Sotto
Fully local voice dictation for Linux. Press a shortcut, speak, and formatted text appears in whatever app has focus.
Sotto is voice dictation that runs entirely on your own computer. Press a global shortcut, speak, and the formatted text lands in whatever application has focus.
Open the Sotto docs
Installing, dictating, models, formatting.
Privacy and licence
What the local-only guarantee actually covers.
What makes it different
Every sample is processed on your machine. There is no account, no server component, and no
telemetry. The only network request Sotto ever makes is downloading a speech model, once, when you
ask it to. The LOCAL badge on the dictation pill is a permanent reminder of that.
Under the hood it runs whisper.cpp with GPU acceleration — ROCm on AMD, or Vulkan, or CUDA on NVIDIA.
What you get
- A global shortcut (Meta+Alt+D by default) that starts and stops dictation
- A small pill at the bottom of your active monitor with a live level meter and a live transcript
- Real writing, not word soup: punctuation and capitalisation from Whisper, paragraph breaks when you pause, and spoken “new line” / “new paragraph” commands
- Three ways to get text out: pasted into the focused app, typed key by key, or left on the clipboard
- A built-in notepad for dictating without inserting anywhere
- Settings for the model, shortcut, microphone, output method and monitor
Platform
Linux, Wayland, KDE Plasma (KWin) first. The dictation overlay uses wlr-layer-shell, so it is
never focusable and never tiled. Hyprland and other wlroots compositors are on the roadmap; GNOME
falls back to an ordinary always-on-top window; Windows and macOS are not supported.
Status: early v1, developed and tested on Arch Linux with Plasma 6 on Wayland and an AMD GPU.
Licence
Sotto is being released as open source under the Apache License 2.0. See Privacy and licence.