Count words in a transcript

What each number means

• Words: whitespace-separated tokens that contain a letter or a digit.
• Characters (with spaces): every character, including spaces and line breaks, counted as the person sees them, so an emoji or an accented letter counts once.
• Characters (without spaces): the same with all spaces, tabs and line breaks removed. This is the figure some forms and translation quotes ask for.
• Sentences: stretches of text ending in a full stop, question mark, exclamation mark or ellipsis, including their Arabic and East Asian forms. Abbreviations such as “e.g.” can make the number a little high, and text with no punctuation at all counts as one.
• Paragraphs: blocks of text separated by one or more blank lines.

How reading time is estimated

Reading time is the number of words divided by a words-per-minute figure. There is no single correct figure. The commonly quoted average for adults reading ordinary prose silently is a little under 250 words per minute, which is why 240 is the default here, but studies report a wide spread between readers and between kinds of text, and reading on a screen, in a second language or with unfamiliar terminology is slower. Change the box to match your audience and the estimate updates straight away. Treat the result as a planning number, good for deciding whether something is a two-minute read or a ten-minute one.

Speaking time is a separate estimate

Reading aloud is slower than reading silently, and it varies with the speaker and the material. The default of 150 words per minute is a conversational rule of thumb, not a measurement of anyone. If you have a recording of yourself, the speaking time calculator works your own pace out from a word count and a duration, and can also tell you how many words fit a time slot.

Counting a subtitle file

A raw SRT file is full of numbers and timestamps that are not part of what was said. Paste one in and the tool detects the timestamps and counts only the caption text, so a one-hour talk shows roughly the number of words that were actually spoken. Here is how little of a file is actual speech:

1
00:00:01,000 --> 00:00:03,500
Welcome to the show.

That cue is four words in a block of three lines. Counting the file as it stands would report seven, and the count gets worse the shorter the cues are. To keep the cleaned text as well, run it through the SRT to text tool.

Using a count for a transcript

Word counts are useful for checking whether a transcript is complete (a one-hour interview that comes out at 800 words has probably lost most of the audio), for estimating translation or editing costs that are quoted per word, and for fitting captions or show notes to a limit. If you need the transcript in the first place, turn a video into text, then bring it here. For the formats those transcripts come in, see transcript file formats explained.