Works offline · v2.23 out now

Hold a key.
Speak. It's typed.

TalkKey turns your voice into text wherever your cursor already is — Slack, Mail, your editor, a terminal. Nothing leaves your Mac unless you ask it to.

◆ Apple Silicon ◆ macOS 14 Sonoma or later ◆ Signed & notarized
Hold Right ⌘ — or press this key
Zero friction

No gap between the thought
and the text.

The audio engine is prepared once when TalkKey launches, so pressing the key starts recording immediately — no clipped first words, no waiting for the microphone to wake up.

~100msFrom key press to listening
Faster than typing it out
0Bytes sent in offline mode
99+Languages recognised
Speech models
Balanced

Installed · 486 MB

ACTIVE
Most Accurate

Best accuracy, slowest · 1.5 GB

DOWNLOAD
Dictionary
Playwright

also heard as: плейрайт, play right

ON
pytest

Corrected automatically

ON
Three keys, three jobs

Paste it, polish it,
or translate it.

Each mode has its own key. Your hands never leave the keyboard, and you decide what happens the moment you let go.

Direct paste

Straight to the cursor

Let go and the text is typed where you were already working. It lands in your clipboard first, so a failed paste never costs you the words.

Review

Read before you send

See the transcription first, edit it, and optionally restyle it — formal, casual, concise, bullet points — before it goes anywhere.

Fn
Translate

Speak one language, write another

Talk in your own language and get the text in the one you need. Useful when you think faster than you translate.

Custom dictionary

It learns the words
you actually use.

Product names, colleagues, technical jargon — the words every dictation app mangles. Add them once and TalkKey gets them right.

Each term works twice: it primes the speech model before it listens, then repairs anything that still slips through. Plurals and inflections are left alone, so it fixes the word without breaking the sentence.

PlaywrightpytestAllure GitHubKubernetes+ yours
Live correction
Let's run the playwrite suite and push the report to git hub before the standup.
Everything in the box

Built like a tool,
not a demo.

Every feature here exists because dictation breaks somewhere in real use. These are the fixes.

Fully offline

Whisper runs on your Mac. No account, no upload, no internet — your voice never leaves the machine.

NEW

Custom dictionary

Teach it the words it keeps getting wrong, with optional mishearings for the stubborn ones.

NEW

Floating control bar

A small pill above the Dock. Click to start and stop, watch the waveform, switch language without opening a window.

NEW

Model library

Five Whisper models, 39 MB to 1.5 GB. See what each takes on disk, install and delete them freely.

Auto language detect

Switch languages between recordings and TalkKey works out which one you spoke — offline too.

Never loses a word

Every transcript hits the clipboard before it's typed. If a paste can't go through, a card hands it back to you.

AI rewriting

Turn a rambling thought into a clean paragraph, a bulleted list, or a professional email.

Local history

Every transcription kept on your Mac, ready to copy again. Clear it whenever you like.

Quiet updates

Signed, notarized, and updated in place. Lifetime updates, no subscription.

Works anywhere
you can type.

TalkKey lives in the menu bar and types into the app you're already using — no plugins, no integrations to set up.

Slack Notion VS Code Cursor Claude OpenAI Gmail Telegram Discord Linear Figma GitHub Obsidian Arc Chrome X
For developers

Brief your agent
out loud.

Good prompts are long, and typing them is the bottleneck. Describe the change the way you'd explain it to a colleague, and let Claude Code, Cursor or Windsurf take it from there.

Put your stack in the dictionary and it will spell your framework names, file paths and libraries correctly every time.

claude code — ~/dev/talkkey
Add a dark mode toggle to settings. Persist it
   in localStorage and apply it without a reload.

Read settings.tsx — 245 lines
Updated theme provider
Writing tests…
Pricing

Free forever.
Pro once, if you want it.

No subscription. Buy it once and every future update is included.

Free

$0

Everything you need for private, on-device dictation.

  • On-device Whisper, fully offline
  • All five speech models
  • Custom dictionary & control bar
  • Auto language detection
  • Cloud mode
  • Review & translate modes
Download free
RECOMMENDED

Pro

$15one-time

Unlock every mode and support a one-person project.

  • Everything in Free
  • Review mode with AI rewriting
  • Translate to 12+ languages
  • Cloud mode with your own API key
  • Unlimited daily transcriptions
  • Lifetime updates
Get Pro license

Questions,
answered.

Does my voice ever leave my Mac?

Not in offline mode — the speech model runs locally and nothing is uploaded. Cloud mode is opt-in and uses your own OpenAI key, so audio goes only to OpenAI and nothing is stored by TalkKey.

Do I need an OpenAI API key?

No. Offline mode is the default and needs no key and no account. A key is only required if you choose Cloud mode or AI rewriting.

Which Mac do I need?

An Apple Silicon Mac (M1 or newer) running macOS 14 Sonoma or later. The app is signed and notarized by Apple, so it opens without warnings.

How much disk space do the models take?

Between 39 MB and 1.5 GB depending on which you install. The model library shows exactly what each one uses and lets you delete any of them at any time.

Can it handle more than one language?

Yes. With auto-detect on you can switch languages between recordings without touching settings, and it works in offline mode too.

Is Pro a subscription?

No. $15 once, and updates are included forever.

Stop typing what you
could just say.

Free to use, works offline, installs in under a minute.

Download TalkKey