Live on-device transcription macOS 26+ · Apple Silicon only

HOLD.
SPEAK.
DONE.

Your voice, unchanged.

Vaaku is an AI dictation app for Mac. It listens while you talk, cleans up what you meant, and types it — without your voice ever leaving this Mac.

macOS 26+, Apple Silicon only  ·  Free beta  ·  No account, ever

RECORDING — HOLD FN speech.local

hey can you pull up the deck from yesterday— wait, let me start that ove

0 network calls 0 accounts 0 third-party dependencies 100% on-device
Privacy

Your voice never leaves the room.

No backend. No accounts. No cloud APIs. Vaaku makes zero network calls, period. Capture, transcription, and cleanup all run locally, entirely on-device, with nothing sent to a server.

  • No URLSession, no analytics SDKs, no telemetry, no API keys — anywhere in the codebase.
  • Cleanup runs against an on-device language model. Not available? Vaaku falls back to a local formatter, never a cloud LLM.
  • Zero third-party dependencies. No SwiftPM packages, no vendored SDKs, nothing phoning home.
100%
LOCAL

Verify it yourself: disconnect your Mac from the internet entirely. Vaaku won't notice — it never needed the connection.

Speed

Speak at full speed. It keeps up.

Hold fn and your words appear as you say them — streamed live while you're still talking, transcribed entirely on-device. Let go, and an on-device model tidies it up and drops it exactly where your cursor was.

  • Live partial transcript overlay while you speak, not after you stop.
  • Native on-device speech engine — no bundled model download, no round trip to a server.
  • Built in Swift for macOS 26. No Electron. No cross-platform compromises.
0ms
Round trips to a server
— there's no server
Your voice

It writes the way you actually talk.

Cleanup trims filler words and fixes punctuation and grammar — but makes only the edits it needs to. Your wording stays yours. Need more structure? Just say so, mid-sentence.

"…as bullet points"Formats the content as a list
"…new paragraph…"Starts a new paragraph
"…new line…"Starts a new line
Concise Mode
Opt-in. Actively cuts filler, redundancy, and hedging. Off by default.
Always ready

Lives in the menu bar. Speaks only when spoken to.

Hold fn to talk, release when you're done. Double-tap fn instead for hands-free recording that keeps going until you press it again. The rest of the time, Vaaku stays out of the way.

HOLD

Press and hold fn, speak, release. One utterance, done.

DOUBLE-TAP

Tap fn twice for hands-free recording. Tap again to stop.

How it works

Four beats. No setup.

1

Hold

Press and hold the fn key. A quiet overlay appears near your cursor.

2

Speak

Say what's on your mind. A live transcript follows every word, on-device.

3

Release

Let go of fn. An on-device model cleans up the transcript, locally.

4

Done

Your words land exactly where your cursor was — already typed.

Why Vaaku

Cloud dictation tools ask you to trust them. Vaaku doesn't have to be.

Typical cloud dictation
  • ×Your audio is uploaded to a server
  • ×Requires an account and a subscription
  • ×Stops working the moment you're offline
  • ×A model you don't control, updated on someone else's schedule
VS
Vaaku
  • Never leaves your Mac
  • No account. Ever.
  • Works fully offline, always
  • Local, on-device models
100%
REAL

Your Mac already understands you. Let it type, too.

macOS 26+  ·  Apple Silicon only, Intel not supported  ·  Free beta