Case Converter

Fixing a heading that arrived in all caps, or a title someone typed entirely in lowercase, is a small job that most converters make slightly worse. They give you a row of unlabelled buttons that rewrite the box you are working in, so the only way to find out what a case does is to press it and lose what you had. Here each case previews itself against your own text before you choose, and the conversion lands in a separate box, so the original is still there when the result is not the one you wanted.

Seventeen cases sit in two groups. For prose: UPPERCASE, lowercase, Title Case, Sentence case, Capitalise Each Word, iNVERT cASE, aLtErNaTiNg cAsE and a random case you can reproduce from its seed. For identifiers: camelCase, PascalCase, snake_case, CONSTANT_CASE, kebab-case, Train-Case, dot.case, path/case and plain space case. The identifier cases work a line at a time, so a list of fifty entries comes back as fifty identifiers rather than one long joined string.

Title Case follows the convention writers actually use rather than capitalising indiscriminately: short words like of, the and by stay lowercase in the middle of a title, while the first word, the last word and the word after a colon always get a capital. The list of short words is yours to edit. Sentence case is similarly careful — it keeps the pronoun I, capitalises through an opening quote, and knows that e.g. and etc. do not end a sentence, so it will not capitalise the word that follows them.

The most useful thing either of them does is leave alone what is already deliberate. NASA, HTML, iPhone and macOS all carry meaning in their capitals, and a converter that flattens them to Nasa, Html, Iphone and Macos has destroyed information you cannot get back. Words that are all capitals, or that carry a capital after their first letter, are preserved by default — and for the proper nouns no rule could guess, an exceptions list forces an exact spelling that wins over everything else.

Case conversion is unusual among text operations in being mostly one-way. Sorting a list can be re-sorted, and a replacement can be replaced back, but casing throws information away: once a heading has been uppercased, nothing in the text records that it used to read "the iPhone by NASA" rather than "The Iphone By Nasa". A tool that transforms the box you are working in therefore destroys something on every click, and two clicks in a row destroy it beyond recovery.

So the input is left exactly as you typed it and the result appears separately, updating as you edit or change any option. When you do want to build on a result — convert to space case to break apart an identifier, then to Title Case to make a heading of it — "Use as input" moves it up deliberately, and undo covers that as it covers loading a file or clearing the box.

Sentence case is where this matters most, because it is the conversion that reaches furthest into the text. Its lowercasing pass can also be turned off entirely, in which case only the first letter of each sentence is touched and every other character is left as you wrote it. That is the right setting for prose that is nearly correct already and only needs its sentence starts fixed.

Turning a phrase into an identifier means deciding where each word begins, and that decision is where most converters go wrong. Spaces, underscores and hyphens are always boundaries, and a capital letter always starts a new word — but three cases need a judgement call, so each is a switch you can see.

A run of capitals is the first. In XMLHttpRequest the boundary sits between the L and the H, not before the whole run, so splitting runs of capitals yields xml_http_request rather than xmlhttp_request. Numbers are the second: utf8Encode can reasonably become utf_8_encode or utf8_encode depending on whether you consider the 8 part of the word, so you choose. Note that a capital after a digit always starts a word either way, so Encode is never absorbed into what precedes it.

Dots and slashes are the third and the most consequential, because they are separators in a dot.case identifier and ordinary characters in a domain name or a file path. They are off by default, which keeps example.com and src/components intact, and a note appears when your text contains one so the choice is never silent. A dot between two digits is never treated as a boundary at all, so 3.14 stays a single number whatever you pick.

Every major style guide defines title case differently, and none of them capitalises every word. The common thread is that short function words — articles, coordinating conjunctions and short prepositions — stay lowercase unless they open or close the title. Where guides disagree is the threshold for prepositions: some lowercase everything under four letters, others under five, and their treatment of subordinating conjunctions varies. Rather than claim an authority it cannot honour, this tool gives you the word list and lets you match whichever guide you answer to.

Proper nouns are the harder limit. Deciding that Paris is a city and not a word needs a dictionary, so no rule-based converter can be trusted to lowercase a body of text without damage. Two things narrow the gap. Capitals that already look deliberate are kept, which covers acronyms and names styled like iPhone. And the exceptions list handles the rest: put GitHub or PostgreSQL in it once and every conversion spells them that way, overriding every other rule.

Two notes on languages. Turkish and Azerbaijani have a genuinely different mapping for the letter i — the capital is İ, with the dot retained, and a dotless ı exists as a separate letter — so a converter that is not told the language corrupts Turkish text. Pick the language and the conversion follows it. German ß has no traditional single-character capital and historically expands to SS, which means an uppercase round trip does not always return what you started with; that is a property of the alphabet, not a bug in the tool, but it is worth knowing before you convert and paste back.