OpenHush
by OpenHush Contributors
Open-source voice-to-text that acts as a seamless whisper keyboard
About
OpenHush is a privacy-first voice-to-text tool that runs entirely on your local machine. No cloud, no subscriptions, no telemetry. Just fast, accurate transcription using OpenAI's Whisper model. Features: - Push-to-talk and toggle recording modes - Auto-paste transcribed text at your cursor - GPU acceleration (NVIDIA CUDA, AMD ROCm, Vulkan) - 99+ language support with auto-detection - Optional translation to English - LLM-powered grammar correction via Ollama - Voice Activity Detection for continuous dictation - Custom vocabulary for domain-specific terms - File transcription for WAV/MP3 recordings - D-Bus integration for scripting and automation OpenHush is designed for privacy-conscious users who need fast, accurate dictation without sending their voice data to the cloud.
Permissions
Everything this app can reach outside its sandbox. Nothing here weakens it.
- inter-process communication (needed for X11)
--share=ipc - Wayland display
--socket=wayland - X11 display, only when Wayland is unavailable
--socket=fallback-x11 - audio
--socket=pulseaudio - GPU acceleration
--device=dri - notifications
--talk-name=org.freedesktop.Notifications - system tray
--talk-name=org.kde.StatusNotifierWatcher - owns bus name org.openhush.Daemon1
--own-name=org.openhush.Daemon1 - network
--share=network - access to xdg-config/openhush
--filesystem=xdg-config/openhush:create - access to xdg-data/openhush
--filesystem=xdg-data/openhush:create
Who publishes this
Nobody has confirmed that the publisher controls this app id. Flatsonar warns before installing.
- org.openhush.OpenHush claims the domain openhush.org; only its owner can prove that
publisher:namespace
Maintenance
Recent activity on the upstream repository, or built and reviewed by Flathub.
No staleness signals: recent activity, or built and reviewed by Flathub.