Local-first default — New installs use on-device Whisper for instant, private transcription with no API keys
New
Background model preload — Whisper model loads on app startup so first dictation is near-instant
New
No more "Processing..." pill — Overlay returns to idle immediately after recording stops
Improved
Streamlined defaults — Cleanup disabled by default for zero-latency raw transcription
Improved
Onboarding refresh — Emphasizes local/private processing and free-forever local mode
Improved