Case Converter

Nineteen cases, one tokenizer

Every case convention in programming and writing comes from the same underlying decision: how do you split text into words, and how do you rejoin them? JavaScript's camelCase starts lowercase and capitalises every word after the first. Python's snake_case lowercases everything and joins with underscores. CSS's kebab-case does the same with dashes. SQL constants are traditionally CONSTANT_CASE — upper snake. Go exports a type as PascalCase. None of these are arbitrary: each grew out of what a particular language's parser or style guide would accept, and a developer who works across two or three of them converts by hand, constantly.

The hard part isn't joining words back together — it's splitting them apart correctly in the first place. Most case converters split on every capital letter with a regex, and that breaks the moment an acronym shows up: parseJSONValue becomes the useless parse-j-s-o-n-value instead of parse-json-value, because a regex can't tell "a run of capitals is one acronym" from "each capital starts a new word". This tool's tokenizer treats a run of two or more capitals as a single unit — the acronym stays together, and only the last capital before a lowercase letter starts the next word. That's the difference between a converter that works on real code and one that only works on toy examples.

Four title-case style guides, because they disagree

"Title case" sounds like one rule; it's actually at least four competing ones, and the differences aren't cosmetic. AP style and APA capitalise any word of four letters or more, including prepositions like "About" and "After" — so "A Study About Sleep" is correct AP style. Chicago and MLA lowercase prepositions and articles at any length but capitalise subordinating conjunctions like "because" and "although" regardless of length — so the same sentence becomes "A Study about Sleep". A tool with a single "Title Case" button is right for half its visitors and silently wrong for the other half. This page ships all four, each with the exact rule stated beside the output rather than buried in a settings menu, plus an editable exceptions dictionary for the brand names and acronyms no rule table can guess.

Why "straße" breaks the round trip

Uppercasing German straße gives STRASSE — seven characters from six, because the letter ß has no uppercase form of its own in the traditional mapping and expands to two S's. Lowercase that back and you get strasse, not the original straße. A tool that markets UPPER and lower as exact opposites is quietly wrong the moment German text shows up, and the same trap catches ligatures (file uppercases to FILE, two characters becoming one) and the Greek final sigma (word-ending σ becomes ς, not the same letter it would be mid-word). Every writer-row transform on this page reports a receipt stating whether the result round-trips, and names ß specifically when it's the reason it doesn't.

Why Turkish "I" breaks builds

In Turkish and Azerbaijani, the uppercase of "i" is not "I" — it's "İ" (with a dot), and the lowercase of "I" is "ı" (without one). Call the wrong locale-sensitive case function on an identifier like Title.toLowerCase() using the visitor's browser locale instead of an explicit one, and a build that compiles fine in one office fails in another because the same source text produced two different strings. This tool never reads your machine's locale automatically — the locale option above is explicit, and "Invariant" (the default) uses the same case mapping on every machine regardless of where it's opened.

Graphemes, code points, bytes: not the same number

"How many characters is this?" doesn't have one answer once text stops being plain ASCII. A family emoji (👨‍👩‍👧) is one visible character — one grapheme — but five Unicode code points and more UTF-16 code units than that again once you count the invisible joiners holding it together. The four counters above this box are measured separately and labelled separately, because merging them into a single "characters" figure is exactly the kind of quiet imprecision that eventually breaks a fixed-width column or a VARCHAR limit that was sized off the wrong number.

No headline score

Some case-converter sites print a number claiming to grade a title's "emotional impact" or SEO strength. That number is not measured from anything — it's invented, and printing one we can't verify would be the exact failure this whole site exists to avoid. What you get instead is a receipt: how many characters actually changed, whether the length changed, and whether the transform is reversible. That's a fact about the text you typed, not a guess about how well it will perform.