Our Text Entropy Calculator measures the Shannon entropy of any text you paste - a single number, in bits per character, that tells you how unpredictable the symbols in that text are. Score the entire text at once, or switch to per-line or per-paragraph mode to compare different strings side by side, with adjustable decimal precision.

Beyond the headline score, the tool reports how many unique symbols it counted, the theoretical maximum for that set of symbols, and what percentage of that maximum your text reaches. It also offers case-insensitive counting, a letters-and-digits-only mode and a word-level mode for measuring vocabulary variety. Everything runs locally in your browser - nothing you paste is sent to or stored on any server.

This tool is useful in many situations: checking whether passwords, tokens or IDs are as random as they should be; estimating how well a text will compress before running it through gzip; spotting base64-encoded or encrypted-looking blobs inside log files and exports; comparing writing samples for vocabulary richness in word mode; cleaning up data analysis and NLP preprocessing pipelines; and teaching information theory with live examples instead of formulas alone.

Entropy, introduced by Claude Shannon in 1948, measures the average amount of uncertainty - or surprise - in a sequence of symbols. For text, each character is a symbol, and the entropy formula is H = −Σ p(x) · log₂(p(x)), where p(x) is the probability that character x appears in the text. The result is expressed in bits per character: a score of 4 means each character carries, on average, about 4 bits of information.

The extremes are easy to picture. A string where every character is identical, like aaaaaaaaaaaa, has zero entropy - once you have seen the first letter, nothing is uncertain any more. At the other end, entropy peaks when every distinct symbol appears equally often, reaching log₂(N) bits for N different symbols. In between, small examples make good demonstrations: ABAB scores exactly 1.000 bit per character because it uses two equally likely symbols, while ABCD scores 2.000 because four equally likely symbols carry two bits each.

Natural language sits between these extremes. Measured character by character, ordinary English prose lands around 4.0 to 4.5 bits per character - high enough to look varied, but well below the ceiling because some letters (e, space, t) appear far more often than others (z, q). Shannon himself estimated that once you account for how predictable words and phrases are, the true information rate of English is closer to 1 to 1.5 bits per character; the gap is redundancy, and it is exactly why text compresses so well.

Paste or upload your text: Type or paste into the Text to Analyze box, click Load an example for a ready-made demonstration, or use Upload a file to load a .txt, .csv or other text file straight from your device. A live counter below the input box shows the total number of characters and how many distinct characters they include as you type.

Choose a scope: Keep Entire text for one overall score, switch to Each line to score lines individually, or pick Each paragraph to split on blank lines and compare paragraphs.

Adjust the counting rules: Set the Precision from 0 to 6 decimal places, tick Case-insensitive to merge upper and lower case, tick Letters & digits only to ignore spaces and punctuation, or tick Count words instead to measure word-level variety.

Calculate and export: Click Calculate Entropy and the score appears in the result box together with a breakdown in the message below. Use Copy or Download to save the output, or Clear to start over.

Scope: Decides what unit of text gets its own score. Entire text produces a single number; Each line scores every non-empty line and prints the line's own text followed by its score (The quick brown fox jumps over the lazy dog. → 4.488); Each paragraph builds a numbered block separated by blank lines - the paragraph number on the first line, the first 100 characters of the paragraph on the next line (ending in ... when it is longer), and Entropy: with the score in bold on the last line - so every score is easy to match to the text it belongs to. This makes it easy to compare candidates - for example, ranking generated passwords or test strings from most to least random at a glance.

Precision: Controls how many decimal places the score shows, from 0 to 6. The default of 3 keeps results readable - our sample text scores 4.729 - while raising it to 5 shows 4.72907, which helps when two very similar inputs differ only in later digits.

Case-insensitive counting: Merges upper and lower case into one symbol, so T and t count together. On the pangram The quick brown fox jumps over the lazy dog. this lowers the score slightly from 4.488 to 4.442, because the merged alphabet is smaller. Use it when capitalization is meaningless for your comparison, such as plain prose analysis.

