Skip to content

v1.1.10 — Live Transcription with Apple Speech Recognition

Choose a tag to compare

View all tags
@iainporter iainporter released this 02 Mar 10:52
· 30 commits to main since this release
v1.1.10
867372a

What's New

Real-time live transcript during recording — replaced the slow WhisperKit streaming (10-second chunked updates) with Apple's SFSpeechRecognizer for near-instant word-by-word transcription. Text appears as people speak, with a scrollable transcript area showing the latest ~15 lines.

WhisperKit batch transcription still runs after recording stops for the high-quality final stored transcript.

Changes

  • Live speech recognition — SFSpeechRecognizer with automatic session restart, thread-safe buffer handling, and 50ms throttled UI updates
  • Scrollable live transcript — flowing text with scrollbar to review earlier text during recording
  • ~1.5GB less RAM while recording — WhisperKit no longer loads at recording start, only when batch transcription begins after stop
  • On-device speech recognition toggle — new setting (defaults to on) keeps audio processing fully local
  • Graceful degradation — if Speech Recognition permission is denied, recording works normally without live text

Bug Fix

  • Fixed permission prompt loop when granting Screen Recording permission on fresh install — Speech Recognition authorization is now requested lazily on first recording, not at app launch

SHA-256

40c2b3ec9cba7a46e2dc6bf0e2b48c35d09f752b82100b0bd95b9497b99c1c0c
Assets 3
Loading