Speed of Sound

Voice typing for the Linux desktop.

• Offline, on-device transcription using Whisper. No data leaves your machine.
• Multiple activation options: click the in-app button or use a global keyboard shortcut.
• Types the result directly into any focused application using Portals for wide desktop support (X11, Wayland).
• Multi-language support with switchable primary and secondary languages on the fly.
• Works out of the box with the built-in Whisper Tiny model. Download additional models from within the app to improve accuracy.
• Optional text polishing with LLMs (Anthropic, Google, OpenAI), with support for a custom context and vocabulary.
• Supports self-hosted services like vLLM, Ollama, and llama.cpp (cloud services supported but not required).