Letter Frequency Analyzer

Paste any text and see a per-character histogram: count, percentage, and a bar per row. Toggle which character classes to include (letters / digits / whitespace / punctuation / other Unicode). Case-folding optional. When letters are enabled, a side-by-side comparison against the standard English letter distribution (E 12.7%, T 9.1%, A 8.2%…) is shown — handy for classical cryptanalysis, language identification, font coverage planning, or just nerd curiosity.

The English reference distribution comes from large bodies of ordinary prose, which is the assumption to watch. A technical document, a list of names, source code or anything with heavy jargon departs from it enough that the comparison stops meaning much, and a short sample departs from it by chance alone. Cryptanalysis by frequency wants a few hundred characters at minimum before the ordering of the common letters settles.

That comparison is the classical attack on a substitution cipher, and it works because a monoalphabetic substitution moves the peaks around without changing their heights. It is defeated by anything that flattens the distribution: a Vigenère cipher with a key of unknown length, a homophonic cipher that gives common letters several symbols each, and every modern cipher, whose output is indistinguishable from random by exactly this test.

Counting is per Unicode code point rather than per glyph. An accented letter typed as a base plus a combining mark counts as two entries and lands under both, while the same letter typed as a single precomposed code point counts as one, so two visually identical texts can produce different tables. Normalising the input first is the fix. Emoji built from several code points split the same way.

The character classes are worth switching off deliberately. Leaving whitespace in puts the space at the top of nearly every text, where it crowds out the letters you were looking at.