Guides
From rambling to finished: formatting dictated text
Nobody speaks in paragraphs. The gap between what you said and what you would have typed is real, and there are three honest ways to close it.
Last updated
What raw dictation looks like
It is worth seeing honestly, because people assume their own speech is tidier than it is. Here is twelve seconds of ordinary talking:
so um I think we should probably push the launch to to Monday because er legal haven't signed off yet and and I don't want to you know ship something that we then have to pull
Nothing wrong with it — that is how people talk, and it is perfectly clear out loud. It is simply not what you would have sent.
Method one: say the punctuation
The oldest method and still the most precise. You say "full stop", "comma", "question mark", "new paragraph", and they appear.
| Say | Get |
|---|---|
| full stop / period | . |
| comma | , |
| question mark | ? |
| new line | A line break |
| new paragraph | A blank line and a new paragraph |
| open quote / close quote | “ ” |
| colon, semicolon, dash | : ; — |
It feels absurd for about a day and then becomes automatic. Models will place punctuation for you, but they are guessing at your intent, and you are not. For anything where a comma changes the meaning — contracts, instructions, anything legal — say it yourself.
Method two: dictate loosely, edit by hand
For short messages this is genuinely the fastest route. Speak the whole thing, then spend ten seconds deleting the "um"s and adding the full stops.
It stops scaling at around a paragraph. Editing four hundred words of unpunctuated speech is slower than having typed them, which is where most people conclude dictation is not for long-form work. It is — see method three.
Method three: rewrite it automatically
A small language model takes the raw dictation and returns what you would have typed: fillers gone, punctuation in, spoken corrections applied, lists laid out as lists.
What you said
so um I think we should probably push the launch to to Monday because er legal haven't signed off yet and and I don't want to you know ship something that we then have to pull
What lands on the page
I think we should push the launch to Monday. Legal haven't signed off yet, and I don't want to ship something we then have to pull.
The same mechanism can target a shape rather than just tidying: the same twelve seconds of speech becomes three bullet points, or a formal email, or a set of notes, depending on what you asked for.
| Style | Use it for |
|---|---|
| Natural | Default. Your words, tidied. Nothing added |
| Concise | Chat and short replies. Cuts hard |
| Professional | External email. Raises the register without inventing content |
| Bullets | Notes, actions, anything someone will skim |
| Adds a greeting and a sign-off around what you said |
The two questions to ask about any rewriting feature
- Where does it run? Many apps recognise speech locally and then send the text to a cloud model to tidy it. That is a second transfer, of the actual content, and it is often on by default. Vaitly Voice's writing model runs on your Mac, which is also why it works on a plane.
- What happens when it is wrong? A model that rewrites can also invent — a figure that was not said, a sentence you did not mean. The important behaviour is what the app does when it is uncertain. Vaitly Voice validates the rewrite against what you actually said and pastes your raw words rather than vouch for something it cannot.
Which method for which task
| Task | Method |
|---|---|
| A chat message | Loose, then fix by hand. Or automatic Concise |
| An email to a colleague | Automatic tidy-up, then read it |
| An email to a client | Automatic, then read it properly. Or say the punctuation |
| Meeting notes | Automatic, targeting bullets |
| Anything contractual or legal | Say the punctuation. Do not delegate a comma |
| A first draft of something long | Loose and fast. Structure it afterwards by keyboard |
| A prompt to an AI | Loose. It does not care about your commas |
Questions
Do I have to say the punctuation?
No. Modern models place it reasonably well, and a rewriting step handles it too. Say it yourself when a comma changes the meaning — instructions, contracts, anything where precision is the point.
Does automatic rewriting change my meaning?
It can, which is why what matters is the behaviour when it is uncertain. VV compares the rewrite against what you actually said and falls back to your raw words rather than paste something it cannot vouch for. Read the result regardless.
Does the tidy-up send my text to a cloud service?
In VV, no — the writing model runs on your Mac, which is why it works with no connection. In many apps that stage is a cloud API call. It is worth checking specifically, because it is the step where a local-first setup most often leaks.
Can I turn the rewriting off?
Yes. Some people want their exact words and nothing else, particularly for notes to themselves. It is a setting rather than a fixed behaviour.
Can it produce bullet points from a spoken list?
Yes — say the items in a sentence and ask for bullets. It is one of the highest-value shapes, because a spoken list is naturally messy and a written one needs to be scannable.