This release expands Soba’s voice features with OpenAI speech support, improves dictation cleanup, and adds a focused Morning Pages progress indicator. It also tightens autosave behavior so note edits remain reliable while navigating quickly between documents.
Added
- Added OpenAI as a speech provider for live transcription and read aloud, alongside the existing ElevenLabs options.
- Added voice settings for provider, model, API key reuse, API key override, and read-aloud voice selection.
- Added a Morning Pages word-count progress indicator that fills toward 750 words and completes as a green status marker.
- Added a Post-Dictation Cleanup setting so cleanup can be turned off when raw transcripts are preferred.
Improved
- Made post-dictation cleanup context-aware for notes, Morning Pages, and AI chat prompts.
- Improved OpenAI realtime dictation startup behavior so the UI only shows recording once capture is actually live.
- Reduced redundant autosave, backup, cache, and watcher work during normal editing and dictation.
- Improved note-switching resilience by scoping editor state per note and collapsing duplicate chat context reloads.
Fixed
- Fixed OpenAI realtime transcription setup for transcription sessions.
- Fixed Morning Pages dictation cleanup so it uses the light prose cleanup path.
- Fixed Settings launch crashes caused by missing voice transcriber environment wiring.
- Fixed a fast note-switch autosave crash after editing, navigating away, and returning to the original note.