spellsafe
Many writers who try dictation find their first-draft word counts jump. Talking is two to four times faster than typing for most people, and the inner editor that makes you rewrite every sentence tends to stay quieter when you’re speaking.
The catch is that dictation only helps if the tool stays out of your way. If you keep stopping to fix words, add punctuation or fight with the app, you lose the flow that made it worth trying.
Here’s a workflow that works, and how spellsafe fits into it.
Draft by voice, edit by keyboard (mostly)
The most common mistake is trying to dictate finished prose. Treat dictation as a way to get the draft out of your head, and keep editing as a separate pass.
- Outline first. Know the beat of the scene or the point of the section before you start talking.
- Talk through it without stopping. Mistakes are fine. Say “scratch that” and carry on if a sentence really went wrong.
- Edit later, by keyboard or with voice commands.
Setting up spellsafe for long drafting sessions
spellsafe types into any app: Scrivener, Ulysses, iA Writer, Word, Pages, Google Docs or your CMS.
- Live mode for flow. Tap Right Option and talk. Words are typed about half a second to a second behind you, and the final pass when you stop fixes anything that changed. Tap again to stop. Some writers like seeing the words land; if it’s distracting, look away from the screen.
- Hold-to-talk for careful passages. Hold Right Option, say a paragraph, and let go. The whole clip is transcribed in one pass, which is slightly more accurate.
- “New paragraph” and “new line” to break up text without touching the keyboard.
- It doesn’t run away on you. A dictation stops itself after 30 seconds without speech (you can change that), so if you get up to make tea, it won’t type whatever the microphone picks up. “Stop dictation” ends it by voice.
- Long sessions don’t slow down. Audio is streamed to disk as you talk, and memory use stays flat however long you go.
Character names and invented words
Fantasy and science fiction writers know the pain: the dictation tool will never spell “Aelindra” on its own. In spellsafe’s Vocabulary you add the name and what it sounds like (“ay lin dra”), and it’s typed your way from then on, in live mode and on release. It works for real-world jargon too: place names, technical terms and brand names.
Editing by voice
When the draft is down, you can stay on voice for a lot of the clean-up:
- “Replace Tuesday with Wednesday” changes the last occurrence of a word you dictated.
- “Select that”, then keep talking, replaces the last sentence with what you say next.
- Command mode: select a passage, hold Right Command, and say “tighten this”, “make this more formal” or “fix the typos”. A language model on your Mac rewrites it in place, and ⌘Z undoes it.
- “Rewrite that…” applies an instruction to what you just dictated: “rewrite that in the past tense”.
AI cleanup is optional. If you want your words exactly as you said them, leave “Clean up with AI on release” off and only rewrite when you ask.
Your manuscript stays yours
A lot of writers are uneasy about where their unpublished work goes, and whether it ends up training someone else’s AI. spellsafe recognises speech on your Mac and runs its AI cleanup locally. Your audio and text are never uploaded, there’s no account, and after a one-time model download it works offline, so you can draft on a plane or in a cabin with no signal.
Track your progress
The Stats page shows words dictated, your speaking speed, words per day for the last two weeks, your streak, and time saved compared with typing. It’s a useful nudge on days when the words come slowly.
Limits to know about
- English only, for now.
- Apple silicon Macs with macOS 14 or later.
- It doesn’t transcribe recordings you made elsewhere, like voice memos from a walk. It types what you say live.
Try it
spellsafe has a 7-day free trial with every feature, then costs $9 a year. Download spellsafe.