Voice notes have a reputation. Ask anyone about them and you hear about the five-minute monologue, the rambling update, the audio you put off for an hour. We wanted to know how often that really happens.
So we counted. Kaption has transcribed more than 100 million WhatsApp voice notes in the last three years. For this article we looked at the most recent 90 days: between 25 July and 22 October 2026, Kaption transcribed about 7.8 million voice notes in WhatsApp Web. This article covers what those numbers show about length, language, place and time. We didn't read any message content to get them. The methodology section at the end explains exactly what was counted.
Updated October 2026
The short version
- The median voice note is 14 seconds long. The average is 26 seconds.
- Only 7.7% run past one minute. Only 0.4%, about 1 in 240, run past five minutes.
- 74% are in Portuguese and 18% in Spanish. We detected more than 40 languages.
- 84% are transcribed between 8:00 and 18:00 local time. The busiest hour is 11:00.
- A Saturday has about a quarter of a weekday's voice notes. A Sunday has about an eighth.
How long is a voice note?
Most are short. Four in ten last 10 seconds or less. More than nine in ten are done within a minute.
| Length | Share of voice notes |
|---|
| 10 seconds or less | 39.2% |
| 11 to 60 seconds | 53.1% |
| 1 to 2 minutes | 5.5% |
| 2 to 5 minutes | 1.8% |
| 5 to 10 minutes | 0.29% |
| 10 to 30 minutes | 0.11% |
| Over 30 minutes | 0.016% |
The average (26 seconds) is almost twice the median (14 seconds). That gap comes from a small number of long recordings. Here is the same distribution as percentiles:
| Percentile | Length |
|---|
| Median (50th) | 14 seconds |
| 75th | 28 seconds |
| 90th | 52 seconds |
| 99th | 3 minutes 7 seconds |
| 99.9th | 10 minutes 49 seconds |
These numbers barely move over a longer window. Across the last 12 months the median was still 14 seconds, the average 26 seconds, and the share over one minute 7.7%.
The five-minute voice note is real, but rare
About 1 voice note in 240 runs past five minutes. About 1 in 800 runs past ten minutes. About 1 in 6,300 runs past half an hour.
Long notes are rare, but each one takes a lot of listening. A 14-second note costs you 14 seconds. A 12-minute note in a work group is a different problem. That's what summaries are for. Across the voice notes that got one, the median summary was 72 words, and the median transcript it came from was 147 words. The summary is about half the length.
Words, not seconds
The median transcript is 38 words long. The average is 66 words, and nine in ten transcripts are 136 words or shorter. For most voice notes the text is a short paragraph, often just a sentence or two.
Which languages?
Each transcription gets a detected language. We have that label for about 85% of transcriptions from 27 August to 22 October 2026, so the table below covers that shorter window.
| # | Language | Share of voice notes |
|---|
| 1 | Portuguese | 74.4% |
| 2 | Spanish | 17.7% |
| 3 | English | 2.0% |
| 4 | Italian | 1.7% |
| 5 | Arabic | 0.8% |
| 6 | German | 0.5% |
| 7 | Hindi | 0.5% |
| 8 | French | 0.4% |
| 9 | Russian | 0.4% |
| 10 | Chinese | 0.3% |
The top two languages account for 92% of voice notes. The top ten account for 98.7%. More than 40 languages appeared at least 100 times each.
English comes third at 2%, far behind the two leaders. The country table below explains why.
Where?
This table uses the country code of the WhatsApp number on the Kaption account. The US and Canada share the +1 code, so they appear together.
| # | Country | Last 90 days | Last 12 months |
|---|
| 1 | Brazil | 74.7% | 71.6% |
| 2 | Argentina | 7.6% | 8.4% |
| 3 | Colombia | 4.5% | 4.1% |
| 4 | Italy | 1.3% | 1.8% |
| 5 | Spain | 1.2% | 1.7% |
| 6 | Peru | 1.1% | 1.3% |
| 7 | Mexico | 1.0% | 1.1% |
| 8 | USA and Canada | 0.7% | 0.9% |
| 9 | Venezuela | 0.7% | 0.8% |
| 10 | Chile | 0.5% | 0.6% |
Over 12 months, Israel (0.9%) also made the top ten and Chile dropped out. The language table and the country table tell the same story: Brazil first, then Spanish-speaking Latin America.
