Speech to text for macOS

Dictation that never leaves your Mac.

Flit turns speech into finished text while you are still speaking, 4× faster than typing. Fillers gone, punctuation right, formatted for whatever app you are in.

7 days free · Then $29 once · No account · macOS 14 or later · Apple silicon

New message 9:41

Does Friday still work for the kickoff?

Can we meet at 10:30 AM on Friday? 

0.7s after keyup

um so can we uh meet at 1030 AM on, on friday

fn

Demo · 2:36

Watch Flit do the typing.

Two and a half minutes, every feature as a real dictation you can hear. Hold a key, talk, let go, and watch the finished text land in Messages, Slack, Mail, Notes and your editor.

0:00 / 2:36

faster than typing

0.7s

from keyup to finished text

99

languages, 25 live as you speak

0

bytes of audio leave your Mac

Types into every app on your Mac, including

SlackMailMessagesWhatsAppTelegramDiscordNotesNotionChatGPTClaudeObsidianLinear
ChromeSafariArcFigmaVS CodeCursorZedXcodeJetBrainsWarpTerminalTextEdit

If you can put a cursor in it, Flit can write in it. Chrome and Electron apps included, which is where most dictation tools give up.

01Before and after

This is what you said. This is what Flit wrote.

Every other dictation tool hands you a transcript and leaves the cleanup to you. Flit does the cleanup, on your Mac, before the text ever lands.

You said

“um so can we uh meet at 1030 AM on, on friday”

Flit wrote

Can we meet at 10:30 AM on Friday?

The engine hears "1030 AM". The clean 10:30 is a fixed rule, not model luck.

You said

“the budget is uh three hundred dollars scratch that four hundred dollars”

Flit wrote

The budget is $400.

"Scratch that" cancels the statement before it. Only the correction survives.

You said

“send it to bob at acme dot com and uh cc the the design team”

Flit wrote

Send it to bob@acme.com and CC the design team.

Emails, URLs, and doubled words are written the way you would have typed them.

02Instant

4× faster than typing, measured on real dictations.

You speak at over 120 words a minute and type at about 31. Flit closes that gap without asking you to change anything: it transcribes while you talk, so when you release the key there is almost nothing left to do.

124 WPM averaged across 500+ real dictations. About 0.7s from keyup to finished, formatted text.

How the 4× is measured

The 124 is a floor, not a benchmark: real delivered words, pauses and thinking included, measured on a speaker whose first language is not English. Published speaking-rate benchmarks run higher, so 4× is the conservative end of the claim.

Transcription is done when you release the key; the 0.7s is the formatting pass. Prefer raw speed? Clean mode lands in about 0.15 seconds. And Flit keeps count on your Mac only: the dashboard shows the words you spoke and the hours you got back.

The same 31-word reply

Typing 31 WPM

still typing…

Speaking with Flit 124 WPM

done · 15 seconds

Fifteen seconds spoken. A minute typed. That difference, a few hundred times a week, is where the hours come from.

Why it matters

The keyboard is the bottleneck.

You think at speaking speed. Every study that measures it finds the same gap: the words are ready long before the fingers are.

3×

faster: speech beat keyboard entry in Stanford’s controlled text-entry study.

Stanford · Ruan et al., 2016

28%

of the knowledge workweek goes to reading and answering email.

McKinsey Global Institute

3.1hrs

a day: what the average worker spends in work email alone.

Adobe Email Usage Study, 2019

Sources: Ruan et al., Stanford HCI · McKinsey, The Social Economy · Adobe, 2019

03The arithmetic

How much of your year do you want back?

Speaking at 124 words a minute against composing at a keyboard at about 31, three of every four writing minutes come back to you. Slide to the day that looks like yours.

2,000

250 · the occasional reply 6,000 · writing is the job

48 min

back every day

4 hours

back every week

5 weeks

of workdays back, every year

124 WPM measured across 500+ real dictations · ~31 WPM composition typing · 5-day weeks, 250 working days, 8-hour days

5 weeks a year.

That is what your slider says. Workdays returned, not minutes shaved.

04Writes like you

It writes the way you would have typed it.

Flit removes fillers and false starts, fixes punctuation and capitalization, and writes numbers, dates, times, prices, emails, and URLs properly. Say “scratch that” and it keeps only the correction.

You said: “um so the launch is moving to thursday can you update the doc”

Messages · casual

the launch is moving to Thursday, can you update the doc?

