GNOME Shell integration for whisper.cpp (Linux desktop)

[bcelary]

Built this for Linux users wanting Wispr Flow-style experience - native shell integration without running a separate GUI app.

And yes, bunch of what comes next is spit out by Claude Code :)

What it is:

• GNOME Shell extension + lightweight D-Bus service
• Keyboard shortcut → speak → get text (auto-type or clipboard or preview)
• Automatically starts/manages whisper.cpp server - no manual server management
• System tray indicator, configurable notification modes
• Whisper.cpp models (runs in auto language mode but can be overridden), VAD support

Architecture:
Three components working together:

• Extension: UI/shortcuts in GNOME Shell
• D-Bus service: Python backend (audio capture, server lifecycle, communication)
• whisper.cpp server: auto-started by service on startup

Why this approach:
Wanted minimal background overhead - no dedicated app, just small service integrated directly with desktop environment. Press shortcut, speak, continue working while transcription happens in background. Service handles server lifecycle automatically.

Target users:
Linux users running local whisper.cpp who want seamless desktop integration without complex scripts or heavy GUI apps.

GitHub: https://github.com/bcelary/gnome-speech2text

Tech note: Developed with AI assistance (Claude) - not vibe-coded. Started from working codebase.

Forked from kavehtehrani/gnome-speech2text, heavily refactored for whisper.cpp support and improved UX.

Settings window:

Notification area during recording:

The feedback is given with the state of the notification area but if you want, you can enable blocking recording and/or preview modals: