Dictation vs Typing: When Speaking Beats the Keyboard

Speaking produces words faster than typing, and in lab tests the gap is large. In a controlled study on an iPhone, English speech entered text about 2.93 times faster than the built-in keyboard ¹. Finished text is a different measure. Dictated text carries more errors into the final copy ², and fixing them takes time. So dictation wins for first drafts, messages and notes, while the keyboard wins for precise edits, code, tables and quiet shared rooms. The practical answer is to speak to get words down and type to fix them.

This guide sets out the evidence on speed, accuracy, editing time and setting. It belongs to the overview of dictation. For the full keyboard and phone speed data, read the typing speed overview. For short answers, see Is dictation faster than typing?.

Raw speech is several times faster than typing, but finished text narrows the gap

Speech wins clearly on raw entry speed. In a laboratory study on an iPhone, English speech reached 153 words per minute, against 52 on the built-in keyboard, and Mandarin speech reached 123 words per minute against 43 on the Pinyin keyboard ^{1 3}. The study counted the finished text, including time spent correcting errors ¹, so the advantage is not only a fast first pass.

The advantage shrinks at the end. Speech produced fewer errors during entry, with a corrected error rate of 5.30 percent against 11.22 percent on the keyboard ². Yet it left slightly more errors in the final text: 1.30 percent uncorrected against 0.79 percent ². Those leftover errors are what an editor must find. Fast drafting does not give clean text without a check.

The chart places these numbers beside other reported speeds. The studies differ, so read the bars as an order of magnitude, not as one test.

Treat the phone numbers as a snapshot, not a forecast. The study used short phrases in a laboratory and Baidu's Deep Speech 2 recognizer ¹. Its authors say that performance in non-laboratory settings still needs study for both methods ⁴, and they note that upper-bound recognition performance has greatly improved compared with earlier systems ⁵. Newer engines may move both the speed and the error figures, and this study cannot say by how much.

Composing, not speaking, limits how fast you write

Your mouth is not the bottleneck. Telephone conversation in English and Chinese averages about 196 words per minute overall, counting silences and listening turns ⁶, and the authors of an earlier dictation study put normal speaking at 125 to 150 words per minute ⁷. Speech delivers words far faster than the composing figures below.

Composing by dictation software ran much lower in earlier reports. Feng, Karat and Sears cite reported composition speeds of 8 to 15 words per minute for speech software ⁸. That is about 5 to 12 percent of normal speaking speed ⁹. In their own longitudinal study, navigation and error correction took a large share of the dictation task early on ¹⁰, and users became more efficient with experience ¹¹.

The gap between speaking and dictating is about more than the words themselves. A recognizer that hears you perfectly still leaves you to decide the sentence, move the cursor and correct errors. The 8 to 15 figure comes from older recognizers, so read it as a historical floor ⁸. The lesson holds: plan each sentence before you say it. The how to dictate guide covers the technique.

Dictation wins for first drafts, messages and notes

When the goal is volume and the words will be revised anyway, dictation gives the most speed for the least effort. Three uses fit that pattern.

• First drafts of articles, reports or chapters, where the argument matters now and polish comes later. The does dictation change how you write guide covers the trade-offs in the sentences that result.
• Messages and replies. The phone study tested short message entry, so its results apply most directly to texts and short notes ¹.
• Notes and ideas, where capture speed matters more than finish. Speaking an idea aloud is a method in its own right, described in the glossary entry for oral composition.

Typing wins for precise edits, code, tables and quiet shared rooms

Precise edits are the first case for the keyboard. When you correct a dictated error by speaking again, the correction can fail too. In a simulated dictation interface, Vertanen found that spoken corrections are often misrecognized themselves, which costs time and patience ¹². A swapped name, a changed comma or a word in the middle of a paragraph is usually faster to fix with keys and a cursor. Homophones, words that sound alike, are one kind of error to watch for; see the glossary entry for homophone.

Code and symbols are the second case. Spoken punctuation covers the common marks, as the dictation punctuation guide explains. Code is dense with brackets, quotes and names that the recognizer has not seen, so typing is the default there. This is a judgment from how code is written, not a measured result.

Tables and layouts are the third case. A keyboard moves between cells with one key, while a dictated command has to name each move. Typing is the simpler route for dense grids.

Quiet shared rooms are the fourth case. The phone study was run in a laboratory ¹, and this guide's sources do not measure dictation in an open office or a busy room. Until you have tested your own room, keys are the safer default when colleagues sit close by. The dictation and background noise guide and the microphone guide cover what changes the result.

Editing dictated text costs time that speed comparisons usually leave out

A speed comparison stops when the words are on the screen. Working text stops later. Vertanen states that, when dictating with a large vocabulary recognizer, correcting errors dominates the task time ¹³. The phone study does count the time spent correcting errors during entry ¹, but it used short phrases, not a full document that must be proofread afterward.