Mail · professional

The launch is moving to Thursday. Could you update the doc?

Notes · neutral

The launch is moving to Thursday. Can you update the doc?

Messages · casual Mail · professional Terminal · verbatim

Matches the app you are in

Casual in Messages, professional in Mail, your exact words in Terminal. Set your own rule per app.

kuber netties Kubernetes

Added to your dictionary

Learns your spellings

Correct a word twice and Flit offers to remember it. Nothing is learned without your approval.

Can you send the deck to Priya?

Sending it to Priya now.

Knows the conversation

Optionally reads the field you are typing into, so replies pick up names and match the thread. Read locally, kept in memory for one dictation, never stored.

The launch is delayed until further notice. Launch delayed, more soon.

“make this shorter”

Edit by voice

Select text anywhere, hold the key, and say “make this shorter” or “fix the tone”. Command Z undoes it.

05Exactness

The details are not left to a model.

A language model is good at judgment and unreliable at precision. So Flit only gives it the judgment. Anything with one correct answer is handled by fixed rules that produce it every time.

Say “ten thirty AM” and you get 10:30 AM. Not usually. Every time.

Fixed rules · always exact

“ten thirty AM” “dash dash force” “three hundred dollars” 10:30 AM --force $300
  • Clock times, written properly whatever the engine heard
  • Spoken punctuation and symbols, more than 60 commands
  • Words from your dictionary, spelled your way
  • Snippets, inserted exactly as you wrote them
  • File and symbol names from your open project

The model · for judgment

  • Tone and register for the app you are in
  • Sentence flow and paragraph breaks
  • Which fillers were filler and which were meaning

And if the model misbehaves, Flit throws its answer away and keeps the deterministic result. Dictation can never break.

06Activation

Three ways to start. All of them one gesture.

fn

Hold

Hold Fn or your own shortcut, speak, let go. Best for a sentence or two.

⌥ ␣

Tap

A quick tap locks recording on so you can talk as long as you want. Tap again to finish.

· · ·

Hands-free

Arm it once and Flit starts the moment you speak and stops when you pause. Nothing is recorded until it hears you.

Esc cancels any dictation. The audio is deleted, nothing is inserted.

Lives in the notch

The live transcript hugs the camera housing, Dynamic-Island style. On external displays Flit draws its own.

9:41
meet at 10:30…

Or floats where you want it

A glass pill in any corner, top or bottom. It grows as the live words arrive and gets out of the way when you stop.

meet at 10:30…

07Snippets

Say it once, insert the whole thing.

Say “insert my scheduling link” and Flit types the link. Snippets hold anything you retype often: sign-offs, addresses, canned replies. Triggers forgive the small differences in how you actually speak, and what lands is exactly what you saved, never reworded.

“insert my scheduling link”

https://cal.com/your-link

Spoken commands

Punctuation, spoken.

Say “comma”, “open paren”, “dollar sign”, “new paragraph”. More than 60 commands, handled without a model touching them, so what you get is exactly what you asked for. 115 emoji by name. Always on in terminals and editors, optional everywhere else.

“open paren”( “dash dash force”--force “camel case …”likeThisInstead “rocket emoji”🚀

08Developer mode

It knows the names in your codebase.

Dictating into an editor or terminal, Flit reads the open project’s file and symbol names locally, so spoken identifiers come out spelled right. It fixes mishearings too, because it can see the real name.

Cursor 9:41

Explorer

HUDController.swift

UserAuthService.swift

VoiceSnippetStore.swift

SpeechEngine.swift

FormatterPipeline.swift

“open user auth service dot swift”

open UserAuthService.swift

“voice snicket store” (misheard)

VoiceSnippetStore.swift

“git log dash dash oneline”

% git log --oneline

Flit reads file names and paths, never file contents. The index stays in memory and never touches disk or network. In terminals and editors your words are never reworded, so a shell command survives intact.

Terminal Xcode VS Code Cursor Sublime iTerm Warp Zed JetBrains and more

09Languages

Speak 99 languages. Switch without touching a setting.

Flit identifies what you are speaking in the first seconds and picks the right engine for it, so a bilingual conversation does not mean a trip to Settings. 25 languages transcribe as you speak; the rest are handled by a slower engine that still runs entirely on your Mac.

English German French Spanish Italian Portuguese Dutch Polish Swedish Danish Norwegian Finnish Czech Ukrainian +85 more

Translation

Think in one language. Write in another.

