Loading
whisrs is the better call if you live on a Linux tiling-compositor desktop, want a free MIT tool you can read and script from the shell, and are happy to choose a backend and set up key permissions yourself. Its default backend is cloud (Groq), so audio only stays on the device when you switch to whisper.cpp or a local sidecar. Typilot wins when you also use macOS or Windows, want keyboard AI commands and autocomplete that need no spoken prompt, or need diarized meeting notes with every model running on the device.
Yes. Both turn a hotkey and your voice into typed text on Linux, and both can run Whisper locally. whisrs is a free MIT Rust daemon for Linux only. Typilot adds macOS and Windows, 27 keyboard-triggered AI commands via local Ollama, ghost-text autocomplete, and diarized meeting transcription.
The software is free and MIT licensed. Privacy depends on the backend: its default is Groq, and Deepgram and OpenAI are also supported, all of which send audio to a cloud API. Choosing whisper.cpp or a local sidecar keeps recognition on your machine. Typilot has no cloud speech option, so audio never leaves the device.
No. The project describes itself as a Linux dictation tool and lists Wayland, X11, and several Linux compositors. Typilot runs on Linux, macOS, and Windows.
whisrs is built for that workflow: a daemon and CLI you bind to any compositor hotkey, with MIT source you can read. Typilot is a desktop app that also adds keyboard AI commands, autocomplete, and diarized meeting notes, so it suits people who want those features and a licence that follows them across operating systems.