When?
Voice notes in WhatsApp Web get transcribed during working hours. The times below are local, based on the timezone of each account's country. They cover nine of the ten countries in the table above (all except the US and Canada), which account for 93% of voice notes. For Brazil and Mexico, which span more than one timezone, we used the capital's.
| Local time | Share of voice notes |
|---|
| 00:00 to 06:00 | 0.8% |
| 06:00 to 09:00 | 6.0% |
| 09:00 to 12:00 | 28.5% |
| 12:00 to 15:00 | 25.2% |
| 15:00 to 18:00 | 25.8% |
| 18:00 to 21:00 | 9.7% |
| 21:00 to 24:00 | 3.9% |
The busiest hour is 11:00 to 12:00, with 10.4% of the day's voice notes. There is a clear lunch dip: the 12:00 and 13:00 hours are about 22% quieter than 11:00. The afternoon then climbs back up and holds steady until 17:00.
The week shows the same pattern, and it shows it more strongly.
| Day | Voice notes, compared with an average weekday |
|---|
| Monday | 88% |
| Tuesday | 104% |
| Wednesday | 99% |
| Thursday | 106% |
| Friday | 102% |
| Saturday | 24% |
| Sunday | 12% |
Saturday and Sunday together account for less than 7% of the week. We think this says less about voice notes than about where people read them. These numbers come from WhatsApp Web, which runs on a computer, and for most people the computer is where work happens. People still get voice notes on weekends. They just don't read them at a desk.
Methodology
What we counted. One record per voice note that Kaption turned into text in WhatsApp Web on a computer. Kaption's phone app and call recordings are not included. If two people transcribed the same voice note, for example in a group, it counts twice.
Date range. 25 July to 22 October 2026 (90 days, UTC), plus 23 October 2025 to 22 October 2026 for the 12-month comparisons. The language figures cover 27 August to 22 October 2026, because that is when the detected-language label starts.
What we excluded. Kaption's own internal and test accounts, records with no account, and records with no length. We also dropped a handful of lengths over four hours as recording errors: 6 records out of the 7.8 million in the sample.
Where the numbers come from. Each transcription leaves a small usage record: when it happened, how long the audio was, how many words came out, the detected language, and the country code of the account's number. Every figure here is a count or a share across millions of these records.
What we did not do. We didn't read, listen to or analyze any voice note or transcript to write this. No message content is part of these records. We didn't look at any single person's activity, and nothing in this article can identify a person or an account.
Questions
How long is the average WhatsApp voice note?
Among the voice notes transcribed with Kaption from late July to late October 2026, the median was 14 seconds and the average was 26 seconds. Fewer than 8% ran past one minute.
How common are long voice notes?
Rare. About 0.4% of voice notes ran past five minutes and about 0.1% past ten. They stand out because each one takes a long time to listen to.
Why is Portuguese the top language?
Most of the voice notes we transcribe come from accounts with Brazilian numbers, and voice notes are a strong habit in Brazil. Spanish is second, from Argentina, Colombia, Peru, Mexico and the rest of Latin America, plus Spain.
Did you read people's messages to get these numbers?
No. The figures come from usage records that hold counts, lengths, times, a language label and a country code. They contain no audio and no text. See the methodology section above.
Can I read voice notes as text on my computer?
Yes. The free Kaption extension for WhatsApp Web transcribes voice notes in Chrome and Edge, and the desktop app does the same on Mac and Windows. Here is a step-by-step guide: how to transcribe WhatsApp voice notes on a computer.
