soba changelog
← Back to all updates

Improved Voice Dictation Pipeline and Sound Effects

bugfixvoiceai

Commit: 297c37d

Enhanced voice dictation functionality and fixed sound effect issues.

Changes

  • Fixed transcript tidy not applying by delaying WebSocket closure for final committed_transcript.
  • Reduced 20s latency on transcript tidy by adding reasoning_effort:‘low’ to GPT-5 API requests.
  • Resolved sound effects not playing by moving setupSoundFXCallbacks() to early startup and adjusting AVAudioEngine timing.
  • Introduced AI panel color customization and cleaned up tag aggregates.
  • Documented voice dictation pipeline and insights in CLAUDE.md.