Dictate in the language you think in and have Flit write in another. Speak German, write English. Off by default, and your words are never translated unless you ask.

können wir uns äh am freitag um halb elf treffen

Can we meet on Friday at 10:30?

10Private, provably

Do not take our word for it.

Flit has no account, no telemetry, and no server. Your voice and your text never leave this Mac, and you do not have to trust us on that.

Open Little Snitch or any network monitor, then dictate all day. You will see exactly two kinds of connection: model downloads you started yourself, and update checks you agreed to. Nothing else, ever.

No account. Flit never asks who you are.

No telemetry. We do not know how you use it, and we like it that way.

Works fully offline. Turn off wifi and everything still works.

Your history stays on your Mac, and you choose how long it lives.

Wispr Flow screenshots your screen and sends it to their cloud for context. Flit reads the text locally instead, keeps it in memory for one dictation, and never stores it.

Read the full privacy statement

11Who it’s for

Made for people who write all day.

The engine is the same for everyone. What changes is what it knows: your apps, your names, your tone. Find the page written for the way you work.

Teams & agencies

The easiest security review your team will ever pass.

Flit has no cloud, so there is nothing to review: no audio leaves any Mac, no vendor holds a copy, no DPA or subprocessor list to negotiate. And a license is $49 once per person, not a per-seat subscription someone has to defend at renewal time.

Nothing uploaded, so client and patient work stays confidential

No accounts to provision, nothing to deprovision

One-time license per person, on up to 2 of their Macs

Buying for a team? Email us

12Compared

Compared, by name.

Flit Wispr Flow Superwhisper SpeakMac Spokenly
Runs on your Mac AlwaysCloudLocalLocalLocal or cloud
Price $49 onceMonthly, forever$250 once$29 onceFree / $9.99 mo
Built for macOS only YesNoYesYesNo, 4 platforms
Text appears as you speak YesYesNoNot statedNot stated
Languages 99100+Many25 + packsMany
Account required NoYesNoNoNo
Exact times and commands Fixed rulesModel decidesModel decidesNot statedNot stated
Knows your codebase Indexes itNoNoNot statedNot stated
Style rule per app YesYesNot statedNot statedNot stated

Compared against each product's own published information. “Not stated” means the product does not publish that detail.

Pricing

One price. Yours forever.

You will buy Flit once and it is yours permanently, on up to 2 Macs, with 12 months of updates. No subscription, no account, no upsell.

Launch pricing

First 100 licenses

Once they are gone, Personal steps to $39, then $49. Lifetime goes to $99.

Personal

Most people

$29

was $49

Yours forever, on 2 Macs. Updates for a year.

Lifetime

$79

was $99

Yours forever, on 2 Macs. Every update, forever.

Not on sale yet. These are the prices Flit will launch at. macOS 14 or later · Apple silicon. 7 days free first, no card, and a 30-day refund.

Both include

  • Every feature. There is no Pro tier and nothing is held back.
  • All 99 languages and every local model, downloadable free.
  • Developer mode, snippets, the dictionary that learns, a style per app.
  • No account. Flit never asks who you are.

Questions? The pricing page answers the usual ones. If a future major version is ever a paid upgrade, you will hear it from us before you pay for anything.

Questions

Fair questions.

Does it work offline?

Yes, completely. The only time Flit uses the network is when you download a model or check for an update.

What are the requirements?

macOS 14 or later on Apple silicon. Some models need macOS 26.

Is my voice used to train anything?

No. Your audio never leaves your Mac, so there is nothing to train on.

Do I need an account?

No. Flit never asks who you are.

How big is the download?

The app is small. Speech models are downloaded separately when you choose one, from 75 MB to 1.6 GB depending on the model.

Does it work in every app?

Yes, including Chrome and Electron apps like Slack and VS Code, which most dictation tools struggle with.

Why is it better at times and punctuation than other tools?

Because it does not ask a model. Anything with one correct answer, like a clock time or a spoken comma, is handled by fixed rules that cannot drift.

What if I do not like it?

Use it free for 7 days first. After that, 30-day refund, no questions.

Will there be a Windows version?

No plans. Flit is built for macOS specifically, and that is why it is this fast.

Your voice was always faster.

Not out yet. $29 USD once when it is, with no account, no subscription and nothing uploaded.

No spam, no sharing. One email when Flit ships, and you can leave in one click.

macOS 14 or later · Apple silicon