I Dictated My First 5,000-Word Hour. Here's What the Page Looked Like.

Established Novelists Have Been Doing This for a Long Time

This isn't a new trick. Henry James dictated most of his late novels to a typist when rheumatism made writing by hand painful. Milton dictated Paradise Lost after going blind. Barbara Cartland dictated from a couch and produced more than seven hundred novels. Richard Powers writes entire drafts on a voice recorder while walking. The modern twist is just that the typist is now a neural network running on your laptop.

Contemporary working novelists who talk openly about dictation — Kristine Kathryn Rusch, Joanna Penn, Monica Leonelle, Cecilia Tan — consistently report the same pattern: their output roughly doubles or triples once the muscle memory is in place. Not on day one. Usually around week three or four.

The Math: Why 5,000 Words Per Hour Is Actually Conservative

Normal conversational speech runs 120 to 160 words per minute. Do the arithmetic: sustained for an hour, that's 7,200 to 9,600 words. Nobody sustains it for an hour because you have to stop and think, re-state sentences, and sometimes stare at the wall.

With realistic pauses — which, counterintuitively, are fewer than most people expect once you've done this a while — experienced dictators land around:

• 2,500–3,500 words per hour for a scene you haven't fully planned
• 4,000–5,500 words per hour for a scene you've thought through
• 6,000+ words per hour for writers who dictate straight from a detailed outline

For reference, a fast typist with a strong session writes maybe 1,200–1,800 words per hour in first-draft prose. Most novelists type closer to 800–1,200. So yes, voice dictation is two to five times faster — but only after you've paid the learning cost.

What a Dictated Draft Actually Looks Like

Here is something I didn't see in any of the “dictate your novel!” blog posts I read before I started recommending this to fiction writers. A dictated first draft looks different on the page than a typed one.

Typed drafts tend toward measured prose, carefully balanced sentence lengths, precise diction, and dense description. You make word-level decisions in real time because the hand is slow enough to let the brain deliberate.

Dictated drafts tend toward dialogue, rhythm, and momentum. They move faster scene-to-scene. The sentences are shorter. The description is lighter. Characters talk more, because talking is what your mouth is already doing, so dialogue comes out more naturally than exposition.

Neither is better. They're different shapes of draft. But if you switch from typing to dictating, your revision process has to adjust. A dictated draft usually needs more structural revision and more descriptive layering than a typed one. You're trading drafting speed for revision load.

Most dictating novelists I've talked to say the trade is worth it — they'd rather write the whole book in three months and revise it in three than write it in twelve months and revise it in two. But the shape of their year changes.

The Three-File Workflow

Almost every fiction writer who dictates consistently uses some variation of a three-file structure. On Mac, the typical setup is:

File 1: The outline file

A short document with beats for the scene you're about to dictate. Not a formal synopsis — more like “Marta confronts her brother at the harbor. He denies it. She notices the scar on his hand. Ends on her leaving without saying goodbye.” Four or five sentences. You read this before you start dictating.

Why: dictating without any scaffolding leads to a kind of aimless drifting where you describe the harbor for 400 words and then realize you forgot what the scene was about. The outline anchors you.

File 2: The dictation target

This is where the words actually land. Most novelists dictate into Scrivener, iA Writer, or a plain text file. Google Docs works too, but can lag on long dictation sessions.

Whatever you use: make sure the dictation tool you're running types directly into the active text field. EmberType works in any text input, which means you can dictate into Scrivener's chapter view the same way you dictate into a browser comment. No integration, no plugin.

File 3: The reference/continuity file

A running document of character names, place names, and small continuity details that you want pronounced and spelled consistently. This pulls double duty: it lets you add those names to your dictation tool's custom vocabulary, and it keeps you from having two characters named “Arin” without realizing it.

Custom Vocabulary for Novelists

Speech recognition models are trained on general English. They will do perfectly reasonable things with “John,” “Maria,” and “London.” They will do increasingly unreasonable things with “Eirian,” “Coldharbour,” “Ixtoc,” or anything you invented.

Every Mac dictation tool worth using has a custom dictionary. Populate it with:

• Every proper name in your novel — characters, places, ships, weapons, gods, made-up words
• Pronunciation hints where available — some tools let you specify “pronounced like [X]”
• Foreign language or genre-specific terms — military technology, magical systems, historical titles
• Common phrases you use that might get mis-parsed — “aft deck,” “port bow,” etc.

