Speaking Time Calculator
Paste your script and watch its spoken length update as you cut — tuned to presentation pace (150 wpm) for talks, videos, podcasts, and ad reads.
How the estimate is made
The calculation is a single division. The tool counts the words in your script and divides by 150, the words-per-minute figure the engine uses for speech. That figure is a typical prepared-speech pace — an average across many speakers and many kinds of talk, not a speed for you to aim at — and the same constant is used wherever this site shows a speaking time, so a script measured on one page of this site gives the same answer on every other page of it. The division produces a fraction of a minute, which the display converts to minutes and seconds, rounding to the nearest second; the rounding rules are set out below. At 150 words per minute each word is worth 0.4 seconds, a figure worth keeping in mind while you edit: a cut of ten words saves about 4 seconds, while a cut of one or two words may or may not move the display, depending on where the exact figure sits against a rounding boundary.
What counts as a word follows the same rule as every other tool on this site. A word is a run of letters and digits; internal hyphens and apostrophes keep a word whole, so "state-of-the-art" and "it's" are one word each, while a comma or a period ends a run, so "2,500" is two words and "3.14" is two as well. The counter is therefore counting typed tokens, not spoken beats, and the gap between those two things is where the estimate drifts — most visibly on numbers, which get their own section below. The tokenization and sentence rules are documented in full on the methodology page.
For the planning side — choosing a length for a slot, budgeting for pauses and questions, and rehearsing against a stopwatch — read the guide Speaking time and speech length. The silent-reading figure shown beside the speaking time is explained in Reading time and words per minute.
Worked example: a short opening
Paste this opening into the tool above and the panel fills in as follows.
“Good morning, and thank you for making time for this. Over the next few minutes I want to show you one chart, explain what changed behind it, and ask for a decision. If you leave with only one number, it should be the one on the last slide. Let’s start with where we were a year ago.”
57 words · 4 sentences · speaking time 23 sec · silent reading 14 sec
The counter finds 57 words in 4 sentences. Divided by 150 and multiplied by sixty, that is 22.8 seconds of speech, which the display rounds to 23 sec. Read silently at 238 words per minute, the same passage takes 14 sec. Both numbers are estimates in the plainest sense: a speaker who pauses after "one chart" for effect, or slows down on the last sentence because it is the one that matters, will take longer, and the counter has no way to see either choice. What the estimate does reliably is rank drafts against each other — cut a sentence and the estimate drops by 0.4 seconds for every word removed — which is the job it is built for.
From minutes to words: the inversion
The tool answers the question "how long will this take?", but the question you have while writing is usually the reverse: "how many words fit in the slot?" Invert the division and the answer is words = minutes × 150. The table gives that budget for common slots, with a round-trip check (the budget fed back through the tool's own formatter) and the silent-reading time of the same word count, which is the figure a reader would see on a reading-time badge for the same text.
| Slot | Words at 150 wpm | Round trip | Silent reading |
|---|---|---|---|
| 30 sec | 75 | 30 sec | 19 sec |
| 1 min | 150 | 1 min | 38 sec |
| 2 min | 300 | 2 min | 1 min 16 sec |
| 3 min | 450 | 3 min | 1 min 53 sec |
| 5 min | 750 | 5 min | 3 min 9 sec |
| 10 min | 1,500 | 10 min | 6 min 18 sec |
| 15 min | 2,250 | 15 min | 9 min 27 sec |
| 20 min | 3,000 | 20 min | 12 min 36 sec |
| 30 min | 4,500 | 30 min | 18 min 54 sec |
| 45 min | 6,750 | 45 min | 28 min 22 sec |
| 60 min | 9,000 | 60 min | 37 min 49 sec |
Two cautions apply to every row. First, a budget is a ceiling on words spoken continuously, and no talk is spoken continuously: a slide change, a demo, a question from the floor, a laugh you wait out — none of these are words, and all of them consume the slot. Subtract them from the slot before you look up the budget, then script to what is left. Second, the inversion is exact in words but not in the display. Because the panel rounds to whole seconds, several neighbouring word counts show the same time, so "the number of words in five minutes" is a narrow band rather than a point, and the round-trip column lands exactly on the slot only because every budget was built from the pace itself, so its seconds come out whole; a word count picked at random would usually not. The speaking-time guide works through applying these budgets to a real slot.
Why number-heavy scripts run long
The estimate divides typed words by a pace, so anything that is voiced in more words than it is typed runs over. Numbers are the main offender, and the reason is tokenization rather than arithmetic. The table shows five fragments as they are usually typed and as they are usually voiced, each counted by the same engine that runs the tool.
| As typed | Counted | As voiced | Counted |
|---|---|---|---|
| 2,500 | 2 | two thousand five hundred | 4 |
| $3,750 | 2 | three thousand seven hundred and fifty dollars | 7 |
| 1.5 | 2 | one point five | 3 |
| HTML | 1 | H T M L | 4 |
| 48 | 1 | forty-eight | 1 |
"2,500" is two words to the counter because the comma ends a run, yet it is four when spoken. The currency symbol is not a word at all, while the "dollars" that replaces it aloud is. A decimal point splits "1.5" into two tokens, and the voiced form adds "point" on top. An acronym read letter by letter is one token but as many spoken beats as it has letters. Notice the last row, though: a two-digit number typed as a numeral and spoken as a hyphenated word counts the same both ways, so the drift does not come from every digit — it comes from thousands separators, currency, decimals, and initialisms. Put the fragments into a sentence and the effect compounds.
