Vociferous

How we count

Insights counts your transcripts as they stand now: a transcript you augmented or edited counts the words of its current version. Measures of how you speak, such as pace and fillers, count the words you spoke. The augmentation figures compare an augmented version with the text it was made from. We count what happened, not what it might be worth.

Words

A word is anything between two spaces in a verbatim transcript, so “don't” is one word. Chinese and Japanese are written without spaces, so each of their characters counts as one word, as word processors count them. Thai, Lao, Khmer and Burmese, also written without spaces, count one word for every four characters.

Your words follow each transcript's current version, so augmenting or editing a transcript changes them, and switching back to the captured words returns them. A transcript continued in later takes counts its current version once, shared among its takes by their captured words; a take added after that version was made counts as captured. A transcript with a speaker left out keeps the words of the speakers who count, because its augmented version cannot be split by speaker.

Sessions and days

A session is one transcript. A transcript you continued in later takes is still one session, and each take counts on the day it was saved. A transcript continued on another day is active on both days, and counts in each period its takes fall in.

Days follow the time zone set in Settings, under Account. Changing it recounts every day in the new zone. A week runs Monday to Sunday. Today, Week and Month run from their start to now. An imported file counts on the day you imported it.

Beside Today, Week and Month, Insights sets the period before. “To date” is the same stretch of it: last week to date is last week from Monday up to the same moment, and last month to date likewise. “All of” is the whole of that period. Yesterday is always the whole day.

Recorded time and silence

Recorded is how long you spent recording, measured on a clock. The timer stops while you pause. An imported file counts the length of its audio.

Silence is included: the quiet inside a recording is you thinking. It is your recorded time less the time your words take, and a gap between words counts as silence when it lasts over 0.15 seconds. Long silences usually don't count against your minutes, but they stay in your recorded time and in your silence. The share of silence is counted only over the recordings that measured it.

Pace and pauses

Recording pace is the words you spoke divided by your recorded minutes, so silence slows it. A transcript with no recorded time is left out of it.

Average pause is the average gap of more than 0.15 seconds between your words, as the speech recognizer timed them. Each transcript counts once in the average, however long it is. For a take with one speaker, the word timings follow the audio sent for transcription, so its pauses leave out the long silences held back while you recorded, though its silence includes them.

Transcripts made before 3 October 2026 keep times measured on the shortened audio, so their pace reads a little high, their pauses short and their pauses per minute a little low.

To type them is your words divided by your typing speed, which you set in Settings, under Preferences. Until you do, 40 WPM is assumed. It is not time saved.

Filler Words

We count “um”, “uh”, “er”, “erm”, “ah”, “eh”, “mm” and “hmm”, with spellings such as “umm”, and the phrases “you know”, “I mean”, “kind of” and “sort of”. Ordinary words such as “so”, “like” and “well” are not counted. The speech recognizer drops some fillers on its own, so the count understates them. These are English fillers, so they are counted only in a transcript that reads as English: most of its letters are Latin, and at least one word in twelve is a common English word such as “the”, “and” or “you”. A transcript of under eight words in plain A to Z reads as English. In any other transcript, fillers are not measured.

Reading grade

Reading grade is the Flesch–Kincaid grade, worked out from the length of your sentences and the syllables in your words, with syllables estimated from spelling. It is an English measure, so like fillers it is worked out only for a transcript that reads as English and has at least three words; any other transcript has no reading grade. It runs from 0 to 20, and each transcript counts once in the average, however long it is. It describes the text. It never scores you.

Augmentation

Augmentation is counted in words; its one time is the median wait. Each figure compares a transcript's current augmented version with the text it was made from: the version it was built on, such as your edit or an earlier augmentation, or else your captured words. Words in, the reading grade before and the fillers before describe that text.

Words changed counts only that augmentation's own changes against it, word by word, never earlier or rejected versions. A change of capitals or punctuation alone counts as a fifth of a word. Fillers erased is how many fewer fillers the augmented version holds.

A transcript counts as augmented when the version it shows is an augmentation, or an edit or correction made over one, or an imported transcript with a correction. A run you discarded or never kept, or a transcript taken back to what you said, does not.

Augmented, never edited is the share of all transcripts in the window that were augmented and hold no edits. Augmented and edited is the share of all transcripts in the window that were augmented and hold one, made before or after augmenting.

What counts in analytics

A transcript you switch out of analytics, on its Analytics tab, does not count in Insights or toward your goals, and its words leave your distinct words unless a transcript that counts says them too. Switching it back returns them.

A transcript you delete leaves every figure the same way, its words included. The time transcription and augmentation take is counted only over the transcripts that count.

In a conversation you can leave out a speaker instead, in the Speakers tab. The transcript then counts only the words, fillers and reading grade of the speakers who count, and its time becomes the time they spent speaking. Its augmented text is left out of the augmentation counts, because it cannot be split by speaker. Leaving out every speaker leaves out the transcript.

A speaker's words leave your distinct words the same way. On an imported transcript made before 3 October 2026, leaving out one speaker can take more than that speaker's share of its time.

Goals, Records and Habits

A day counts toward your weekly target when any take of a transcript was saved on it. When you speak places each take in the hour it was saved.

Busiest day is the day with the most words, as every word count is, so imported transcripts count in it. Longest session, fastest recording pace and When you speak leave imported transcripts out, because they describe how you speak. Longest session is the transcript with the most recorded time, its takes added together. Fastest recording pace counts only transcripts of at least 15 seconds, and ignores any above 300 words a minute.

Milestones count your words and sessions as above, and your distinct words: each dictionary word counts once, from the first time you say it. The dictionary is English, so words in another language add nothing to it.

Rounding

Counts are whole numbers. Under ten minutes, times show minutes and seconds; from ten minutes, whole minutes, rounded down. Pace is rounded to a whole word a minute, shares to a whole percent, words changed to a tenth, pauses to a tenth of a second and reading grade to one decimal.