These are my custom scripts used to interact with Whisper / OpenAI through keyboard shortcuts and voice input.
I am using ArchLinux and the xfce Keyboard settings for key bindings.
voice.sh is bound to F2 - it activates voice recording. Once stopped by the other script, it sends a wav file to whisper.cpp.
stop_voice.py is bound to F4 - it stops voice recording (uses Linux pipe).
voice_tr.sh is bound to F3 - it performs the same voice recording, but with a more capable model. It is used for translations.
This is how it works for me:
process_clipboard.sh is bound to F1 - it sends the currently selected text to the ChatGPT aichat project.
edit_clipboard.sh is bound to F5 - it opens an editable window with the current clipboard contents. After the window is closed, it sends the contents to process_clipboard.sh.
This is how it works:
notify_history.sh is used to read the notification history. It is bound to a shortcut in the system tray.
aichat - as mentioned previously
dunst, zenity - for notifications
whisper.cpp - for voice transcription