Skip to content

v3.14.0

Choose a tag to compare

View all tags
@snowy-0wl snowy-0wl released this 27 Aug 16:07
· 23 commits to main since this release
v3.14.0
c8fb276

Added

  • Insert transcription into the active field after Listen (Settings → General, on by default)
  • Recording overlay with input meters when Visual Alerts are on

Changed

  • Default speech-to-text engine is Parakeet; Whisper is optional. Switching to Whisper does not start a download until you tap Download on a model
  • Whisper acceleration is Metal or Neural Engine (macOS 14+), chosen in Models. Neural uses the Core ML encoder for every listen
  • Voice quality is Fast / Balanced / Quality (2 / 8 / 10 flow-matching steps). The old 3–30 slider sat past where this voice improves
  • Voice playback primes the audio graph so the first syllable is not clipped
  • Whisper language default stays Auto-Detect (or English if auto is missing). Opening Models no longer snaps the language to Bulgarian
  • First-run download is about 1 GB (TTS plus the selected STT engine)
  • Sparkle waits until first-run setup finishes, and no longer shows a start-up error when XPC helpers live only in the Sparkle framework
Assets 3
Loading