v1.43.0

What's Changed

Added

• Added hyprwhspr transcribe INPUT for transcribing WAV and MP3 files from the command line.
  • Write transcripts to stdout or use -o PATH.
  • Override the language with --lang.
  • Apply configured text cleanup with --clean.
  • Reuses an idle service's loaded model, falling back to standalone transcription when the service is unavailable.
  • Supports local and REST backends. The realtime WebSocket backend remains live-capture only.
• Added KDE Plasma focused-window detection through the AT-SPI accessibility bridge.
  • hyprwhspr setup can enable the bridge.
  • hyprwhspr config focused-window now provides actionable diagnostics when window detection is unavailable.

Fixed

• Prevented PipeWire/PulseAudio output monitors from being selected as microphone inputs.
• Re-test previously rejected monitor sources when the default audio source changes, instead of permanently latching the rejection.
• Improved microphone diagnostics and live-input testing guidance.
• Updated pywhispercpp to 1.5.1 and its upstream whisper.cpp revision, restoring source builds with CUDA 13.3. Thanks to @BestSithInEU.

New Contributors

• @BestSithInEU made their first contribution in #247.

Full Changelog: v1.42.3...v1.43.0