feat: add voice-to-text push-to-talk transcription

[jhwheeler]

Summary

• Adds a push-to-talk voice input system using ffmpeg for recording and faster-whisper for local transcription
• voice-to-text (bash): toggle script — first press starts recording, second press stops, transcribes, and types the result into the focused window via wtype + copies to clipboard
• transcribe.py (python): transcribes a WAV file using faster-whisper (base model, CPU, int8 quantization)

Setup

sudo pacman -S wtype
python3 -m venv ~/venvs/voice-to-text
~/venvs/voice-to-text/bin/pip install faster-whisper

Also requires config changes (not in this repo):

• Hyprland keybinding: bind = $mainMod, R, exec, ~/scripts/voice-to-text
• Mako notification rule for app-name="Voice Recorder" with ignore-timeout=0

Test plan

• Press ALT+R — "Recording..." notification appears and persists
• Speak, press ALT+R again — "Transcribing..." then "Transcribed" notification with text
• Verify transcribed text is typed into terminal and available in clipboard (wl-paste)

🤖 Generated with Claude Code