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).