Voice-to-text that runs on your machine

Press a hotkey → speak → text appears wherever you're typing.
Free. The default local Whisper backend works offline with no API key and no account.

~600 MB download · Powered by Whisper AI · Open source · MIT license · All downloads

How it works

Step 1

Press your hotkey

A small indicator appears in the corner. You're recording.

Step 2

Speak naturally

Talk at your normal pace. The waveform shows the mic is active.

Step 3

Text appears at cursor

Press the hotkey again. Your words are transcribed and pasted wherever you're typing.

The default local Whisper backend works offline and keeps audio on your machine. An optional NVIDIA NIM backend, if you enable it with your own NVIDIA API key, sends audio to NVIDIA.
No account and no subscription; the default local backend needs no API key
Works in any app: browser, editor, terminal, chat
Linux, macOS, and Windows support
Draggable indicator, remembers its position
Optional wake word and end word detection
Free and open source (MIT)

Microphone: granted automatically by ALSA/PipeWire
Hotkey (Wayland): add yourself to the input group:
sudo usermod -aG input $USER
then log out and back in (one-time setup)
Linux X11: no extra steps needed

Microphone: macOS prompts automatically on first use. Click Allow.
Accessibility: required for the auto-paste feature. macOS prompts you; or go to System Settings → Privacy & Security → Accessibility.

Need the full install walkthrough, troubleshooting, or developer guide?