Is speaking faster than typing? What the research shows

This page is for anyone deciding whether to move part of their writing to voice and wanting the evidence first. The short answer: speech input has measured faster than phone typing in a controlled study, desktop typing averages around 52 words per minute, and neither number tells you how fast you reach a message you would actually send. The method for using voice day to day is in the Wispr Flow guide; this page is the evidence and a test you can run in five minutes.

Start with the self-test near the end. It takes less time than reading another review.

What did the Stanford study measure?

In 2016, researchers from Stanford, Baidu and the University of Washington compared speech recognition with typing on an iPhone keyboard. The paper is Speech Is 3x Faster than Typing for English and Mandarin Text Entry on Mobile Devices (arXiv).

What they found:

• English input with speech was 3.0 times faster than the smartphone keyboard. Mandarin was 2.8 times faster.
• Error rates were lower with speech, according to the Stanford report.

What the study did not cover:

• Phones only. It compared speech with a touchscreen keyboard, which is slower than a physical one.
• Short standard phrases. Participants entered short phrases from a standard text-entry phrase set. Reports and structured emails were outside the test.
• A small, young sample. 32 participants aged 19 to 32, experienced texters.
• One engine from 2016. Baidu's Deep Speech 2. Current tools add AI cleanup, which changes both speed and the kind of errors you see.

It is good evidence that speaking beats thumb typing for short messages. It is not evidence that voice is three times faster for everything you write at a desk.

How fast do people type on a keyboard?

The largest public dataset is Observations on Typing from 136 Million Keystrokes, published at CHI 2018 by researchers from Aalto University and the University of Cambridge. About 168,000 volunteers took an online typing test. The average was around 52 words per minute, with wide variation between participants.

Two things follow. First, the benefit of voice depends heavily on how fast you already type. A 90-word-per-minute typist gains less than a 35-word-per-minute one. Second, a typing test measures copying text. Composing is a different task. When you write, thinking is often the bottleneck, and speaking does not remove the thinking.

Why is sendable-text speed different from input speed?

Every message has four phases:

1. Thinking about what to say.
2. Input: typing or speaking.
3. Review: reading what came out.
4. Correction: fixing errors and restructuring.

Studies of input speed measure phase two. Your day is made of all four.

Voice shortens input and often changes thinking, because people explain more when speaking. It can lengthen review and correction: a misheard name has to be found and fixed, and a spoken ramble may need restructuring. AI cleanup in tools like Wispr Flow removes filler words and adds punctuation, which cuts correction time, but you still have to read the result.

That is why the useful number is time to a sendable message, measured on your own work.

Where does voice lose?

• Code and identifiers. File paths, function names, regex and commands come back mangled. Type them.
• Exact figures. Prices, dates in contracts, version numbers. One wrong digit costs more than the time saved.
• Small edits. Changing one word is faster with a keyboard and a cursor.
• Noise and privacy. Open offices, cafés, calls nearby, or anything you would not say aloud in the room.
• Material that must stay local. Cloud dictation sends audio to the vendor. Check the tool comparison for on-device options.

The AI architect article shows the split I use: speak the context, type the exact values.

How do you test your own speed in five minutes?

You need a timer and one message type you send often.

1. Typed take (2 minutes). Write a 150-word update you would really send: a project status, a customer reply, a brief. Time from the first keystroke to the moment you would press send.
2. Spoken take (2 minutes). Dictate a comparable update on a different but similar topic, so you are not repeating the first one. Time from starting to speak until it is sendable, including reading it back and fixing errors.
3. Count errors (1 minute). Note every misheard or wrong word you had to fix in the spoken take.

Record it:

Repeat once a day for a week. The first day is unfair to voice, because the dictionary does not know your names yet. If by day five the spoken take is faster and the corrections column is shrinking, move that message type to voice. If not, keep typing it and try a different message type.

What does Wispr claim, and how should you read it?

Wispr Flow's own marketing says "talk 4x faster". That is a vendor claim. It does not measure your work. The research above supports the direction for short phone messages and says nothing about your desk, your vocabulary or your editing habits. The five-minute test does.

For a visual walkthrough of what dictation into everyday apps looks like, Tool Finder's tutorial is a good reference point before you run the test.

YouTube

Wispr Flow Tutorial: Talk to get things done

Click to load YouTube

Disclosure: the Wispr Flow links on this page are affiliate links. If you subscribe through them I may earn a commission at no extra cost to you.

The Wispr Flow free plan includes 2,000 words a week on desktop, according to its pricing page on 20 October 2026, which is enough for the week-long version of this test. Try Wispr Flow, or run the test with your operating system's built-in dictation first.

Frequently asked questions

Is speaking really three times faster than typing?

On a smartphone keyboard, one 2016 study measured English speech input at 3.0 times the speed of typing. That study used short standard phrases, 32 participants aged 19 to 32, and Baidu's Deep Speech 2. It did not test desktop keyboards or long documents.

What is the average typing speed?

A 2018 CHI study of about 168,000 volunteers and 136 million keystrokes measured an average of roughly 52 words per minute on physical keyboards, with wide variation between people.

Why does dictation sometimes feel slower?

Raw input speed is only part of the time. Reading the transcript, fixing misheard words, and restructuring a spoken ramble all count. For short, precise edits, the keyboard usually wins.

What should I measure instead of words per minute?

Measure time from starting a message to having a version you would send. Include review and corrections. Compare the same message type typed and spoken.

Does Wispr Flow make me four times faster?

Wispr markets the product as talking four times faster. That is the vendor's claim. Your result depends on the message type, your typing speed, your accent and vocabulary, and how much editing the output needs.

Where does voice typing lose?

Code identifiers, file paths, exact numbers, one-word edits, noisy rooms, and anything you cannot say aloud where you are. Keep those on the keyboard.

Next step

Run the five-minute test today. Then read the Wispr Flow guide for setup, and the role routines in voice typing for founders or voice writing for authors. The voice typing learning path collects the tutorials in one place.