Voice writing means writing by speaking. You say your draft out loud, software turns it into text, and then you do what writers have always done: read it, cut it and shape it. The speaking replaces the slowest part of typing, which is getting the first version out of your head.
The phrase has a second, older meaning in court reporting, which I'll explain near the end. Most of this guide is about the everyday practice: how to dictate drafts that are worth editing, and how to edit them well.
I make LumeVoice, a dictation app, and I write most of my own first drafts this way.
Why writing by voice works#
Two things make voice writing appealing.
Speed of capture. Most people speak much faster than they type. A study by Stanford University, the University of Washington and Baidu found speech input on a phone was about three times faster than typing on the phone's keyboard (Ruan et al., 2016). That gap matters most in a first draft, where the goal is volume, not polish. For more on what that means in practice, see voice dictation vs typing.
Separating drafting from editing. When you type, it's tempting to fix each sentence before writing the next one. When you speak, you can't easily do that, so you keep going. Many writers find that this separation is where the real benefit lies: you get the whole idea down, then you improve it.
It also suits people for whom typing is hard, whether because of pain, injury, dyslexia or anything else. Dictation can help reduce typing; our accessibility page has more.
Drafting habits that make spoken drafts good#
Talk to one person#
Pick a real reader: a colleague, a client, a friend who asked you a question. Explain it to them. Spoken language aimed at a person comes out clearer than spoken language aimed at "the reader". If you catch yourself sounding like a press release, stop and imagine their face.
Outline out loud first#
Before the draft, dictate a rough list of the points you want to make, in any order. Then read it, reorder it with the keyboard, and use it as your map. Speaking a list is fast, and it stops you wandering mid-draft.
One section at a time#
Long monologues drift. Dictate one section or one point, stop, glance at it, then do the next. Short bursts also make it easier to spot a recognition error while you still remember what you meant.
Don't fix while you talk#
If a sentence comes out wrong, say it again the right way and keep going. Many AI dictation tools apply spoken self-corrections automatically ("on Tuesday, actually Wednesday" becomes "on Wednesday"). If yours doesn't, you'll fix it in the edit. Either way, stopping to tidy breaks your flow.
Mark what you'll come back to#
Say a marker you'll find later, like "check this" or "figure needed", where you're unsure of a fact. Then keep talking. A quick search for the marker during editing finds every gap.
Set up your tool once#
Add names, product terms and jargon you use to your tool's custom vocabulary so they come out spelled right. Use a decent microphone close to your mouth; our microphone guide covers the options.
The editing loop#
A spoken draft is raw material. Here's a loop that turns it into finished writing.
- Read it through once without changing anything. Just notice where it's strong, where it repeats itself and where it loses the thread.
- Fix the structure with the keyboard. Move paragraphs, delete repeats, add headings. Structure is visual work, and a keyboard and mouse are better at it than your voice.
- Re-dictate weak paragraphs. If a paragraph is muddled, don't fiddle with it word by word. Select it, explain the point out loud again in two or three sentences, and replace it. The second attempt is usually clearer, because now you know what you're trying to say.
- Tighten sentences. Spoken drafts lean on phrases like "the thing is" and "basically". Cut them. Shorten long sentences.
- Check facts and names. Speech recognition predicts words; a confident wrong guess looks just like a right one. Look hardest at names, numbers and technical terms.
- Read it aloud. It sounds odd to finish a voice-written draft by speaking again, but your ear catches clumsy rhythm your eye skips.
An illustrative example#
This is invented, to show what a spoken draft and a cleaned version look like. It isn't a measured result.
Spoken (illustrative): "So the main reason, um, the main reason people abandon the onboarding is the, uh, account setup step, it asks for too much, like company size and role and everything, before they've seen anything useful, so I think we should move those questions to after the first project, actually after the second session."
After AI cleanup: The main reason people abandon onboarding is the account setup step. It asks for too much, like company size and role, before they've seen anything useful. I think we should move those questions to after the second session.
After your edit: Most people who abandon onboarding leave at account setup, which asks for company size and role before they've seen anything useful. We should move those questions to the second session.
The cleanup removes the noise; the edit makes it yours.
Where voice writing works best#
- First drafts of articles, reports, essays and chapters. Writers' tips: dictation software for writers.
- Emails and messages, where the spoken version is often close to final. See how to write emails faster using voice.
- Notes and journaling, where nobody else reads it and speed is everything.
- Scripts and talks, which are meant to be spoken anyway. See our content creator guide.
- Thinking out loud, for example explaining a topic to learn it, as in the Feynman technique.
It's harder for tightly structured documents such as tables, code and legal clauses, where every character matters. Dictate the surrounding prose and type the rest.
"Voice writing" in court reporting#
If you searched for voice writing because of a court reporting course, you're looking for something different, and it's worth explaining properly.
In court reporting, a voice writer makes the official verbatim record of a proceeding by repeating every word spoken, including who said it, into a stenomask: a hand-held mask with a built-in microphone that muffles the reporter's voice so nobody else in the room hears it. Speech recognition software turns that dictation into text, either in real time or for a transcript prepared afterwards. The Bar Association of San Francisco describes voice writers repeating "verbatim every word into a stenomask" and credits Horace Webb, a shorthand writer, with developing the method in the early 1940s (Cram and Costa, 2024).
In the United States, the National Verbatim Reporters Association (NVRA) is the professional association for voice writers, and it offers certification. It describes voice writers as professionals trained to capture the spoken word and convert it to text with computer-aided transcription software, working in courtrooms, depositions and broadcast captioning, among other settings (NVRA). Rules on who may report in court vary by state; California, for example, opened its Certified Shorthand Reporter exam to voice writers through a 2022 law, according to the same Bar Association article.
It's a demanding profession with its own training, equipment and exams. The everyday voice writing in this guide borrows the name and the core idea, speaking instead of typing, but it isn't a substitute for court reporting, and LumeVoice is not a court reporting tool.
Where LumeVoice fits#
LumeVoice is an AI dictation app for Mac, Windows, iPhone and Android. Hold fn on a Mac or Ctrl + Win on Windows in any app, talk, and let go; or double-press for hands-free mode to dictate a whole section. Filler words disappear, self-corrections are applied, punctuation and paragraphs are added, and spoken lists become real lists. It needs an internet connection, and your audio isn't stored after processing.
Draft your next piece out loud
Talk through the first draft, then edit on the page. LumeVoice works in any app on Mac and Windows, with iPhone and Android on the same plan, and new accounts get 3,000 free words, no card required.
The short version#
- Voice writing is drafting by speaking and editing on the page.
- Talk to one person, outline first, and dictate one section at a time.
- Don't fix while you talk; fix structure by keyboard, then re-dictate weak paragraphs.
- Check names and numbers, and read the final draft aloud.
- In court reporting, "voice writing" is a separate, certified method using a stenomask.




