
A minimalist, privacy-focused desktop application for offline speech-to-text. Converts voice
input directly into any active window (editors, browsers, IDEs, AI assistants) using Whisper or
NVIDIA Parakeet models locally for speech recognition.
For system tray integration on GNOME and automatic typing on Wayland, see the Desktop Environment
Support Guide.
Features:
<ul><li>Offline speech-to-text, privacy-first: all processing happens locally</li><li>Portable AppImage and Flatpak packages</li><li>Cross-platform support for X11 and Wayland</li><li>Linux desktop integration with GNOME, KDE, and others</li><li>GPU and CPU support with Vulkan acceleration and CPU fallback</li><li>Voice typing or clipboard mode</li><li>Flexible audio recording with ALSA or PulseAudio/PipeWire</li><li>Multi-language support, custom hotkey binding, visual notifications</li><li>Model management: switch between Whisper models (base, small, medium, large-v3-turbo, large-v3)
and NVIDIA Parakeet TDT 0.6B v3 via tray or CLI</li></ul>