Skip to content

Understanding transcription

How Many Words Is an Hour of Audio?

About 7,000 to 10,000 English words for an hour of speech, and fewer in Arabic. Sourced speaking rates and the word counts in our own test clips.

The Scriptorium team ·

An hour of English speech is roughly 7,000 to 10,000 words, depending on how fast people talk. At the 150 words per minute that VirtualSpeech's page on speaking rate gives for conversation, an hour is 9,000 words. In our own test clips, read English ran at 171 words per minute (about 10,280 an hour) and read Arabic at 97 (about 5,840 an hour). Arabic gets fewer "words" per minute because it attaches many small words to the next one when it is written.

Speaking rates and what they give per hour

SourceWords per minuteWords in an hour
Conversation, US English (VirtualSpeech, citing the National Center for Voice and Speech)about 150about 9,000
Conversational, listed without attribution (VirtualSpeech)120 to 1507,200 to 9,000
Presentations, comfortable pace (VirtualSpeech)100 to 1506,000 to 9,000
Audiobooks, radio and podcasts (VirtualSpeech)150 to 1609,000 to 9,600
Our six English test clips (audiobook speech, 1,391 words in 487 seconds)171about 10,280
Our three Arabic test clips (read sentences, 399 words in 246 seconds)97about 5,840

The first four rows are VirtualSpeech's own statements, only one of them attributed to a research source, so read them as typical ranges. The last two are our measurements: counts of words in the human reference transcripts, divided by the seconds of audio. They come from read speech, and conversation, interviews and phone calls will differ.

Why does Arabic give fewer words?

In written Arabic, the conjunction "and", many prepositions and the definite article are attached to the word that follows. A phrase that English writes as four words, "and with the book", is one written Arabic word. So the same amount of speech makes fewer whitespace-separated words. This is a property of the script, not a slower pace of speech, and it means Arabic word counts and English word counts should not be compared as if they measured the same thing.

A quick planner

Recording lengthWords at 150 per minuteEnglish at our 171Arabic at our 97
15 minutes (the free preview limit)2,2502,5691,460
30 minutes4,5005,1382,921
60 minutes9,00010,2765,841
90 minutes13,50015,4148,762

These are multiplications of the rates above, so they carry all the uncertainty of the rates.

Where does the number matter?

  • Anything priced per word. Translation and editing are often quoted per word, so the transcript's word count sets the bill. Count it rather than estimating from the length.
  • Subtitle and reading length. A transcript of an hour runs to thousands of words; our word counter and the formats guide help you split it.
  • Word error rate. It is divided by the number of reference words, which is why Arabic and English scores are not directly comparable (how it is calculated).

Count your own transcript

Paste the text into our transcript word counter, which runs in your browser, or in Python:

print(len(open("transcript.txt", encoding="utf-8").read().split()))

That counts runs of characters separated by spaces, the same way our word-error-rate script splits words. A word processor may count slightly differently around hyphens and numbers.

How we counted

We took the reference word counts and clip durations from our measurements page (clips of read speech, measured 2026-10-24) and divided. Nine clips are too few to describe a language; they show why the two languages differ, not an average for either.

Checked on 2026-10-26. VirtualSpeech's page on average speaking rate (the 150 words per minute figure and its attribution, and the other ranges). The rows for our own clips are arithmetic on our published data file. The Arabic attached-word example is a standard feature of the written language, not a measurement. Menus, prices and limits change; confirm on the vendor's page before you rely on them.

Questions

How many words is a one-hour recording?
Roughly 7,000 to 10,000 English words for normal speech. In our test clips, read English was about 10,280 words an hour and read Arabic about 5,840.
What is the average speaking rate?
VirtualSpeech cites about 150 words per minute for conversation in US English, attributed to the National Center for Voice and Speech, and lists 100 to 150 for presentations and 150 to 160 for audiobooks and podcasts.
Why does Arabic have fewer words per minute than English?
Because written Arabic attaches the conjunction, many prepositions and the definite article to the next word, so the same speech has fewer space-separated words.
How do I count the words in my transcript?
Paste it into the transcript word counter on this site or count whitespace-separated pieces in a script.

Transcribe a file

Free for a first file, with no account.

More pages like this