How to dictate a book (and actually finish it)

You can dictate a book — authors have been drafting entire novels by voice for decades, and modern AI dictation has removed most of the friction that once made it a specialist's craft. Most people speak at 130–150 words a minute and type at around 40, so a spoken first draft arrives roughly four times faster than a typed one. What that number hides is that dictating a book is a skill with a method: the authors who finish outline their scenes, dictate in short sessions and edit with their eyes. The ones who quit tried to "just talk". Here is the method.

Why authors dictate (it is not just speed)

Start with arithmetic. An 80,000-word novel typed at 40 words a minute is more than 33 hours of pure typing; spoken at a comfortable 140, it is under ten. Nobody sustains pure output speed — you pause, think, re-speak — but even at half that ratio, the drafting hours of a novel shrink by weeks.

Speed is only half the appeal. Dictation separates writing from the desk: science-fiction author Kevin J. Anderson has drafted most of his 190-plus books by talking into a handheld recorder while hiking — around ten thousand words over a single hiking weekend, by his own account — and co-wrote On Being a Dictator, a book about the method. For writers with RSI, a long commute, or twenty free minutes and no laptop in reach, speaking is often the difference between writing and not writing.

The third reason is the strange one: speech is a forward-motion machine. You physically cannot fiddle with the previous sentence while saying the next one. For perfectionists who edit every line to death mid-draft, that constraint alone can transform daily output.

Pick your workflow: live or recorded

There are two ways to dictate a book, and starting with the wrong one is the top reason people quit.

• Live dictation — you speak, and text appears in Scrivener, Word or Google Docs as you go. You keep visual contact with the draft, catch engine mistakes immediately, and stay oriented inside the scene.
• Record now, transcribe later — you talk into your phone or a recorder on a walk, then transcribe the audio at your desk. Deeper flow and total freedom of movement, but you come home to a wall of raw transcript that still needs shaping.

Start live: watching your speech become prose teaches you to speak in complete written sentences far faster than a recorder does. Graduate to recorded sessions once dictating feels natural — most dictating authors end up mixing both.

The chapter-by-chapter method

Dictation punishes improvisation harder than typing does: at a keyboard you can stall quietly, but out loud a stall becomes silence and the session dies. The fix is a light outline — enough to keep talking, never so much that you are reading.

• Before each session, jot three to five bullets for the scene: who wants what, what goes wrong, where it ends. This is exactly what Anderson carries on a hike — a short outline for the two or three chapters he intends to speak.
• Dictate one scene per session, fifteen to twenty-five minutes. A scene fits entirely in your head, which is what fluent speaking requires; a full chapter often does not.
• Never stop to fix. Misspoke a sentence? Say it again cleanly and keep moving — repairs belong to the edit, and hunting for the cursor mid-flow is how momentum dies.
• Close by dictating a note to your future self: "Next scene: Marta finds the letter, ends on the doorbell." Tomorrow you start talking instead of staring.

The pitfalls — learn them now, not in chapter 12

Proper nouns. Every speech engine guesses at invented names, and it guesses differently on Tuesday than it did on Monday: "Aelric" arrives as "a relic". Pick character names you can actually pronounce, add them to your tool's custom vocabulary if it has one, and run a find-and-replace consistency pass at the end of every chapter — not at the end of the book.

Dialogue and punctuation. Raw speech-to-text makes you speak the punctuation — "comma", "period", "open quote" — which in dialogue-heavy fiction is more tiring than typing. This is the single strongest argument for an AI-formatted dictation tool that punctuates from your natural delivery instead.

The spoken-draft voice. Dictated prose runs warmer and looser than typed prose — longer sentences, more connective tissue. That is fine; it is what edits are for. Budget one extra tightening pass, and proofread names, numbers and dates with your eyes: those are the three places every engine still slips.

The tools, honestly

Free built-ins. Apple Dictation and Google Docs voice typing cost nothing and are the right way to test whether dictation suits you at all. Apple Dictation auto-punctuates from your pauses (imperfectly, and you can still speak "comma" and "period" yourself for precision); Google Docs voice typing still expects the punctuation spoken. Either way you get close to your literal words, and accuracy on names drifts over long sessions. Fine for a chapter; wearing for a book.

The classic. Dragon Professional is the tool dictating authors of the 2000s swore by, and its voice-command control is still the deepest around. Today it is Windows-only — the Mac version was discontinued in 2018 — at ~$699 one-time (price checked July 2026; confirm with the vendor). And it transcribes what you said rather than cleaning it up.

Modern AI dictation. Swifly is an AI voice dictation app for Mac, Windows, iPhone and Android that turns speech into polished, correctly formatted text in any app — with instant translation. For book work that means you hold a hotkey and talk straight into Scrivener, Word, Docs or Ulysses, and an AI pass strips the false starts and punctuates from your delivery, no spoken commas. One honest limit: processing happens in the cloud, so there is no offline mode — on a signal-free trail, do it the Anderson way and record first, then transcribe back home. There is a free plan to start.

The takeaway

At speaking speed, 1,000 words is well under ten minutes of actual talking. Make it a daily twenty-minute session — pauses, re-speaks and outline glances included — and an 80,000-word first draft lands in about three months, spoken. The editing is still entirely yours. The typing no longer is.