Ten minutes of vocabulary setup at the start of a new novel saves you an hour a week of correcting the same misrecognitions.

Dialogue and Punctuation Without Breaking Flow

The thing that scares most writers off of dictation is the idea of saying “comma” and “period” every few seconds. Here's the good news: modern speech recognition (Whisper-based tools included) punctuates automatically based on the rhythm and structure of your speech. You do not need to dictate punctuation.

You do need to speak with intent. A pause at the end of a sentence becomes a period. A small pause within a sentence becomes a comma. If you trail off without punctuation, the transcription will trail off without punctuation. Speak your sentences as complete thoughts and let the rhythm land.

For dialogue, there are two schools:

1. Verbal markers: Speak “open quote” and “close quote,” say the tag as “said Marta,” and clean up in revision. Most dictation tools recognize “open quote/close quote” as punctuation commands.
2. Inline cleanup: Dictate the dialogue naturally without markers, then do a fast pass after each scene to add quotation marks. This is faster to dictate and slower to clean up.

Most novelists settle on a hybrid: verbal markers during exchange-heavy scenes, inline cleanup for monologue-heavy sections. Do whatever keeps the scene alive in your head.

Where You Dictate Matters More Than the Software

Voice drafting is a weirdly physical activity. After the first hour, your face gets tired. After the first day, you notice that certain rooms work better than others. A few practical notes:

• Rooms with soft surfaces transcribe better than rooms with hard walls. Hard walls produce echo that confuses the speech model. Bedrooms and carpeted offices are ideal; bathrooms and kitchens are hostile.
• Cars are surprisingly good for dictation — the interior is dampened — but only if you're stationary. Dictating while driving is a bad idea for obvious reasons.
• Walking outside is what a lot of novelists say produces their best drafting. Joanna Penn and Richard Powers both talk about walking-dictation as their primary drafting mode. The body moving seems to help the prose flow.
• Whispering does not work well. Speech models are trained on normal conversational volume. If you mumble, the model guesses — often badly. You do have to commit to talking at roughly the volume you'd use in a coffee shop.

The Productivity Paradox Nobody Warns You About

Here's the thing that surprised every novelist I've talked to about this: your effective output does not go up by 3x the first month, even though your drafting speed does.

The reason is that dictated prose needs more revision. The clean 1,200 words/hour you'd type in a session was already somewhat edited in your head as you went. The 4,000 words/hour you dictate is rawer; at least a third of them will need substantial rework.

Do the math: in hour one, typing, you produce 1,200 polished words. In hour one, dictating, you produce 4,000 rough words. In hour two, typing, you produce another 1,200 polished words (total 2,400). In hour two, dictating, you spend revising, and you end up with maybe 2,800 polished words.

Over time, the dictation path wins. Over a short burst, it's closer than the raw numbers suggest. Go in knowing that the productivity gains are real but deferred — they show up in the second draft, not the first.

The Mac-Specific Setup

1. Download EmberType. Free 7-day trial; Apple Silicon Mac, macOS 14+.
2. Pick your Whisper model. For long drafting sessions, Medium is a good balance of speed and accuracy. For final proofing passes or unusual proper nouns, switch temporarily to Large v3.
3. Add every proper noun from your novel to the custom dictionary before you start drafting.
4. Set a global shortcut — Option+Space is common. Make sure it doesn't conflict with Scrivener or your note app shortcuts.
5. Do a five-minute test session. Dictate a paragraph of your current scene into a test document. Check how the punctuation lands. Adjust your speaking pace.
6. Target 30 minutes of real drafting for your first session. Build up from there. Most novelists need about ten hours of practice before dictation stops feeling unnatural.

Bottom Line

Dictation is not a cheat code. It's a different way of getting words on the page — one that trades keyboard speed for a draft that comes out messier. Novelists who switch tend to stay switched, because the extra revision load is often worth it for the lower-friction drafting. But it's not magic.

If you write on a Mac, have some proper nouns you care about, and are willing to spend two weeks past the initial awkwardness, dictation can do exactly what everyone says. The 5,000-word hour is a real thing. The page that hour produces just won't look like the one your typing hand would have made.

---