On-device dictation
Just talk. It’s already typing.
Hold a key, speak, and clean text appears wherever your cursor is — transcribed and polished on your own Mac. Nothing is uploaded, nothing is stored on anyone else’s server.
Free · Apple Silicon · macOS 14+ · signed & notarized by Apple
Most dictation apps send your voice to a server.
The popular voice-to-text tools stream your microphone to the cloud, transcribe it there, and send text back. It works — until the wifi drops, or you stop to think about whose servers just heard your contract, your patient note, your unreleased idea.
Just Talk was built the other way around. The model runs on your Mac. Your speech is transcribed and cleaned up locally and then it is gone — there is no account to create and no audio to delete, because none of it ever left the machine in front of you.
Your voice is yours. Just Talk keeps it that way.
How it works
Hold. Talk. Done.
- 1
Hold your key, talk
Press and hold your activation key anywhere — email, Slack, a doc, a code editor. A live pill shows your words as you speak.
- 2
Transcribed on-device
A Whisper-class model runs on your Mac's Neural Engine. Your voice never touches a server — no upload, no account, no wait for the network.
- 3
Cleaned up automatically
An on-device model strips the ums, false starts and filler, fixes punctuation and casing — and keeps every word you meant.
- 4
Pasted where you were typing
The finished text drops straight into whatever app you were in. Let go of the key and keep working.
Why on-device
Running on your machine isn’t a limitation. It’s the point.
🔒
Private by construction
Audio and transcript are processed on your machine and never sent anywhere. There is no cloud to leak — the privacy is architectural, not a policy promise.
✈️
Works offline
On a plane, in a basement, on hotel wifi — dictation works with no connection, because there is nothing to connect to.
🌏
Speaks how you speak
English, Hinglish, and Gujarati — code-switch mid-sentence. The on-device multilingual model was picked for exactly this, not just clean American English.
∞
No per-word cost
Cloud dictation pays a vendor for every minute you speak. On-device costs nothing to run, so it never has to meter you.
Not just transcription
It writes down what you meant, not every “um.”
Raw speech-to-text gives you the false starts, the repeated words, the “you know”s. Just Talk cleans that up on-device — filler removed, punctuation and casing fixed, sentences tidied — while guarding your actual content so a clean-up pass never quietly drops a clause you said.
You say
“um so the the best data would be like, you know, the last quarter i think”
Just Talk types
The best data would be the last quarter, I think.
Asks for two permissions. Not one more.
To type for you, Just Talk needs your microphone and macOS Accessibility — the same access a dictation tool has to have, and no more. No screen recording, no input monitoring, no analytics account.
🎙️
Microphone
To hear you while you hold the key. Nothing is recorded to disk beyond the moment it takes to transcribe.
⌨️
Accessibility
To place the finished text into the app you were typing in, and to use your activation key from anywhere.
Where it runs
Mac today. Your phone and PC next.
The same on-device engine is coming to the devices you type on most — each one keeping the work on the device, not in the cloud.
macOS
Available now
Apple Silicon Mac, menu-bar app.
iPhone / iPad
In progress
System keyboard — dictate into any app.
Android
In progress
Input method — on-device, multilingual.
Windows
Planned
On-device on RTX/CUDA and CPU.
Requirements (Mac): Apple Silicon, 8 GB memory, macOS 14 or later.
Try Just Talk on your Mac.
Free to download. Open it, grant Microphone and Accessibility, and start talking. Your voice never leaves the machine.
Apple Silicon · macOS 14+ · signed & notarized by Apple