How long will your speech actually take?

The arithmetic, and why it is the easy part

Converting words to minutes is one division. A script of W words delivered at R words per minute takes W ÷ R minutes, so 500 words at 165 wpm is 3.03 minutes, which is 3:02. Going the other way is a multiplication: a five-minute slot at the same pace holds 165 × 5 = 825 words. There is no hidden model, no smoothing, and nothing about your writing style in it.

All of the difficulty sits in R. Choose a rate and the arithmetic is settled; choose the wrong rate and every downstream number is wrong by the same proportion. On a 500-word script the gap between the slowest and fastest measured presets on this page — 140 and 185 wpm — is 52 seconds. Stretch that to a 2,000-word conference talk and the same proportional gap becomes minutes — the difference between finishing inside your slot and being waved off stage mid-sentence.

The number everyone quotes, and what we did instead

A range of 125–150 words per minute circulates as the speaking rate. This site does not use it, for one plain reason: we have not read a source for it that we could show you. Passing along a figure whose origin we cannot demonstrate is exactly the habit that makes numbers like this unkillable, so instead of repeating it we went and measured.

The measurement needed material where a full transcript and an exact running time are published for the same delivery. Recorded conference talks qualify: their pages carry a complete English transcript in which every caption cue is stamped with a millisecond offset. That gives an exact word count and an exact elapsed interval for one performance, with nothing estimated in between. We ran it over 39 single-speaker talks — 78,841 words across 8.0 hours of delivery.

The median came out at 165.7 words per minute, with the whole set spread from 102 to 210. Against the commonly quoted band, 10 of the 39 talks fall inside 125–150, 28 are faster than 150, and 1 is slower than 125. Whatever that band describes, it does not describe rehearsed public speaking as this sample performs it. The full method, every talk with its own figures, and a sensitivity check on the one judgement call in the counting rule are laid out on the speaking rate page.

Two honest limits travel with that median. It is measured on rehearsed, coached, time-limited conference talks, which is the closest public material to a prepared speech but is not the same thing as a nervous first run-through in a classroom. And the transcripts are lightly cleaned by their publisher, so filler words are absent from the counts — which happens to match your situation exactly, because your script does not contain them either.

Every word count against every rate

Both tables below are complete cross products, computed when the site is built and checked cell by cell in its test suite. The first covers 60 script lengths against 14 speaking rates; the second inverts it for 60 slot lengths. Together they are 1,680 cells, and none of them was typed by a person.

Reading the answer without fooling yourself

The clock gives you the length of the script. An event is longer than its script, and the difference is where most overruns come from. Before you accept a number, subtract the parts of the slot that are not speaking:

• Questions. If the slot includes them, they are not yours. A ten-minute slot with three minutes of questions is a seven-minute script — about 1,155 words at 165 wpm, not the 1,650 the full slot would hold.
• Anything you do rather than say. A demonstration, a video, a poll, a pause while a volunteer comes up. The slowest talk in our whole set measured 102 wpm, and it earned that figure honestly: it is a demonstration, and the speaker spends long stretches silently doing the thing with no narration over the top.
• Applause and laughter. Already inside our rates, because they were inside the deliveries we timed — but only in the ordinary amounts a conference audience produces.
• Your own start. Rehearse the first minute on its own and time it. If your pace is going to drift away from the rate you planned with, that is where you will see it first, and it is the cheapest place to correct.

Fifteen minutes, five of them questions

You have written 2,000 words and been given fifteen minutes with questions at the end. At 165 wpm the script runs 12:07, which looks like a comfortable fit until you take the questions out. Reserve five minutes for them and your real budget is ten minutes, which holds 1,650 words — so roughly 350 words have to go, about 18% of the draft. That is a structural edit, not a tightening pass, and knowing it before you rehearse is worth more than any other output on this page. The 2,000-word page tabulates the cut for every common slot and pace.

The three-minute speech that ran out at two

A student is told to speak for exactly three minutes and writes to fill it at the commonly quoted 138 wpm, producing about 414 words. On the day, adrenaline puts them nearer the fast end of our measured set, 185 wpm, and the script empties in 2:14 — they finish with most of a minute of silence to fill. The fix is not a better calculator. It is measuring your own rate once, on your own script, which is what the meter below is for.

Common mistakes

• Using a blog reading-time estimate. That number models silent reading, which is a different activity at a different rate.
• Timing the first read-through. A cold read is slower than a rehearsed one. Time the third pass, and time it aloud at full volume.
• Counting the words your slides say. Only the words you speak take speaking time. Bullet text you never read out is free.
• Trusting one run. A single timing is a sample of one. Three runs and the median is a small amount of extra work and a much better number — the same rule we applied to our own dataset.
• Planning to the second. Speaking rate is not stable to two decimal places. Build in a paragraph you can drop without anyone noticing.