OpenWhispr

OpenWhispr turns speech into text with local Whisper and Parakeet models or a cloud provider of your choice. It includes dictation, transcription history, and configurable shortcuts.

This community package downloads the official, unmodified Linux tarball at install time. It runs using X11 or XWayland and keeps models and settings in the app's private directories.

Dictation and automatic paste are tested with X11 target applications. Native Wayland targets and compositor-specific shortcuts are unverified. Browser sign-in is unavailable in this package; email sign-in may work but is untested, or you can continue without an account. Direct input-device and PipeWire access are not granted by default.

Transcribing audio from a URL may update the bundled yt-dlp to a nightly build in the app's private cache. These updates are managed by yt-dlp rather than FlatPark.

Optional system-audio capture needs direct PipeWire access, which also exposes other audio and video streams: flatpak override --user --filesystem=xdg-run/pipewire-0 com.gizmolabs.openwhispr

Undo optional overrides: flatpak override --user --reset com.gizmolabs.openwhispr