everyday mac tools

How much faster dictation is than typing, and where it is not

· 6 min read

Almost everyone speaks faster than they type, often by a factor of two or three, and you can confirm that on your own Mac in ten minutes. What that does not tell you is whether dictation makes you faster at your actual work, because the number that matters is finished text per hour, not words per minute. On loose prose the gain mostly survives editing. On code, tables, numbers and anything you are revising rather than drafting, it disappears and then goes negative.

Measure it yourself rather than trusting a figure

Published typing and speaking averages are not about you. Two tests, ten minutes, and you will have real numbers.

Test one: raw transcription. Pick a paragraph of about 200 words from something you wrote. Start a timer, type it, stop. Then start again, dictate the same paragraph, stop. Divide 200 by the minutes each took. This is your ceiling, and the ratio between the two is usually larger than people expect.

Test two: the one that counts. Write something real, from nothing, twice. A reply you owe somebody, or a paragraph of notes. Once by keyboard, once by voice, on different but comparable tasks. Time it from starting to the point where you would actually send or save it, corrections included.

Test one measures your mouth against your fingers. Test two measures your workflow, and it is the only one worth deciding on; the ratio there is usually much smaller, and still positive for prose.

Some dictation software tracks this for you as you go. Voice Keyboard Pro shows words per minute and usage over time, which is a fairer number than a one-off test because it is measured across weeks of real work rather than a paragraph you chose.

Where the speed comes from, and where it leaks away

Speaking is something you have done since before you could read, at a pace set by thought rather than by hands. Typing puts a mechanical stage between the sentence and the screen, and for most people that stage is the bottleneck.

The gain then leaks away in three places.

Correction. Suppose recognition gets 97 words in 100 right. On a 500-word draft that is fifteen wrong words to find and fix, and finding them means reading the whole thing carefully. If they are proper nouns you may be retyping rather than correcting. That reading pass is real time, and it is the single biggest reason the impressive test-one number does not show up in your day.

Formatting. Anything that is not a stream of sentences costs you. Lists, headings, indentation, quotation marks, moving a paragraph. Voice commands cover the common ones (“new line”, “select all”, “undo” and similar), but they are slower than the keystroke you already know.

Composition. Speaking a first draft is different from typing one. You cannot see the last sentence while you produce the next one, and that changes what you write. For some people the result needs more editing, which eats the gain. For others the drafts are looser and better, and they were over-editing while typing anyway.

Where dictation is genuinely faster

First drafts of prose. The clearest win. Nobody is editing yet, the sentences are long, and speed of production is the constraint.

Messages and replies. Short, prose-shaped, and there are a lot of them. This is where a normal working day contains the most dictatable text. Clearing an inbox by dictating the replies sets out a way to run a whole inbox this way.

Notes while your hands are busy or your eyes are elsewhere. Reading a document and capturing thoughts, working through a physical task, standing at a whiteboard.

Anything long that you would otherwise put off. The honest gain here is not words per minute, it is that a 600-word update gets written today instead of Thursday. Lower activation energy is a genuine productivity effect even though it does not show up in a speed test.

Working around hand pain. Where typing hurts, the comparison is not speed against speed. Any pace you can sustain beats a faster pace you cannot.

Where it is slower, and worth admitting

Code and shell commands. Punctuation density is the problem: braces, brackets, underscores, camel case, exact case sensitivity. You can dictate code, and it is slower than typing it for anyone who can type.

Numbers, tables and structured data. Formatting is ambiguous by voice and the error cost is high, because a wrong digit does not look wrong.

Editing existing text. Voice is good at producing text and poor at surgery. Selecting the third sentence and changing one clause is a keyboard job.

Precise punctuation. Automatic punctuation is good at commas and sentence ends and weak at anything else: colons, dashes, quotation marks, parentheses. Saying them explicitly works and is slow.

Names and specialist terms. Every unfamiliar noun is a potential correction. A custom vocabulary raises the hit rate a lot; dictating names, jargon and technical terms accurately covers building one.

A room where you cannot speak comfortably. Accuracy falls, corrections rise, and the arithmetic flips. Dictating in a noisy room: microphone and technique covers how much of that is recoverable.

Latency is a different number from speed

Words per minute is not the only thing that decides whether dictation feels fast. The pause between finishing a sentence and seeing it appear matters more than it should, because that pause is where you lose your place in the sentence you were about to say next.

macOS Dictation shows words as you speak, which keeps you connected to the output. Systems that transcribe after you stop are quicker per word and can feel slower if the wait is long. Voice Keyboard Pro releases most dictations at the cursor within 200 to 500 ms of letting go of the key, which is short enough to keep a train of thought going. The general point applies whatever you use: a system that is more accurate but pauses for three seconds a sentence can end up feeling worse to work in.

A realistic split for a working day

What most people who stick with dictation end up doing:

  • Speak: mail, chat, meeting notes, first drafts, journal entries, comments and reviews, anything long and prose-shaped.
  • Type: code, terminal commands, spreadsheets, search fields, passwords, file names, form fields, and every edit to text that already exists.
  • Switch mid-task without ceremony. The keyboard is right there. Dictating a paragraph and then typing the three-word title is normal, and stopping to think about which mode you are in costs more than the switch.

The realistic claim is not that you will work twice as fast. It is that a specific and quite large slice of the day, the prose slice, gets quicker and costs your hands less, and the rest stays exactly as it was.

Questions

How long before I am as fast by voice as by keyboard? For simple replies, a few days. For drafting anything with structure, longer, because you are learning to compose out loud rather than just to operate the software. Judge it after a week of real work, not after an afternoon.

Is a paid dictation app faster than what macOS includes? Where it wins is usually not raw recognition speed. It is the things around it: a held key rather than a toggle, a custom vocabulary that removes repeated corrections, and text landing in any app without fiddling. Those reduce correction time, which is where the day actually goes.

Does accuracy improve as it learns my voice? Do not count on a system quietly adapting. What reliably improves accuracy is the things you control: microphone position, a steady pace, and a vocabulary list for the words it keeps missing.

Can I dictate and type at the same time? Yes, and it is the normal way to work once the novelty passes. Hold the key for the sentence, release, type the URL, hold again. Treating dictation as a mode you enter and leave is what makes it feel slower than it is.