Editing your text by voice

Dictation is wonderful for getting words out. Editing those words is a different job, and it is worth being honest about how well voice handles it. The short version: voice is excellent for drafting and only sometimes good for editing. Knowing where the line falls will save you a lot of frustration.

This post lays out what voice editing realistically can and cannot do, and suggests a workflow that plays to the strengths of both your voice and your hands.

Why editing by voice is hard

Drafting and editing are not the same activity. Drafting is generative — you produce a stream of words and your voice is a fast, natural way to do that. Editing is positional and precise. It means putting a cursor between two specific characters, selecting exactly the right span of text, and making a small surgical change.

Speech is bad at precision of that kind. To delete one word in the middle of a sentence by voice, you have to describe which word, unambiguously, in language the tool can parse. Your hand can do the same thing with one click. The mouse and the arrow keys are pointing devices, and pointing is exactly what editing needs.

So the honest framing is this: voice editing is not a replacement for hands-and-eyes editing. It is a partial tool with a few genuine strengths and several real limits.

What voice editing actually does well

There are cases where speaking a change is genuinely faster or more comfortable than reaching for the keyboard.

Appending. Adding a sentence to the end of what you have is just more dictation. This is the most reliable "edit" of all, and it is really still drafting.
Replacing a whole chunk. If a paragraph is wrong, it is often faster to delete it by hand and re-dictate it fresh than to repair it word by word.
Dictating into a found location. Use your mouse to click exactly where you want text, then dictate. The hand does the precision part; the voice does the words.
Simple, unambiguous commands in tools that support them — "delete that", "new paragraph", "scratch that" right after a misrecognition.

Notice the pattern. Voice editing works best when it does not have to locate anything precisely, or when your hand has already done the locating.

What voice editing struggles with

Be realistic about the rest.

Precise selection. "Select from 'however' to 'therefore'" works in some dedicated accessibility software, but it is fragile when either word appears more than once, and most general dictation tools do not support it well at all.
Small surgical fixes. Changing "their" to "there", adding a missing comma, fixing one transposed letter — these are faster with a click and a keystroke than with any spoken instruction.
Navigating long documents. "Go to the third paragraph" is vague and error-prone. Scrolling and clicking is instant and exact.
Formatting. Bold, italics, headings, indentation. A few tools handle some of this; most do not, and it is rarely worth the spoken syntax.

If you find yourself constructing an elaborate sentence just to describe a one-character fix, that is the signal to stop and use your hand.

A sane workflow: draft by voice, edit by hand

The most productive approach treats the two modes as separate phases rather than trying to do everything by voice.

Phase one — draft by voice, fast and loose. Speak the whole thing through without stopping to fix anything. Misheard word? Leave it. Clumsy sentence? Leave it. The goal of this phase is volume and momentum. Stopping to correct each error kills the flow that makes dictation fast in the first place, and you will revisit everything anyway.

Phase two — edit by hand, slow and careful. Now switch tools. Read the draft on screen, with your hands on the keyboard and mouse. Fix the misrecognitions, tighten the sentences, move things around. This is ordinary editing, and your hands are the right instrument for it.

Phase three — re-dictate, not repair. When a whole sentence or paragraph needs to be rebuilt rather than nudged, do not edit it. Select it, delete it, and dictate a fresh version. Re-dictating a clean sentence is usually faster and produces better writing than patching a broken one.

The one exception worth keeping is the immediate correction. If you misspeak and notice right away, fixing it on the spot — by voice or by a quick keystroke — is fine, because you have not lost your place yet. It is the delayed, hunt-for-the-error edits that belong in phase two.

Catching misrecognitions efficiently

Since editing is mostly hands-and-eyes, the practical skill is reading a dictated draft well.

Read it once, end to end, before fixing anything. You will catch more if you understand the whole passage first.
Watch for real words in wrong places. Dictation errors are rarely gibberish. They are plausible words — "form" for "from", a name spelled as a common noun. These slip past a quick skim, so read for sense, not just for typos.
Check names, numbers, and technical terms specifically. These are the highest-risk items in any transcript. Give them a deliberate second look. More on this in dictating technical terms.
Trust your ear. If a sentence sounds slightly off when you read it back, it probably is. Re-dictate it.

When to skip voice entirely

Sometimes the right call is not to dictate at all. A dense paragraph that needs three precise edits is faster to fix purely by keyboard. A short reply — a couple of words into a form field — is faster to type. Voice earns its place on the generative work: first drafts, long messages, notes, anything where you have something to say and just need it on the page.

Treating voice as a drafting tool rather than a universal interface is not a limitation to apologize for. It is just using each tool for what it is good at.

The honest summary

Voice editing has a few genuine wins — appending, replacing whole chunks, dictating into a spot your hand has already chosen — and a lot of cases where a click and a keystroke simply win. The productive pattern is to keep the phases separate: draft by voice with momentum, then edit by hand with care, and re-dictate anything that needs more than a nudge.

Lispr is built for the drafting phase. Hold the right Option key, speak, release, and the text lands at your cursor in whatever app you are in. When it is time to edit, put your hands back on the keyboard — that is still the best editing tool you own. Used that way, voice and hands complement each other instead of competing.