TobiVoice is a native macOS app that turns speech into text almost instantly. Transcription runs on local AI models by default, so processing happens offline and your audio stays on your Mac. It’s built to be the kind of voice-to-text you actually keep using all day: fast, accurate and private.
What it does
- Accurate, near-instant transcription using local models, no cloud round-trip. Whisper and Parakeet both run on-device.
- Local by default: the on-device models need no account, no network and no API key. Cloud and streaming providers are available if you want them, but they are opt-in and off unless you add a key yourself.
- Global shortcuts for quick recording and push-to-talk, with a warning that tells you when macOS Secure Input is silently swallowing them and which app to close.
- Context-aware modes that adapt to the app you’re writing in, plus a personal dictionary for custom terms, per-mode replacements and preserved brand casing.
- Meeting mode for long recordings: you get the full transcript and a bullet summary in your history instead of a wall of text pasted into whatever had focus.
- Read Aloud: have a summary or a selection spoken back to you, with your choice of voice and speed. It can use a local speech engine, so this stays offline too.
- Quick language switching from a shortcut or the menu bar, for anyone who dictates in more than one language.
Requires macOS 14 or later. The app is a focused, local-first alternative to cloud dictation services.