Talking and writing by voice are not the same thing

The first time most people try to dictate something real — an email, a paragraph, a chapter — they speak the way they speak to a friend, watch a wall of meandering text appear, and quietly conclude that dictation isn't for them. The conclusion is wrong, but the experience is real. Learning how to write with your voice is a skill with its own technique, and nobody is born knowing it any more than they were born knowing how to touch-type.

The good news is that the technique is small. A handful of habits separate people who find dictation clumsy from people who draft entire documents by voice without a second thought. None of them require talent. They require knowing what to do with your mouth, and knowing what to leave for later.

Separate the drafting from the editing

This is the whole game, and almost everyone gets it wrong at first. When you write by hand or keyboard, you edit constantly — backspacing, rephrasing, nudging a comma — and you barely notice you're doing it. People bring that same reflex to dictation and it falls apart immediately, because you can't smoothly backspace your voice. So they stop, restart the sentence, stop again, and the experience feels like driving with the parking brake on.

The fix is to consciously split the two jobs. When you're speaking, only speak. Let the wrong words stand. If you misspeak, just say the sentence again and keep going — you'll cut the bad version later. Editing happens after, with your eyes, once the raw material exists. This feels unnatural for about a day and then becomes second nature, and it is the single biggest unlock. Drafting and editing are different mental modes, and dictation finally forces you to stop interleaving them, which is part of why dictated first drafts often come out faster than typed ones.

Speak in phrases, not in words and not in paragraphs

There's a sweet spot for how much to hold in your head before you say it. Word-by-word is too granular — you sound robotic and the rhythm dies. Trying to compose a whole paragraph in your head before opening your mouth is too much; you'll lose the end of it before you reach it, which is working memory doing exactly what working memory does.

The natural unit is the phrase or the clause — roughly the amount you'd say in one breath. Think the shape of the next chunk, say it, think the next. This matches how speech actually wants to flow, and it keeps the cognitive load low enough that you can attend to meaning instead of mechanics. If you find yourself stalling, you're probably trying to hold too much. Shorten the chunk.

Say the punctuation only until you don't have to

Early dictation tools made you narrate every comma and period out loud, which is exhausting and one reason the whole category got a reputation for being fiddly. Modern on-device transcription infers most punctuation from the rhythm and pitch of your speech — a falling tone and a pause read as a period, a held tone reads as a comma. For most prose you can simply talk naturally and let the sentence boundaries land where your voice puts them.

Where it helps to be explicit is the structural stuff: a new paragraph, a list, a colon before a quote. A short pause before a new thought also gives the system a cleaner seam to work with. Beyond that, resist the urge to micromanage the marks. You'll spend more attention announcing punctuation than you'd spend fixing the occasional comma later.

The deeper reason to let punctuation infer itself is that narrating commas pulls you out of the flow of thinking. Every "comma," "period," "new paragraph" is a little context-switch away from what you actually mean and toward the mechanics of the page — which is exactly the friction dictation is supposed to remove. Trust your voice's natural cadence to carry most of it, and reserve your conscious attention for the words.

Write the way you'd explain it, not the way you'd perform it

People who struggle with dictation often do so because they're trying to speak in fully formed literary sentences — the kind they'd write, not the kind they'd say. That's a recipe for paralysis, because you're now doing the hard composition work and the speaking work at once, with none of the freedom that made speech easy in the first place.

A better frame: imagine you're explaining the thing to a smart colleague sitting across the desk. Talk to them. The register that comes out is conversational but coherent — looser than your final draft, but full of the actual substance. That's exactly the raw material you want. You can tighten the register in the edit; you cannot edit a paragraph that never got spoken because you were waiting for the perfect sentence to arrive.

Let the cleanup close the gap

Even with good technique, spoken English carries traces of speech — a stray "you know," a doubled-back start, a sentence that trails off. The final move in writing with your voice is to not do that cleanup by hand if you don't have to. When filler removal, punctuation, and light tidying happen automatically, the gap between "what I said" and "what I'd write" mostly closes before you ever read it. What you're left editing is substance and style, not litter.

This is the difference between dictation that feels like a chore and dictation that feels like a superpower. In the first, you trade typing labor for cleanup labor and barely come out ahead. In the second, the messy part is handled and you spend your attention only on the parts that need a human — the argument, the tone, the cut.

Give it a week

Like touch-typing, this clicks through use, not study. Pick something low-stakes — a few emails, a journal entry, notes after a meeting — and dictate it for a week, holding to the one rule that matters: speak when you speak, edit when you edit. By day three the parking-brake feeling lifts. By the end of the week you'll catch yourself reaching for the mic before the keyboard, not because someone told you to, but because it's simply less work.

Quill is designed to make that week easy. It transcribes and cleans your speech entirely on your iPhone or Mac, so the filler and punctuation are handled the moment you stop talking — and because it all runs on-device, nothing you dictate is sent anywhere. You speak in phrases, polished text lands where your cursor is, and the editing you're left with is the good kind. If you want to learn how to write with your voice for real, you can start at quill.lumenlabs.works.