01
On-device & private
Apple’s on-device engines — including the new macOS 26 Speech — transcribe right on your Mac. Audio never has to leave the machine.
macOS dictation · on-device & private
(Est. 2026 — v1.0)
Stop typing. Start talking.
Dictation so native you’d swear Apple shipped it. Private on-device transcription, the best cloud engines when you want them, and a model picked per language.
Looks like Apple
made it.
(Features 01—09)
Only the Mac — never a cross-platform port. Built in Swift & AppKit to feel like it shipped with the system.
02
Truly native
Built in Swift & AppKit — tiny and instant, never an Electron wrapper. Made for macOS only, and always will be.
03
Lives in the menu bar
A global hotkey, a live waveform, and auto-insert straight into whatever app you’re typing in.
04
Cloud, your choice
Deepgram, OpenAI and ElevenLabs with your own key — and pick the exact model per provider.
05
Per-language engines
Route each language to the engine that handles it best — automatically, or by hand.
06
Downloadable models
Pre-fetch on-device language models so dictation starts the instant you press the key.
07
Made for your accent
Built-in dictation trips on a strong accent — VTT doesn’t. Switch to a cloud engine like OpenAI Whisper or Deepgram, trained on millions of voices, and your accent stops being the thing that breaks dictation.
08
You pick the language
VTT follows your keyboard. Flip your input source to Russian and you get Russian back; flip to English and you get English — with a keystroke you already use. No menu-diving, and no engine quietly translating what you said into a language you never asked for.
09
Never lose a transcript
Every dictation lands in a local history, newest first. Close the app, paste into the wrong window, change your mind — it’s still there. Re-paste any recent transcript with a click, or grab the latest straight from the menu bar. Kept on your Mac, nowhere else.
(Privacy)
Your voice never leaves your Mac.
Pick on-device and nothing ever touches a server — no account, no sign-in, no tracking. Reach for a cloud engine only when you want extra accuracy, and even then it’s your own API key talking straight to the provider.
One app, every engine worth having.
(Engines)
Apple Speech on-device
Deepgram
OpenAI
ElevenLabs
Bring your own key
Questions, answered.
(FAQ)Is VTT private? Does my audio leave my Mac?
By default VTT transcribes on-device using Apple’s Speech engine, so your audio never leaves your Mac. There’s no account and no analytics. If you enable a cloud engine, audio is sent directly to that provider using your own API key.
Is VTT free?
Yes — VTT is free to start, with no account required. On-device dictation costs nothing. Cloud engines are pay-as-you-go on your own provider API key, only if you choose to use them.
Which speech engines does VTT support?
Apple’s on-device Speech (including the macOS 26 models), plus optional cloud engines: Deepgram, OpenAI, and ElevenLabs. You can pick the engine per language.
What are the system requirements?
macOS 14 or later, on Apple Silicon or Intel. Download the build that matches your Mac — the site recommends the right one automatically.
Does VTT work offline?
Yes. On-device dictation works fully offline. An internet connection is only needed if you choose a cloud engine or to download additional language models.
How is VTT different from Apple’s built-in Dictation?
VTT keeps the on-device privacy of Apple Dictation but adds engine choice (including large cloud models via your own key), per-language routing, downloadable models, and a real menu-bar workflow with a global hotkey and auto-insert.
Does VTT handle strong accents?
Yes — that’s one of the best reasons to use it. If built-in dictation keeps mishearing your accent, switch VTT to a cloud engine like OpenAI Whisper or Deepgram. These models are trained on millions of voices across the world, so they’re far more forgiving of non-native and regional accents. You can even set it per language, so each one routes to the engine that understands you best.
Can I control which language I dictate in?
Completely. By default VTT transcribes in whatever language your keyboard is set to, so you choose the spoken language the same way you switch typing — one keystroke, no menus. Speak Russian and you get Russian text; switch the keyboard and you get that language instead. It never silently translates your words into English (or any other language) the way some dictation tools do. You can also pin a fixed language in settings if you prefer.
What happens to my transcripts — can I get one back?
Yes. Every dictation is saved to a local history on your Mac, newest first, so a transcript is never lost to a misclick or a wrong window. Re-paste any recent one from the History tab, or paste the latest straight from the menu bar. The history stays on your device — nothing is uploaded — and you can clear it whenever you like.
(Download — free)