Roman Numeral Converter

Type a number and get the numeral; paste a numeral and get the number. The card splits the answer into its place values, each one as a numeral and as a number — 1994 is M + CM + XC + IV, which is 1000 + 900 + 90 + 4 — and refuses anything that is not strict classic form, because IIII is not four and IC is not ninety-nine. All 3,999 numerals from I to MMMCMXCIX are covered, and nothing above 3,999 is faked: that needs an overline, which plain text cannot draw.

Four place values, three letters each

The thirteen building blocks, and nothing else

Every digit, in every place

Years people actually look up

The current decade

Refused forms, what they look like they mean, and what is correct

Why IIII is not four

Roman numerals are a positional system wearing letters. That is not how they are usually taught, and it is the fastest way to learn them. Break the number into thousands, hundreds, tens and ones; write each place with its own three letters — I V X for the ones, X L C for the tens, C D M for the hundreds, M alone for the thousands; then join the pieces largest first. The nine shapes are identical in every place: 1, 2, 3 repeat the unit; 4 is the unit before the five; 5 is the five; 6, 7, 8 are the five followed by repeats; 9 is the unit before the ten. 1994 is 1000 + 900 + 90 + 4, which is M + CM + XC + IV, and no step of that needs memorising a list.

IIII sums to four and is still not four. This is the objection everybody has, and it deserves a straight answer rather than a rule quoted at it. The strict classic form allows I, X, C and M to repeat at most three times and V, L and D not at all, and it allows exactly six subtractive pairs: IV, IX, XL, XC, CD, CM. IIII breaks the first rule; IC and IL and IM break the second, because I is only ever subtracted from V and X. The point of those limits is uniqueness: with them, every number from 1 to 3,999 has exactly one spelling and every legal string means exactly one number. Without them, 1994 could be written MDCCCCLXXXXIIII or MCMXCIV or a dozen things in between, and a converter could not tell you that your input was wrong because nothing would be.

The clock faces are real, and they are not a counter-example. Public clocks, tower dials and a great many watches have used IIII for four for centuries, usually beside a correct IX for nine. The usual explanations are visual — IIII balances VIII on the opposite side of the dial, and four identical strokes read faster at a distance than a subtractive pair — and the practice is old enough that no one needs to defend it. It is a typographic convention on dials, not an alternative arithmetic, and the same dial almost never writes XXXX for forty. A converter that accepted IIII would have to accept VIIII, XXXX and MDCCCCLXXXXIIII too, and would then be unable to tell a reader that IC is a mistake.

Validation here is a round trip, not a rulebook. When you paste a numeral, the page reads it as a number and then writes that number back out in strict form. If the result is not character-for-character what you typed, the input is refused. That single test enforces every rule at once — the repeat limits, the six legal pairs, the ordering — and it cannot drift out of step with the spelling the page produces, which a hand-written list of rules eventually would. It is also why the refusal is trustworthy: the page is not applying an opinion about your input, it is showing that your input is not what that number looks like.

3,999 is a real ceiling, not a limit of this page. M is the largest letter and it repeats at most three times, so MMMCMXCIX is the largest number the plain letters express. Rome had answers: a vinculum, an overline over a letter multiplying it by a thousand, so an overlined V is 5,000 and an overlined X is 10,000; and the apostrophus, where CIↃ is 1,000 and IↃↃ is 5,000. Neither survives into plain text — an overline is a typographic mark, not a character, and there is no honest way to write it in an input box or a headline. So this page explains both and implements neither, which is better than a half-feature that renders as MMMM or as V with a stray hyphen.

There is no zero and there are no negatives. The system has no place-holder — that is exactly what the positional decimal system brought, and why arithmetic in Roman numerals is so awkward — and no sign. A year 0 does not exist in this notation any more than it does in the Julian calendar. Roman accounting did use fractions, in twelfths, with their own marks, but they are a different notation from the letters and nothing on this page represents them. If you entered 0, a negative number or a decimal, the card refuses it and says why rather than rounding to something.

Some numbers about the numbers. There are exactly 3,999 numerals, one per value from 1 to 3,999, and they are all distinct. 2,047 of them need no subtractive pair at all; 1,536 use one, 384 use two, and 32 use all three places subtractively. The longest is 3,888 — MMMDCCCLXXXVIII — at 15 symbols, and it is the only 15-symbol numeral; the shortest are the seven single letters. 999 values need no M at all. Every one of those figures, and every numeral on this page, was generated from the place-value tables above and then checked by converting it back.

Frequently asked questions

Related calculators

References

1. Wikipedia, Roman numerals (read 30 September 2026), for the history of the standard subtractive form and its variants: the repeat limits, the six subtractive pairs, the vinculum (an overline multiplying by one thousand) and the apostrophus (CIↃ for 1,000, IↃↃ for 5,000), and the long-standing use of IIII on clock and watch dials alongside IX. It is explicit that the strict form is a modern regularisation of a notation that was used inconsistently for centuries — MCMXCIV and MDCCCCLXXXXIIII are both attested shapes — which is the honest framing for a page that refuses one of them.
2. The Unicode Standard, Number Forms block (U+2150–U+218F), which encodes Ⅰ Ⅴ Ⅹ Ⅼ Ⅽ Ⅾ Ⅿ and their lower-case forms as compatibility characters, present for round-tripping legacy East Asian encodings rather than as the normal way to write a Roman numeral. Cited, not reproduced: the block is the Consortium’s copyrighted work. This page follows that advice — it emits Latin M, D, C, L, X, V and I, and it refuses the single-character forms on input rather than silently normalising them, so a numeral pasted from a document that uses them is reported as unreadable instead of half-understood.
3. The arithmetic on this page is not quoted from anywhere. All 3,999 numerals, the symbol counts, the subtractive-pair counts and every table cell were generated from the four place-value tables shown above, and then verified two ways: each numeral was parsed back to its number, and the whole domain was cross-checked against the calculator engine’s own independent greedy algorithm. Every malformed form in the refused-forms table was put through the same parser the page ships.

Unit definitions vary by region and manufacturer. Where a spread exists we show it. For food safety, follow your national authority's current guidance. Full disclaimer at calcengines.com/disclaimer/