How AI voice typing works from your voice to clean text
Plain dictation writes down every sound you make. AI voice typing goes further: it understands what you meant, removes what you did not mean to say, and formats the result for the app you are writing in. Here is what VoiceTyping does with every recording, in five steps, in about a second.
Step 1. Your voice
You hold your shortcut or the mic button and talk, the way you would explain something to a colleague. You do not need to say "comma" or "period", and you do not need to speak slowly.
| What you said |
|---|
| "um so hi Dana, uh, quick one, can we move the call to Thursday at 3, no, at 4, and uh can you send the deck before, like, Wednesday thanks" |
Step 2. Transcript
Speech recognition turns the audio into words. This is where built-in dictation usually stops. Your personal dictionary is applied here, so names, brands and jargon come out spelled the way you spell them.
| Raw transcript |
|---|
| um so hi Dana uh quick one can we move the call to Thursday at 3 no at 4 and uh can you send the deck before like Wednesday thanks |
Step 3. Fillers removed
Language models remove filler words (um, uh, like, you know), repetitions and false starts. Self-corrections are resolved, so "at 3, no, at 4" becomes "at 4".
| After cleanup |
|---|
| hi Dana quick one can we move the call to Thursday at 4 and can you send the deck before Wednesday thanks |
Step 4. Punctuation and paragraphs
Sentences get punctuation and capital letters. Long dictation is split into paragraphs, and anything you list out loud becomes a list.
| With punctuation |
|---|
| Hi Dana, quick one. Can we move the call to Thursday at 4? And can you send the deck before Wednesday? Thanks. |
Step 5. Style for the app
The text is shaped for where it is going. In Gmail the Email style adds structure and a sign off, in Slack the Chat style keeps it short.
| Style | Final text |
|---|---|
| Hi Dana, Could we move our call to Thursday at 4? It would also help to get the deck before Wednesday. Thanks, Alex | |
| Chat | Hi Dana, can we move the call to Thu at 4? And could you send the deck before Wednesday? Thanks |
Edit by voice after you dictate
Say a command right after a dictation and VoiceTyping changes the last result.
- "New paragraph" starts a new paragraph.
- "Make it a list" turns what you said into bullet points.
- "Make it shorter" keeps the meaning in fewer words.
- "More formal" adjusts the tone for a client or a manager.
- "Delete last sentence" removes the last sentence.
- "Translate to Spanish" (or another language) translates the result. Translate while dictating is part of Pro and Team.
Your words, spelled your way
The personal dictionary holds names, brands, product terms and abbreviations, so "Siobhan", "Kubernetes" or your company's product names come out right every time. Personal holds up to 200 words, Pro up to 2,000, and Team adds a shared dictionary for everyone.
Snippets are text you say instead of type. Say "my calendar link" and your real link is inserted. Personal includes 10 snippets, Pro and Team have unlimited snippets.
Languages and automatic detection
VoiceTyping works in English plus Spanish, French, German, Portuguese, Italian, Polish, Dutch and more. Leave the language on Auto and it detects the language of each recording, so you can write one message in English and the next one in Spanish without touching a setting. Cleanup and styles work in every supported language.
What happens to your audio
- Audio is sent over an encrypted connection and processed to produce text.
- It is deleted right after transcription and is never used to train models.
- Your text history is encrypted at rest. You can switch history off completely.
- You can delete all your data in one click, and your data is not sold.
Full details are on is voice typing private.
How minutes and limits work
An AI minute is one minute of recorded speech. A minute of normal speaking is roughly 130 to 150 words.
| Plan | AI minutes per month | About | Daily cap |
|---|---|---|---|
| Personal | 400 | 55,000 words | 40 min |
| Pro | 900 | 125,000 words | 90 min |
| Team | 1,500 per seat | 1,500 minutes of speech per person | 120 min per seat |
Your remaining minutes and today's usage are shown in the app. Minutes renew with each billing period. If you reach a limit, you can move to a higher plan right away. Compare plans on voice typing app pricing.
Where the cleaned text goes
VoiceTyping runs in the browser and as an installed web app on Mac, Windows, Linux, iPhone and Android. The speech to text Chrome extension for Chrome and Edge inserts the text into any text field on the web. On a phone you dictate in the installed web app and copy or share the result into any app with one tap. See every option on voice typing software features.
FAQ
Do I have to say "comma" and "period"?
No. Punctuation, capital letters and paragraphs are added for you. You can still say "new paragraph" when you want a break in a specific place.
How fast is it?
A short dictation comes back in about a second. Longer recordings take a little more, and you see progress while you wait.
Will it change what I meant?
No. Cleanup removes fillers and repeats and fixes formatting, but it keeps your words and your meaning. Commands like "make it shorter" change wording only when you ask.
Does it work with accents?
Yes. Speech recognition handles a wide range of accents, and your dictionary helps with names and terms that are easy to mishear.
What if I run out of minutes?
You see your usage in the app. You can move to a higher plan right away, or your minutes renew at the start of your next billing period.
See all five steps on your own voice
Record a messy sentence and watch it go from raw transcript to clean text. Try it without an account, no card needed to try it.