macOS · On-device AI dictation

We've typed for 150 years.

Time to speak.

Speak, and get clean prose. Your voice, the transcript, the AI cleanup — all of it stays on your Mac. Not a single word goes to the cloud.

Apple SiliconFreeNo account

Dictation runs on macOS 14+. AI formatting needs a Mac with Apple Intelligence (macOS 26+).

Fn

Speak and watch it appear. Let go, and AI cleans it up and inserts it.

Dictation, until now

Your words were sent to a distant server.

Cloud dictation is fast and smart. But your voice and text always pass through someone else's machine.

Thinka is different

Speak. Done. Nothing leaves your Mac.

Apple Intelligence's Foundation Models and Apple Silicon-tuned WhisperKit. Recognition and AI cleanup run entirely on your Mac.

One key — Fn

As fast as you think.

INSTANT MODE
Hold Fn for quick text — straight into Slack or a search bar.
STREAMING MODE
Fn + Space for continuous speech. Live captions flow at your cursor; stop, and AI reshapes and replaces them.
EDIT MODE
Select text and press Fn. Say “make it polite,” “delete line 3,” “translate to English” — it rewrites the selection.

After you speak

It doesn't stop at transcription.

AI CLEANUP
Removes fillers, adds punctuation, breaks paragraphs — on device. Spoken words become readable prose.
CONTEXT
Matches the front app: formal in Mail, casual in Slack, symbols preserved in code.
CUSTOM RULES
Set standing rules in plain language — “no emoji,” “British English,” and so on.
DICTIONARY
Absorbs name spellings and expands your shorthand automatically.

How it's different

Private and smart. No compromise.

Cloud tools are smart but send your voice out. The OS default is safe but plain. Thinka fills the gap.

 ThinkaCloud appsOS default
Where your voice goesStays on your MacExternal serversStays on your Mac
AI formattingyesyesno
Tone & custom rulesyesyesno
Edit by voiceyesPartialno
Works offlineyesnoyes
PriceFree+MonthlyFree

Want a large cloud model? Connect an OpenAI-compatible API any time (off by default).

What's stored, and where

Your voice never leaves your Mac.

Voice
With WhisperKit, zero is sent — processed on your Mac. Only Apple's recognizer sends audio to Apple, as the OS standard.
Transcript
Local only. Nothing kept on a server.
AI formatting
On-device, via Foundation Models.
Settings & history
Saved only on your Mac.

Pricing

Start free.

$0Free

The core of voice input, free today.

  • Instant / Streaming / Edit modes
  • On-device AI formatting
  • WhisperKit + Apple speech
  • Context formatting · multilingual
¥1,500/ mo · Pro

Smarter. Every install starts with a free 7-day trial.

  • Higher-accuracy formatting with large cloud models
  • Advanced custom rules
  • Priority support
  • Meeting mode (long transcription, speaker separation) — soon
Buy Pro¥12,000 / yr — 4 months free

FAQ

The things you're wondering.

Which Macs are supported?

Dictation runs on Apple Silicon, macOS 14+. AI formatting needs a Mac with Apple Intelligence (macOS 26+). On other Macs, voice punctuation and dictionary expansion still work.

Does it really work offline?

Yes. WhisperKit and Apple's on-device AI handle recognition and cleanup with no network.

Is my voice ever sent anywhere?

By default, not a single word. Connect a cloud model only if you choose to.

What languages?

Japanese, English and many more, with automatic language detection.

How is it different from built-in dictation?

It doesn't just transcribe — it removes fillers, adjusts tone, and edits by voice, all on your Mac.

Is my voice used to train AI?

No. Everything runs on your Mac — we never receive your data in the first place.

Is setup painful?

Just three permissions — microphone, speech recognition, accessibility — with a guided flow. One or two minutes, no account.

What does it cost?

The core is free. Pro is ¥1,500 / month or ¥12,000 / year. Every install includes a 7-day Pro trial, with no automatic billing.

Speak. Done. Nothing leaves your Mac.

Apple Silicon · Dictation on macOS 14+ / AI formatting needs macOS 26 + Apple Intelligence