As typed: “Revenue rose from $2,500 to $3,750 over 12 months, a ratio of 1.5. The full HTML report runs to 48 pages.”
24 words · speaking time 10 sec
As voiced: “Revenue rose from two thousand five hundred dollars to three thousand seven hundred and fifty dollars over twelve months, a ratio of one point five. The full H T M L report runs to forty-eight pages.”
36 words · speaking time 14 sec
The voiced transcript carries 12 more words than the typed sentence — 50% more — with nothing added in meaning. A script that is figures from end to end will drift by a comparable share, and a script with one price in it will hardly drift at all; the counter cannot tell you which yours is, but you can. For a passage dense with numbers, either paste a spoken-out version of it, as above, or read that passage aloud against a stopwatch and use the measured time in place of the estimate. Symbols that are neither letters nor digits behave the same way: a percent sign or a degree sign is not counted, but it is spoken. A unit abbreviation is different: "kg" is a run of letters, so it is counted as one word, and whether it is voiced as one word or several is a choice the counter cannot see.
Reading time versus speaking time
The panel shows two times because the engine carries two constants. The silent-reading figure divides by 238 words per minute, the adult silent-reading average from Brysbaert's 2019 meta-analysis of reading-rate studies; the speaking figure divides by 150, the prepared-speech pace described above. Both are averages over populations and genres, presented as estimates. Because the two constants are fixed, so is the ratio between them: a passage is spoken in about 1.6 times the time it is read. That ratio is the number to internalize when a script arrives as an article draft, because the writer's sense of its length was formed by reading it. A "2 min read" — 476 words at the reading average — is delivered in 3 min 10 sec.
“Before we look at the plan, I want to set out the problem in plain terms. Last spring our support queue doubled in six weeks, and the team that handled it did not grow at all. We tried the obvious fixes first: longer shifts, a triage rota, and a template for the most common replies. Each one helped for a while, and each one stopped helping as the queue kept climbing. What finally worked was not a faster way to answer tickets but a slower, more careful look at why the same three questions kept arriving. Today I will walk through those three questions, show what we changed in the product to remove them, and put a number on what that change saved.”
123 words · 6 sentences · silent reading 31 sec · speaking time 49 sec
At paragraph scale the gap is already visible — 31 sec on the page against 49 sec at the lectern — and it grows in proportion to length, since both estimates are linear in the word count. Pasting an article draft into a video pipeline without cutting is how a topic that reads in a few minutes becomes a recording that runs much longer, and the speaking-time panel is the place to catch it. For the reading side of the comparison, including where the reading average comes from and what it does not cover, see Reading time and words per minute.
How the seconds are rounded
The formatter turns a fraction of a minute into the string you see. It multiplies by sixty and rounds to the nearest whole second. A result of zero shows "0 sec"; anything under five seconds shows "under 5 sec" rather than a misleadingly precise figure; under a minute it shows seconds only; an exact number of minutes shows minutes only; and everything else shows both. The table runs word counts around those thresholds through the same function.
| Words | Exact seconds at 150 wpm | Display |
|---|---|---|
| 1 | 0.4 | 0 sec |
| 2 | 0.8 | under 5 sec |
| 4 | 1.6 | under 5 sec |
| 12 | 4.8 | 5 sec |
| 13 | 5.2 | 5 sec |
| 150 | 60.0 | 1 min |
| 151 | 60.4 | 1 min |
| 152 | 60.8 | 1 min 1 sec |
Three things follow. A single word showing "0 sec" is rounding, not a fault: 0.4 seconds rounds down to zero. The first count to display a non-zero whole number of seconds is 12 words, and 13 words display the same, so on very short texts the panel is quantized more coarsely than the underlying estimate. And at exactly 150 words the display reads "1 min" — one more word still reads "1 min", and the first extra second appears at 152. The display has one-second precision; the estimate behind it does not, because the pace is an average carrying far more uncertainty than a second. Read the figure as a rounded quantity, not to the last second, and treat two drafts a few seconds apart as the same length.
Scripts with labels and stage directions
The tokenizer counts every run of letters, and a production script is full of runs of letters that nobody says aloud: speaker names, bracketed directions, camera notes. The short exchange below is counted twice, once as written and once with only the spoken words.
HOST: Welcome back. [PAUSE] Before the break we were talking about deadlines.
GUEST: Right, and I said the deadline is never the problem. [LAUGHS] The problem is the week before it.
HOST: [TO CAMERA] We’ll pick that up after this.
