Punctuation and formatting
How VoiceSnap Pro punctuates what you say and takes out filler words, why there are no spoken punctuation commands, and how to get structure into a dictation.
On this page
Automatic punctuation
You speak in plain sentences and VoiceSnap Pro punctuates them. The transcription model reads phrasing, grammar and the pauses between clauses to place full stops, commas and question marks, and it capitalises the first word of each sentence and the names it recognises. You never say the word “comma” out loud.
It also tends to write things the way people usually write them rather than the way they say them: numbers, times, percentages and amounts generally come out as figures, and an email address or web address spoken aloud usually comes out without spaces. “Usually” is the honest word — these are the model's habits rather than settings, so check anything where the exact form matters.
No spoken punctuation, and why
VoiceSnap Pro has no spoken-punctuation mode. Saying “comma” or “new paragraph” is not a command it acts on: the words go to the transcription model like the rest of what you said, and what comes back depends on the model rather than on a rule.
In practice that is rarely a loss. For prose — email, messages, documents — automatic punctuation does the work, and constantly saying “comma” breaks the rhythm that makes dictation faster than typing in the first place. When you need one exact mark, type it.
Paragraphs and line breaks
VoiceSnap Pro does not insert paragraph breaks, and there is no spoken command for one: a dictation comes back as running text. If you need structure — paragraphs, a list, an address, a code comment — dictate it in pieces and press Return between them yourself.
In a chat box where Return sends the message, VoiceSnap Pro never presses it for you. The text is pasted, and sending stays your decision.
Filler words and false starts
Clean up dictation, on by default in Settings → AI, takes out filler words such as “um”, “uh” and “you know”, stutters and repeated words, and a slip you corrected out loud — say “Tuesday, no, Wednesday” and only “Wednesday” is kept. It only ever removes words and fixes punctuation and capitals: it never adds, rewords, respells or translates, and if the result contains a word you did not say, or the clean-up fails, the transcript is used as it was. It adds about a second.
It runs on dictations and voice notes of up to about five minutes. Longer recordings, re-transcribed notes, dictations into a locked note and dictations of fewer than three words keep every word, so for those, pausing silently while you think still gives the cleanest result. Turn it off to keep every word everywhere.
If a dictation needs more than that, do it afterwards and on purpose: Fix grammar & spelling or Make shorter, run as an AI action, reworks it and saves the result as a new note.
Names and jargon
Punctuation is rarely what needs fixing; names are. Add the people, products and terms you use to your personal vocabulary, and the model is given your spellings with every dictation. Custom vocabulary covers it.
Fixing it up afterwards
When a dictation needs reshaping rather than correcting, run an AI action on it: Fix grammar & spelling, Make shorter or Turn into bullet points. On a dictation the result is saved as a new note, so the original is never changed. In a note, the action works in place, and Undo takes it straight back. On the free allowance each action uses one of your AI actions; Pro has no allowance to use up.
AI actions are never applied to a dictation on their own. Apart from Clean up dictation taking out fillers and corrected slips, what you said is what gets pasted; anything more is a choice you make afterwards.