The fair comparison is the time from first word to finished text. For dictation, that is the speaking time plus correction time plus proofreading. For typing, it is the typing time plus proofreading, with corrections made as you go. Dictation saves time when the correction load stays light. Heavy correction, from technical vocabulary, names, or a noisy microphone, can take the saving back. Even in that study, speech left more errors in the final text: 1.30 percent uncorrected against 0.79 percent on the keyboard ².

Two practical limits affect editing rhythm. In Apple's current Mac guide, Dictation stops by itself when it detects no speech for 30 seconds ¹⁴, so a long pause to think ends the session and you must start again. Recognizer accuracy is often reported as word error rate, and the overview of speech recognition explains how that figure is calculated. The dictation mistakes guide covers how to reduce them.

The best workflow combines voice for drafting and keys for fixing

The strongest pattern is not a choice between the two. Use voice to produce, keys to correct, and the keyboard again to proofread. The steps below follow the Apple and Microsoft documentation for each platform.

1. Draft by voice in one pass. Speak whole sentences and resist the urge to edit mid-draft. The how to dictate guide covers the rhythm.
2. Type during dictation if your Mac has Apple silicon. On a Mac with Apple silicon, you can type while dictating. The cursor stops pulsing while you type and resumes when you stop, so you can keep dictating ¹⁵.
3. Restart after silence. On a Mac, Dictation stops after 30 seconds without speech ¹⁴. Start it again with the shortcut before you continue.
4. Check the connection on Windows. Windows voice typing needs an internet connection, a working microphone and a cursor in a text box ¹⁶. Confirm it works before you rely on it, following the Windows voice typing guide.
5. Fix with the keyboard, then proofread. Use keys and the mouse for corrections and the final read. Keep spoken commands for simple moves, as the dictation commands guide describes.

Yap, which we make, is an on-device dictation app for Mac and Windows that follows the same pattern: dictate, then correct by keyboard. For many people the built-in tools are enough, and the Mac dictation and Windows dictation overviews cover them in depth. You can see what Yap includes on the download page.

Your own typing speed decides how much dictation saves you

Typing averages depend on who was measured. In a study of 136 million keystrokes from an online typing test, the average participant typed 51.56 words per minute ¹⁷. Those were self-selected volunteers who took a test online, so the average is higher than the population figure ¹⁷. The average typing speed guide gives the population data.

Time per word is the simple arithmetic behind the choice. Each word takes a fixed time to type and a fixed time to speak, and dictation saves the difference between them. The slower you type, the larger the keyboard time per word and the bigger the saving from speech. A fast typist gains less, because the keyboard time per word is already small. The speaking vs typing speed guide compares the two rates in more detail. The same trade-off on a phone is covered in the iPhone dictation and Android voice typing guides.

Test both methods on your own writing before you switch

The studies above use short phrases, lab recognizers and volunteers. Your own writing decides the answer for you, and a short test is more reliable than a general rule. Use a piece you would normally write, such as a status update, a reply or a short section of a report.

1. Pick one piece of text and write it twice. Write the first version by voice and the second by keyboard, on the same topic, so the thinking is comparable.
2. Time the draft stage only. Note how long each version took to get onto the screen, before any editing.
3. Count the corrections in each draft. Count the words you changed while drafting, and the errors you found later. Note which kinds of error appear, such as homophones, names or punctuation, because each kind has a different fix.
4. Time the proofread. Read each version once, with the keyboard, and time it. This is the stage that the speed figures leave out.
5. Add the totals. The method with the lower total for your kind of text is the one to use for that task. Repeat the test for a different kind of text, such as technical notes, before you generalize.

If dictation wins on drafts but loses on proofreading, the answer is the workflow below: speak the draft and type the fixes.

What the research does not settle yet

The studies in this guide answer some questions and leave others open. The phone study measured short messages in a laboratory, and its authors ask for study in non-laboratory settings ⁴. The longitudinal dictation study is the oldest source cited here, and upper-bound recognition performance has improved greatly since earlier systems ⁵. The speed of conversation and the speed of composing are measured in different tasks, so the gap between them is an estimate rather than a single measured number ⁶.

None of the sources used here measures a full adult document written in each method, proofread afterward, in a noisy room. That is a common case, and it is where your own test above matters most. The dictation background noise guide explains how room sound changes recognition, and the dictation mistakes guide explains how to see where errors come from.

Accuracy also varies by engine and speaker. Radford and colleagues report that, compared with humans, their Whisper models approach human accuracy and robustness on benchmarks ¹⁸. A benchmark is not your microphone, accent or room, so test with your own voice before you judge.

The short rule: dictate when you are producing words, and type when you are editing, when the text is technical or dense, or when the room is quiet. For two related questions, see When is typing better than dictation? and Can you dictate and type at the same time?. On a Mac with Apple silicon, the answer to the second is yes, as the fact above shows.