Letters & digits only: Ignores spaces, punctuation and all other symbols, scoring just the letters and numbers. The same pangram rises from 4.488 to 4.593 under this rule, since removing the frequent spaces takes away the distribution's biggest skew. It is handy for comparing alphabets without punctuation noise.

Count words instead: Treats whole words as the symbols being measured, giving a vocabulary-variety score in bits per word. For The cat sat on the mat and the cat saw the bird. the exact-case score is 3.022 bits per word across 9 unique words; with Case-insensitive also ticked, the repeated The/the merge into one and the score falls to 2.752 across 8 unique words.

Result breakdown: Alongside the score, the message area reports the number of unique symbols found, how many symbols were counted, the maximum possible entropy for that alphabet (log₂ of the unique count) and the percentage of that maximum reached. For one sample line that reads: Entropy: 4.010 bits per character | 19 unique symbols | 56 counted | max possible 4.248 (log2 of 19) | 94% of maximum.

Load an example: Inserts four contrasting lines - an English pangram, a run of twelve identical letters, a typical Lorem Ipsum sentence and a random-looking password string - and calculates them immediately. With Each line selected they score 4.488, 0.000, 4.010 and 3.585, neatly showing repetitive, natural and high-variance text.

Upload a file: Loads a .txt, .csv, .log or similar text file directly into the input area instead of copy-pasting. The file is read locally by your browser and never uploaded anywhere.

Reference points: 0.000 means every symbol is identical - pure repetition like aaaaaaaaaaaa. Natural English prose typically scores around 4.0 to 4.5 bits per character. Data encoded as hexadecimal can never exceed 4.000 (only 16 possible symbols), base64 tops out at 6.000 (64 symbols), and truly random printable ASCII approaches the ceiling of about 6.57 bits (log₂ of 95 printable characters).

Percentage of maximum: Because the ceiling depends on how many distinct symbols your text actually contains, the percentage shown next to the score is often the fairer comparison. A short password string using 12 different symbols scored 100% in our example - every symbol appeared equally often - while natural prose usually sits between 85% and 95% because common letters dominate.

Two caveats: Shannon entropy ignores order entirely - AAAABBBB and ABABABAB both score exactly 1.000 bit per character even though one looks structured and the other alternating - so treat the score as a measure of variety, not of unpredictability over time. And short samples are noisy: ten characters can score highly by sheer luck, so compare strings of similar length whenever the decision matters.

FAQ 1: Is high text entropy good or bad?
Answer: Neither on its own - it depends on your goal. High entropy is desirable for passwords, keys and tokens, and expected in compressed or encrypted data. Low entropy is normal and healthy for human-readable writing, which needs redundancy to be readable. The tool is best used for comparisons: this token versus that one, this draft versus that draft.

FAQ 2: What score should English text get?
Answer: Counting characters, ordinary English prose lands around 4.0 to 4.5 bits per character. In word mode, expect roughly 5 to 7 bits per word - exact values grow with the length and variety of the text, because a larger sample draws on a bigger vocabulary.

FAQ 3: Does the order of characters matter?
Answer: No. The calculation only counts how often each symbol appears, so AAAABBBB and ABABABAB receive exactly the same score. Two texts with identical character frequencies always get identical entropy, however differently they read.

FAQ 4: Is this the same as password entropy?
Answer: They are related but not identical. Password strength calculators use length times log₂ of the possible character pool - a property of the generation rules. This tool measures the actual distribution of symbols in the specific text you give it, so it can reveal that a supposedly random-looking string repeats characters far more than true randomness would.

FAQ 5: Why did my emoji or accented text change the score?
Answer: Every Unicode symbol counts as one symbol - emoji, accented letters and CJK characters included - and rarer symbols widen the alphabet, raising the maximum and usually the score too. Mixing scripts into a text almost always increases its entropy.

FAQ 6: Is my text stored anywhere?
Answer: No. The calculation runs entirely in your browser - nothing you paste or upload is sent to or saved on any server. Once you close or refresh the page, everything is gone.