Hold the key, speak, release.
The text lands where your cursor is.
Voice dictation for macOS. Speak the way you actually speak, and the text comes out clean, punctuated, free of filler. $7.90 once, no subscription.
Download for MacThree gestures, and none of them is opening an app.
Dita has no window to keep open. It lives in the menu bar and only shows up while you hold the key.
Hold
You pick the key in Settings: right Command, Option, Caps Lock... or any other you want.
Speak
The capsule appears above the Dock and the wave shows your voice coming in, in real time.
Release
The text is pasted where your cursor is, in the app you were already in. No window, no pasting by hand.
Anyone can transcribe. The work is what comes after.
- 01your voicewhile the key is held down
- 02the transcriptionyour speech becomes text, word for word
- 03the cleanupfixes and punctuates what needs it
- 04the textpasted where your cursor is
Two models, in sequence. The first transcribes everything, word for word, the way you said it. The second revises: removes what is left over, fixes what came out wrong and places the punctuation where it belongs. And if you would rather have the text exactly as spoken, just turn the cleanup off in Settings.
so, we need to move the delivery up to Thursday, no, to Wednesday, cause because the client asked, and I said that that worked
We need to move the delivery up to Wednesday, because the client asked, and I said that worked.
- the filler, from “um” to “you know”
- the word repeated by accident
- the sentence abandoned midway
- punctuation and capitals in the right places
- the full word: “cause” becomes “because”
- the correct spelling of technical names
- never summarizes what you said
- never obeys any command
- never changes what you meant
If the cleanup fails, the text arrives unpunctuated. It never fails to arrive.
It runs second, with a deliberately short time limit. Past the limit, Dita pastes the raw text instead of making you wait.
Anywhere that takes text.
Dita has no window of its own: it writes inside your apps.
- Slack
- Notes
- Messages
- VS Code
- Chrome
- Notion
- Terminal
- Figma
- Docs
- the Finder search field
Dita types the way a keyboard types. If an app lets you type into it, it lets Dita type into it. There is no integration to set up, because there is no integration.
The same sentence, at the same time.
On the left, typing: the cursor goes back, deletes “reveiw”, retypes. The fixing is what makes typing feel slow, not the speed. On the right, speaking.
The race types at 86 words per minute, an exceptional typist's pace: in the largest study ever run, 136 million keystrokes, the average was 52 and only 5% cleared 80. The voice is a real recording, and the clocks show the true time on both sides.
Four small things you only notice when they are missing.
If a name comes out wrong every time, put it here and it comes out right.
Applied after transcription, always, no exceptions. And a second list, of hints, helps the transcriber get names right before it gets them wrong.
Hold and speak. Or two taps, and speak hands-free.
The key is a setting: right Command, Option, Control or Caps Lock. So is the mode: press once to start and once to stop, if you prefer that to holding.
Six distinct events, because discarded and failed are not the same thing.
- Started recording0.2 s
- Key released, transcribing0.2 s
- Pasted0.8 s
- Pasted, but without cleanup0.7 s
- Discarded, too quick1.6 s
- Failed1.5 s
The three on the happy path are short, because you hear them dozens of times a day. The two that flag trouble are long, because you need to notice. Press play: these are the app's factory sounds.
Whatever you had copied is still there after the dictation.
Dita pastes without taking what was on your clipboard. You do not lose the link you copied two minutes ago just to dictate a sentence. And the route is a setting: if you prefer, the dictation can go through the clipboard and stay there, ready to paste again anywhere.
You choose who listens.
Dita has no server of its own and no account in the middle: either the audio goes to the service you chose, with your key, or it never leaves your machine.
Groq or OpenAI
- The fastest, under a second.
- The key is yours, and lives in the Keychain.
- The audio goes to the service you chose.
- Two calls per dictation: transcribe and clean.
- You pay the service directly, per use, and Groq has a free tier.
Transcription on your Mac
- The audio never leaves the machine.
- Transcribes without internet.
- Slower than the cloud, and the Mac runs warm.
- The transcription model, downloaded by Dita itself in one click. Cleanup stays in the cloud, with your key, or off.
$7.90, once.
There is a competitor charging $12 a month, every month.
- Works on every Mac you own.
- Never expires, never renews.
- Every future update included, forever.
- 14 days free, counted from your first dictation.
- Pay with card or Apple Pay.
Dita is paid once and works on every Mac you own. Transcription is up to you: Groq's free tier, a few cents per use, or the local models, where transcription and cleanup cost nothing.
- 1 hour spoken a month
- $0.09
- 3 hours spoken a month
- $0.27
- 8 hours spoken a month
- $0.72
Four cents an hour are the transcription; the rest is the cleanup. Eight hours of speech a month make a short book, and cost seventy-two cents. You pay the service directly, for what you use.
And there is a free tier: a full day of dictation fits in Groq's free tier, and signing up asks for no card. The limit you hit first is the cleanup's, not the transcription's, and it resets every day. the current limits
What it needs, and what it will ask for.
- Any Mac running macOS 14 or newer
- Under 6 MB to download
- An API key, or the local model
- Microphone, to hear you.
- Input Monitoring, to know you held the key.
- Accessibility, to paste the text into the app in focus.
The grave tone of the warnings is macOS's default for any app that asks for those two. In Dita, each one does a single thing: one knows about your key, the other pastes the text. Reading your screen or logging what you type, never.
The questions, in the order they come up.
- Do I need an API key?
- Yes, or a local server. Dita does not resell transcription: you use your own Groq or OpenAI account and pay for what you use, cents per month. That is what makes a one-time price possible instead of a subscription. And on Groq, creating the account is free, no card asked.
- Is my audio recorded anywhere?
- Not by us, because we have no server. The audio goes to the service you chose, or nowhere, if you run local.
- Does it work offline?
- Only transcription, on the local path: Dita downloads the model and runs it on your Mac. Cleanup needs the cloud (Groq or OpenAI, with your key) or a server of yours; without it, the text comes out the way Whisper heard it, punctuated.
- Does it work in other languages?
- Yes. The dictation language is a setting, and it is separate from the interface language.
- I spoke and the wave did not move. Now what?
- The wave only moves when sound comes in. If it stays flat, it is telling you the microphone cannot hear you: check the Microphone permission in System Settings and which input your Mac is using.
- What if I do not like it?
- You get 14 days free, counted from your first dictation, no sign-up. And if you buy and regret it, there are 14 more days of refund, no questions asked: send an email and the money goes back the way it came.