Speaking Time Calculator

About this tool

Writing a script and reading it out loud are two different skills, and the gap between them is where most timing goes wrong. A page of text looks short; spoken at a comfortable pace it can easily run three minutes. This calculator closes that gap by converting between the two, in both directions.

The first thing it refuses to do is pretend that Chinese and English are measured the same way. Chinese is counted in characters per minute — a news reader moves at roughly 300 of them — while English is counted in words per minute, where 140 is a normal reading pace and news readers actually slow down rather than speed up, because clarity matters more than rate. Applying one number to both languages gives an answer that looks plausible and is wrong by a wide margin. So the two counts are kept separate and mixed scripts are counted on both sides at once.

The second thing it refuses to ignore is punctuation. Pauses are not free: a full stop costs about four tenths of a second, a comma about half that, and a paragraph break longer still. On a two thousand character script that is fifteen seconds or more of silence, and the error grows with the length of the script — which is exactly when you need the estimate to hold. Every figure here includes a pause budget, and the breakdown table shows how much of the total it accounts for.

The backward direction matters just as much. If you are producing a sixty second video, the useful question is not "how long is this script" but "how much script fits in sixty seconds" — and the answer has to subtract the pauses first, otherwise you write to the optimistic number and overrun in the recording booth.

Why numbers are timed digit by digit

A spoken number is not one word. 2026 is read as four separate parts, so the tool times it per digit at about three tenths of a second each, and a four-digit number costs over a second of speech on its own.

Counting words alone would badly underestimate a script full of figures. A date, a price or a phone number is spoken as a sequence of digit sounds, not as a single token, and the longer it is the further the two measures drift apart.

The digit count is kept separate from both the character count and the word count, so the results card shows exactly how much of the time goes into numbers. That matters most for scripts that are mostly figures — a financial summary or a statistics readout — where the pause budget and the digit budget together can dominate the total.

The rate per digit is a single constant rather than a language-by-language table. It is deliberately coarse: how an English three-digit number turns into words varies by convention, and the gain from modelling that is small next to the complexity it would add.

Punctuation inside a number is not a pause

Decimal points and separators inside numbers are masked before punctuation is counted, so 3.14 is not read as a sentence ending, 1,000 as a comma pause, or 12:30 as a clause break.

Taken literally, those characters are exactly the ones the timer counts as pauses. A time like 12:30 would add a clause pause, a price like 1,000 a comma pause, a version number like 3.14 a full sentence stop — inflating the estimate for text that has no intention of pausing there.

The masking only affects the punctuation count. The original text is untouched, the digit count still sees the full run of digits, and the number is still read as the digits it contains. What is removed is only the mistaken silence.

It is a small piece of the calculation that would stay invisible until it was wrong: on a script full of figures the overcount is large enough to change the answer noticeably, and nothing in the output would explain where the extra seconds came from.

What the backward estimate assumes about punctuation

Working back from a target length, the tool has no script to count commas in, so it assumes one comma every twenty units in Chinese and every twelve in English, and a full stop at five times that spacing.

The forward direction counts the punctuation actually present. Backward, the punctuation has not been written yet, so it can only be estimated from the density a script of that size usually carries. That assumption is the one uncertain number in the reverse calculation, and it is stated on the page rather than hidden.

The result is given twice: the count that fits once those typical pauses are subtracted, and a higher figure that assumes no pauses at all. Writing to the second number is writing to a script that will overrun.

A script denser in punctuation than the assumption will run longer than predicted even at the recommended count — which is why the assumed spacing is spelled out rather than buried: text with tighter punctuation takes longer than the number suggests.

220 characters a minute and 140 words a minute are not the same unit

Chinese is counted in characters and English in words. The two rates look close but measure different things: 140 words is roughly nine hundred characters, more than four times 220 characters a minute. Carrying the Chinese rate over to an English script yields a duration four times too long.

This page counts CJK text in characters and Latin text in words, and the normal pace is 220 characters a minute for Chinese against 140 words a minute for English. The numbers are close; what they measure is not.

An English word with its space averages five or six characters, so 140 words is around nine hundred characters. Measured in characters per minute, English is therefore more than four times faster. The comparable quantity is how many units of meaning are spoken per minute, not how many characters.

The ratio between paces differs too. In the normal pace Chinese is 1.57 times English, 220 against 140; in the broadcast pace it becomes 1.94 times, 300 against 155. Broadcast delivery speeds Chinese up more, because a Chinese character is a short syllable with more room to compress.

That factor of 1.57 roughly matches the translation density of one Chinese character to about 0.6 English words. So the same content estimated in both languages should come out close; a large gap usually means the two scripts are not the same length, rather than that the wrong pace was chosen.