Speaking rate calculator: words per minute

How many words per minute do people speak?

There is no single number, because people speak at different speeds in different situations. These are the ranges most often cited, with where each comes from:

The calculator's presets come from these ranges: slow is 110 wpm (the middle of the slide presentation range), average is 150 wpm (the conversational figure), fast is 160 wpm (the top of the audiobook and podcast range, described in the sources as the upper end of what people comfortably hear and say), and very fast is 250 wpm, auctioneer speed.

What changes your speaking rate

• Nerves. Many people speed up when they are anxious, especially at the start of a talk.
• Pauses. Words per minute counts the whole time, so pausing for slides, questions or emphasis lowers it.
• The material. Numbers, names and technical terms take longer to say and longer for listeners to absorb.
• Reading versus talking. Reading a script aloud is usually steadier than unscripted talking, which has more false starts and fillers.
• Your audience. A room of non-native speakers, or a recording people will play back, benefits from a slower pace.

How to measure your own rate

1. Record yourself giving two to three minutes of the talk the way you would really give it.
2. Get a transcript, by typing it or with an automatic transcription tool, and count the words.
3. Divide the word count by the minutes. 420 words in 3 minutes is 140 words per minute.
4. Type your number into the calculator above to time the rest of your script at your real pace.

Words per minute quick reference

How many words fit in common time slots at the slow, average and fast rates:

So a 1,000-word script takes about 9 minutes at a slow pace, 6 minutes 40 seconds at an average pace, and 6 minutes 15 seconds at a fast one. A one-hour meeting with steady talking produces somewhere around 6,600 to 9,600 words of transcript, before you count the gaps when nobody is speaking.

How long does it take to transcribe 1 hour of audio?

Typing one hour of audio by hand usually takes about 3 to 6 hours, and 4 hours is the figure most often used for planning. Clear recordings of one person are at the fast end; noisy recordings, crosstalk and many speakers are at the slow end or beyond.

• A study presented at LREC 2008 (Bazillon, Estève and Luzzati, "Manual vs Assisted Transcription of Prepared and Spontaneous Speech") reported that an experienced annotator needed about 4 hours per hour of audio. Correcting an automatic draft was up to 4 times faster than typing from scratch for prepared speech, but only about 2 times faster for spontaneous speech.
• SpeechText.AI's guide plans 3 hours of work per hour for a clear single speaker, 4 for a clear two-person interview, and 6 to 10 for difficult audio.

Why it takes so long

People speak much faster than they type. In a large study of 168,000 volunteers by researchers at Aalto University and the University of Cambridge, the average participant typed 52 words per minute, and the fastest about 120. An hour of conversation at 150 words per minute is about 9,000 words; at 52 words per minute that is almost 3 hours of nonstop typing. Real transcription adds rewinding, replaying unclear words, labeling who is speaking and proofreading on top. That is why the calculator shows two manual numbers: the typing alone is a floor, and the hours-per-hour estimate is closer to what it really takes.

Automatic transcription

Automatic speech recognition does the typing for you, so your time goes into checking the draft: fixing names, jargon and the words it misheard, and checking who said what. How long the computer takes depends on the tool and your computer, but your own time is mostly one proofreading pass. If you listen through once at 1.5× speed, an hour of audio takes 40 minutes. The guide to transcribing an audio recording to text compares the main ways to do it, free and paid.

Where Hark fits

Hark is a free desktop app for Windows, Mac and Linux. It records meetings in any app without a bot joining the call (on a Mac it records the microphone only), and it can import existing recordings such as .mp3, .m4a, .wav and .mp4 files. Speech recognition runs on your computer, so the audio is not uploaded to be transcribed. It is English only.

For proofreading, Hark highlights each word as it plays and lets you click any word to replay that moment. You can play back faster without chipmunk voices and skip the silences, which makes the checking pass shorter. Your edits are kept. It works out who spoke by voice, and once you name someone it recognizes them in future meetings. Tell people when you are recording and get their consent, because recording laws vary by place.

If the transcript is for meeting records, our meeting minutes template shows how to turn it into minutes, and the meeting cost calculator shows what the meeting itself cost.

Embed this calculator on your site

Writing about public speaking, podcasting or transcription? Paste this code into your page as HTML. It shows only the calculators, free, with a small link back here.

<iframe src="https://externalsoftware.com/hark/tools/speaking-rate-calculator/?embed=1" title="Speaking rate calculator" width="100%" height="1100" style="border:0;max-width:760px" loading="lazy" allow="clipboard-write"></iframe>
<p style="font-size:14px"><a href="https://externalsoftware.com/hark/tools/speaking-rate-calculator/">Speaking rate calculator</a> from Hark</p>

Methodology and sources

• Speaking time is words divided by words per minute; word count is minutes times words per minute. Neither includes pauses you add for slides or questions.
• Pasted text is counted by splitting on spaces and line breaks, so hyphenated words count once and numbers count as words.
• "Typing it by hand" is the recording length times the hours-per-hour figure you choose (3, 4 and 6 from the sources below).
• "Typing alone" estimates the words in the recording at 150 words per minute and divides by your typing speed (52 by default, the study average). It leaves out rewinding and checking.
• "Automatic, then proofread" is the time to listen through once at the playback speed you pick. It does not include the time the software takes, which varies by tool and computer.
• Everything runs in your browser. When you change a value, the numbers (never pasted text) are written into the page address so a copied link reproduces your results.

Sources

• VirtualSpeech: Average speaking rate and words per minute
• Wikipedia: Words per minute (Williams 1998; Wong 2014)
• Bazillon, Estève and Luzzati (LREC 2008): Manual vs Assisted Transcription of Prepared and Spontaneous Speech
• SpeechText.AI: How long does it take to transcribe an interview?
• University of Cambridge: What makes a faster typist? (Dhakal, Feit, Kristensson and Oulasvirta, CHI 2018)