Novelists, journalists, newsletter writers — the ones who produce a lot of words — have quietly moved their first-draft phase to voice. Not all the writing. Editing still happens at a keyboard. But the part where you have to get words out for the first time? Speaking is faster, and the internal editor stays quieter when you can't see what you just said.
This is the setup on Mac, what works, and where it breaks.
Why voice changes drafting specifically
Three reasons:
Speech is faster than typing. Most people speak at 130–150 words per minute, type at 40–60. A 1500-word essay draft is around 10 minutes of speech versus 30+ minutes of typing. The speed alone isn't the whole story — but it's part of it.
The internal editor stays quieter. When you type, you can see every word as it lands. The instinct to backspace and rewrite kicks in instantly. When you speak, the words are gone before you can second-guess them. You keep moving. First drafts need that.
Voice draws out more conversational, more honest prose. The voice you use when you talk is closer to how you actually think than the voice you use when you type. For nonfiction, this usually means a friendlier, more direct first draft. For fiction, it often means dialogue that sounds like dialogue and narration that has a natural rhythm.
Voice drafts are messier. Filler words, false starts, run-on sentences. You'll edit more in the second pass. The math still works out — the first draft happens fast, and the editing is something you'd do anyway.
What's needed on Mac
You need:
- A reasonably current Mac (Apple Silicon strongly preferred)
- A decent microphone (the MacBook built-in is okay; a USB mic or a headset is better)
- A quiet-enough room
- A dictation app
The app is where the choice is.
Apps for long-form writing on Mac
Apple Dictation. Free, on-device, ships with macOS. Works but built for short bursts — it auto-stops after about a minute of speech in older versions, longer in newer ones. For paragraph-by-paragraph drafting it's awkward but doable.
Dragon Dictate. Used to be the gold standard for professional dictation, especially in legal and medical fields. Nuance discontinued the Mac version years ago. If you're seeing recommendations for Dragon on Mac, they're outdated.
Vext — $49 once. Local, Apple Silicon only. Hands-free dictation for long passages, Enhance for cleanup, transcripts saved alongside. Built for the speed-but-clean output that writers want.
Superwhisper — $249 lifetime. Polished dictation with mode-based prompts. The "writing fiction" mode and "writing nonfiction" mode produce different output styles.
MacWhisper — €64 Pro. Originally a file-transcription tool; live dictation in Pro. Strong if you sometimes record yourself talking through ideas and want to transcribe afterward.
Otter, Rev, Descript — cloud-based, subscription. Strong for journalists who record interviews and need transcription, less ideal for solo drafting because of cost and cloud processing.
Voice Memos + Whisper later — record into Apple's Voice Memos app, transcribe later in MacWhisper or with the OpenAI Whisper CLI. Works but means you don't see the text until later. Good for walking-and-talking but not for "draft while sitting at the computer".
For the specific use case of "I want to draft a book chapter / essay / article on my Mac, by speaking, with the cleanup happening automatically", a local app with cleanup (Vext, Superwhisper, MacWhisper Pro) is the right answer.
Setting up Vext for long-form writing
The configuration we'd suggest for someone using Vext primarily for drafting:
- Install:
brew install muvon/tap/vext - Open Settings > Modes
- Enable Enhance with default Gemma 3 4B model
- Set up the Hands-free hotkey (separate from the push-to-talk one) — drafting wants toggle mode, not hold-to-talk
- Disable YOLO Mode (you don't want auto-Enter in a writing app)
- Pick the Whisper Medium model — best balance of speed and accuracy for prose
Then in your writing app (Scrivener, Ulysses, Pages, Word, iA Writer, whatever):
- Click into the document at the point you want to start writing
- Press the hands-free hotkey
- Talk through the section, start to finish, without stopping
- Press the hotkey to stop
- Cleaned-up text lands in the document
A 1000-word section is roughly 7 minutes of speech. Add 30 seconds for the cleanup pass to finish. You get a draft to edit.
How to actually talk through a draft
Two patterns work:
Outline first, dictate against the outline. Write a quick bullet outline of what the section needs to cover. Then dictate against each bullet, taking each one in turn. This keeps you on track and prevents the draft from wandering.
Storytelling mode. For more narrative content, just talk through it like you'd explain it to a friend. "So the thing about this is..." "Okay so what happens next is..." The conversational tone reads well once the cleanup pass removes the obvious filler.
A few tips that come up consistently:
- Don't pause to rephrase. If you say something wrong, just keep going and say it correctly. The cleanup pass usually drops the wrong version anyway. If it doesn't, fix it in editing.
- Speak punctuation if the cleanup is missing it. "Period", "comma", "new paragraph" all transcribe. Cleanup-equipped apps add most punctuation automatically, but for unusual breaks (em dashes, quotation marks) saying them helps.
- Skip dialogue formatting. Don't try to say "open quotation mark Margaret said close quotation mark" — say the dialogue normally, format it in editing.
- Embrace the rough first draft. If you try to make it perfect while dictating, you'll lose the speed advantage. The point is to get the draft out.
The editing pass
Voice drafts need more editing than typed drafts. This is unavoidable. The trade-off is that the editing happens to something that exists rather than to a blank page.
What you're editing for:
- Filler words the cleanup missed
- Sentences that sound right spoken but read awkwardly
- Repetition (we repeat ourselves more when speaking)
- Structure (paragraphs, section breaks, headings)
- The bits where you wandered off topic
For nonfiction, this editing pass typically takes 30–50% as long as it would have taken to type the original draft from scratch. So a 30-minute typed draft becomes a 10-minute voice draft + 12-minute edit = 22 minutes total. The savings compound across a full book or article series.
For fiction, the math depends on how much dialogue and how much description. Description tends to benefit most from voice. Dialogue is often faster to type because you're already in a specific voice that matters.
When voice doesn't help
A few honest cases where typing wins:
Highly technical writing. API documentation, code-heavy explanations, math, anything with a lot of named entities you'd rather type than dictate.
Short pieces. A 200-word blog comment isn't worth the voice setup. Type it.
Poetry or carefully constructed prose. If each word matters and you're already going slowly to get them right, typing is the right pace.
Writing in noisy environments. Cafés, shared offices, open homes. Voice requires either a quiet space or noise-canceling headphones with a good mic.
The first 2-3 weeks of trying it. Voice writing feels weird at first. You'll write less well at the start because you're learning a new motor pattern. Stick with it for two weeks before judging.
Workflows specific writers have settled on
Bloggers and essayists. Outline in Notion or Obsidian. Open the post draft in their writing tool. Dictate the body in 3-5 minute sections. Edit the same day.
Nonfiction book authors. Talk through one chapter section per writing session (~1000–2000 words). Edit the previous day's section before starting today's. The morning starts with feedback (editing) and ends with new output (dictation).
Journalists. Voice for the article draft after the reporting is done. Quotes typed because they have to be exact. The connective tissue between quotes — voice.
Newsletter writers. Dictate the main piece in 10–15 minutes. Edit. Format. Send. Total cycle is short enough that voice + edit is faster than typing for a daily or 3x-weekly newsletter.
The mental shift
The hardest part of voice writing isn't the technology. It's giving yourself permission to produce a worse first draft than you'd type, on the bet that the editing pass will fix it.
Most writers who try voice and stick with it describe the same thing: their typed drafts were already first drafts they'd edit anyway. The typed first draft just felt cleaner because they edited as they wrote — and that editing-while-writing was the slow part. Voice forces a clean separation between drafting (fast, messy) and editing (slow, careful). Once that separation feels normal, voice is faster overall.
The blank page problem doesn't go away. But it gets smaller when talking is the ask instead of typing.