password has to be typed or read out loud rather than pasted from a manager.

FAQ

Is it safe to use an online password generator?

It is safe as long as the password is generated locally and never sent anywhere, which is exactly what this tool does: everything happens in your browser via the Web Crypto API, with no network request involved in creating or displaying the password.

Password or passphrase, which is better?

Both are just points on the same entropy scale described above. A password mixes character classes into a shorter string; a passphrase spreads the same or more entropy over more, easier-to-type characters. Use whichever you can type and store reliably at the length your account allows.

If a service is breached, can my hashed password still be cracked?

It depends entirely on how the service hashed it. A fast, unsalted hash makes offline cracking with precomputed rainbow tables or brute force realistic. A slow, salted hash designed for passwords, such as bcrypt, scrypt, or Argon2, makes the same attack orders of magnitude slower per guess. You cannot control which one a service uses, which is exactly why password length and uniqueness are your side of the defense.

How do attackers actually get passwords in practice?

Rarely through brute force against a live login form, that is usually rate-limited. The common paths are phishing pages that simply ask for the password, credential stuffing with combinations leaked in unrelated breaches, and malware that logs keystrokes or reads a browser's saved-password store. A strong, unique, generated password defeats credential stuffing and offline cracking, but not phishing, that is what two-factor authentication is for.

Are password managers safe to rely on?

A reputable password manager with strong end-to-end encryption is safer than the realistic alternative, which is reused or written-down passwords. It concentrates risk into one master password, so that one should be long, unique, and protected with two-factor authentication.

Do quantum computers break these passwords?

Grover's algorithm gives a quantum computer a quadratic speedup against brute-force search, which in practice halves the effective bit strength of a password rather than eliminating it. A password with 103 bits of classical entropy would still correspond to roughly 51-52 bits against such an attack, and no hardware anywhere near that capability exists today.

What do genuinely weak passwords look like?

Short strings, keyboard sequences, and predictable substitutions repeatedly top real-world breach statistics: 123456, password, qwerty, 123456789, and names or dates tied to the account owner. All of them are near the bottom of the entropy scale described above, regardless of how they are formatted.