40 words with labels · speaking time 16 sec
33 words spoken · speaking time 13 sec
The labels and directions add 7 words here — 21% on top of what the audience hears. On a long two-hander with a name at the head of every line, the overhead scales with the number of lines, and the estimate drifts long by that much. The clean fix is to strip labels and directions before pasting. A useful variant is to paste one speaker's lines on their own, which gives that speaker's time rather than the scene's. Outlines have the same problem in a different costume: headings and bullet stubs are words to the counter whether or not you intend to voice them.
Where the estimate breaks
- The pace is an average, not yours. Nervous speakers rush, careful ones slow, and the same speaker varies between a rehearsal and a room. Measure your own pace once — words divided by rehearsed minutes — and scale the estimate by 150 divided by that figure.
- Silence is invisible. Pauses, applause, slide transitions, demos, and questions are not text. The estimate is the talking, not the slot.
- Typed and voiced diverge. Numbers, currency, decimals, acronyms, symbols, and URLs are voiced in more words than they are counted, as shown above.
- Markup is counted. Speaker labels, stage directions, headings, and bullets inflate the count.
- Very short texts are rounding. Below a few dozen words the display's whole-second steps dominate; use a full paragraph for anything you intend to act on.
- One pace for every text. The counter is Unicode-aware and counts accented and non-Latin words, but the pace is a single figure that does not adjust for language, speaker, or genre.
A working method
- Strip speaker labels and stage directions, or paste one speaker's lines only.
- Paste the script and note the speaking time. Ignore the last few seconds; they are rounding.
- Subtract your non-speaking time — slides, demo, questions — from the slot, and compare the estimate with what is left, not with the slot itself.
- Cut to the budget. Each ten words is about 4 seconds, so the count in the panel tells you roughly how much to remove; a cut of a word or two may or may not move the display, depending on where the exact figure sits against a rounding boundary.
- Rehearse against a stopwatch. Divide the word count by the minutes it took to get your own pace, and from then on scale the page's estimate by 150 divided by that pace.
- Re-check the number-dense passages separately, either by voicing them aloud on the clock or by pasting a spoken-out transcript of them.
Common misreadings
- Treating 150 as a speed to hit. It describes an average; the right pace is the one your material and your room need.
- Reading "speaking time" as "slot time". The estimate covers the words only; everything else in the slot is on you to add.
- Confusing the two panels. The silent-reading figure is the smaller one and describes a reader, not a speaker. If you are deciding how long to stand up, use the larger.
- Expecting long words to cost more. The estimate weighs every word the same, so a script of long words and a script of short words show the same time at the same count; only a rehearsal distinguishes them.
- Trusting the last second. The display's precision is a property of the formatter, not of the estimate.
The guides index collects the longer explanations behind every tool on this site.
Frequently asked questions
What pace does it assume?
150 words per minute, a typical prepared-speech pace. It is an average across many speakers and many kinds of talk, not a target for you to hit: conversational delivery, deliberate keynote delivery with pauses, and a nervous first rehearsal all land elsewhere. The estimate is your starting point; a rehearsal timer is the truth.
How many words is a 10-minute talk?
About 1,500 at 150 wpm, before any of the things that are not words: slide changes, demos, questions, laughter. None of those are visible to a word counter, so subtract the non-speaking time from the slot first and script to what remains.
Does this work for video and ad scripts?
Yes. A voiceover or ad read is a word-budget problem: a 30-second read holds about 75 words at this pace and a 60-second read about 150. Edit the script here until the seconds match the cut, then time a read-through, because copy dense with prices and figures is voiced in more words than it is typed.
Speaking vs reading time — why are they different?
Silent reading runs faster than speech. The reading figure divides by 238 wpm, the adult silent-reading average from Brysbaert’s 2019 meta-analysis of reading-rate studies; the speaking figure divides by 150 wpm. The 123-word passage on this page reads in 31 sec and speaks in 49 sec. The reading side is covered in the reading-time guide.
Can I change the words-per-minute figure?
Not on the page: the estimate uses one fixed pace so that a script measured today compares with one measured next month. To adapt it to your own delivery, time a rehearsal, divide the word count by the minutes it took to get your personal pace, then multiply the page’s estimate by 150 divided by that pace.
Why does a one-word test show "0 sec"?
Each word is worth 0.4 seconds at 150 wpm and the display rounds to the nearest whole second, so a single word rounds down to "0 sec" and two words show "under 5 sec". The first count to display a non-zero whole number of seconds is 12 words. Tests that short are dominated by rounding; paste a full paragraph.
Does it count speaker names and stage directions?
Yes. To the word counter, "HOST:" and "[PAUSE]" are words like any other. The short labelled exchange on this page counts 40 words with its labels and 33 without them, an overhead of 7 words the audience never hears. Strip labels and directions before pasting, or read the estimate as slightly long.
Why did my rehearsal run longer than the estimate?
The usual causes: numbers, currency, and acronyms that are voiced in more words than they are counted; pauses for slides, demos, and audience reaction; and a personal pace slower than 150 wpm. The first is measurable here by pasting a spoken-out version of the passage. The other two only a stopwatch can see.
Estimates assume steady presentation pace; real delivery varies. Scripts are processed locally in your browser and never transmitted. See the methodology page.