
Description
Want to dictate into a chat box, editor or terminal without sending your voice to a cloud or paying a subscription? With Wavo you hold a hotkey, speak, release and the text lands where you were typing, with recognition done entirely on your machine: no account, and your voice never leaves your computer.
On macOS 15 or later it runs on the Neural Engine, and the author reports 11 s of speech becoming text in 61 ms on an M2. It also runs on Windows and Linux (X11 and Wayland) in about 10 MB, MIT licensed.
Hold to talk: Hold the hotkey and speak, or tap to start and stop; Esc cancels.
Pastes anywhere: Text goes where you're typing and the clipboard keeps what it held.
Fully local: Only the model is downloaded, once, then it works offline.
100 languages: Parakeet Ultra for 25 European languages, Whisper for the rest.
History: Every dictation with its recording, searchable, playable, re-transcribable and exportable.
Picks the mic: A USB mic first, then the built-in one, Bluetooth only when nothing else is available.
On macOS 15 or later it runs on the Neural Engine, and the author reports 11 s of speech becoming text in 61 ms on an M2. It also runs on Windows and Linux (X11 and Wayland) in about 10 MB, MIT licensed.
Features
Hold to talk: Hold the hotkey and speak, or tap to start and stop; Esc cancels.
Pastes anywhere: Text goes where you're typing and the clipboard keeps what it held.
Fully local: Only the model is downloaded, once, then it works offline.
100 languages: Parakeet Ultra for 25 European languages, Whisper for the rest.
History: Every dictation with its recording, searchable, playable, re-transcribable and exportable.
Picks the mic: A USB mic first, then the built-in one, Bluetooth only when nothing else is available.
