100% on-device · macOS 14+ · Apple Silicon

Speak. It types.
Nothing leaves your Mac.

TrivialVoice is a private, local-only dictation app for macOS. Hold a key, speak, release — your words are transcribed on your device and inserted into whatever app you're using. No account. No cloud. No telemetry.

  • 🔒 Never sends your voice anywhere
  • ⚡️ Works in any text field
  • 🆓 Free in v1
0bytes sent to a server
<2sto insert short dictations
2on-device speech models
apps you can dictate into

How it works

Three keys to your keyboard-free flow

Setup takes a minute. After that it disappears into the background and just works.

  1. 1

    Hold your key

    Pick fn or right ⌘ as your dictate key during setup. Press and hold it anywhere in macOS.

  2. 2

    Speak naturally

    A floating indicator shows the target app and a live waveform. Say “new line” or “new paragraph” for breaks.

  3. 3

    Release — it's typed

    Let go and your transcribed text lands in the focused field. Double-press for hands-free mode; esc to cancel.

Private by architecture

Your voice is nobody else's business

Privacy isn't a setting you toggle — it's how TrivialVoice is built. Audio, transcripts, your dictionary, history and stats are processed and stored entirely on your Mac.

  • Speech and transcription run fully on-device
  • No account, no sign-in, no license activation
  • Zero telemetry — nothing is phoned home
  • Works completely offline once models are downloaded
  • History retention is yours to set: from never to forever

Features

Small app. Serious dictation.

Everything you need to talk instead of type — and nothing you don't.

🖥️

Types into any app

Native apps get direct Accessibility insertion. Terminals and Electron apps (Ghostty, VS Code, Zed) get a reliable ⌘V paste fallback.

🎯

Menu-bar native

A tidy microphone icon lives in your menu bar: open the app, jump to your dictionary, check for updates or send feedback.

🗂️

Searchable history

Every dictation is saved locally. Search it, delete items, clear everything, or export to a plain-text file.

📖

Personal dictionary

Teach it names, jargon and acronyms. Map a heard phrase to the exact replacement you want, with optional case sensitivity.

Smart formatting

Say “new line” or “new paragraph” for breaks. Deterministic voice commands and dictionary rules apply automatically.

📊

Writing stats

See dictations, words, 7-day totals and averages on the Home tab — computed and kept entirely on your Mac.

On-device speech models

Two engines, zero cloud

Choose the model that fits your language and your Mac. Both run locally on the Apple Neural Engine.

Recommended default

Parakeet

Fast, accurate transcription optimized for Apple Silicon. Batch-decodes a short clip in a fraction of a second on the ANE.

  • Best speed & accuracy on Apple Silicon
  • Runs on the Neural Engine
  • Downloaded once during setup

Language fallback

Whisper

Broad language coverage for when Parakeet doesn't support your preferred dictation language. Still fully local.

  • Wide language support
  • Runs entirely on-device
  • Swap it in from Settings

Optional on-device cleanup (punctuation & casing polish via MLX) is on the roadmap and disabled by default — deterministic voice commands and dictionary rules work today.

Ready to stop typing?

TrivialVoice is free in v1. Requires an Apple Silicon Mac running macOS 14 or later.

On first launch you'll grant Microphone and Accessibility permissions, choose your dictate key, and download a speech model. That's it.

FAQ

Questions, answered

Does my voice or text ever go to the cloud?

No. TrivialVoice processes speech and stores everything locally. It never transmits audio, transcripts, your dictionary, history, stats or telemetry to any remote service. The only network use is a one-time speech-model download.

Do I need an account or subscription?

No. There's no sign-in, no license activation and no payment in v1. Local dictation is never gated behind a paywall.

Which Macs are supported?

Apple Silicon Macs running macOS 14 or later. The speech models are optimized for the Apple Neural Engine.

Can I use it offline?

Yes. After the initial model download, dictation works with no internet connection at all.

Which apps can I dictate into?

Any focused text field. Native apps get direct Accessibility insertion; terminals and Electron apps (like Ghostty, VS Code and Zed) use a clipboard-paste fallback. If Accessibility is off, your text is safely left on the clipboard to paste with ⌘V — it's never lost.

Which speech model should I pick?

Start with Parakeet — it's the recommended default and the fastest on Apple Silicon. Switch to Whisper if you need a language Parakeet doesn't cover.