v1.6.8

# FluidVoice v1.6.8

## Changes

- Added optional offline speaker labeling for file transcription, with timestamps, speaker-aware history, and text/JSON export.
- Added a saved microphone priority list with drag-to-reorder, device history, availability status, and quick access from microphone-change notices.

## Fixes

- Fixed modifier-only dictation shortcuts incorrectly firing after unrelated Shift-key combinations.
- Corrected the displayed download size for Whisper Large Turbo.
- Fixed silent or incorrect microphone selection in clamshell mode and after disconnects, reconnects, or audio-route changes.
- Fixed first-run microphone selection to follow the macOS default while preserving existing choices, with an onboarding picker and seamless live input preview.
- Fixed temperature settings being ignored for non-reasoning OpenAI models accessed through compatible providers.
- Fixed streamed tool calls being lost when compatible providers include reasoning content.
- Fixed restoring focus after dictation bringing every window of multi-window apps to the front.

## Contributors

Thanks to @postoso for offline speaker labeling and streamed tool-call handling, @Erik5000 for offline speaker labeling, @double-thinker for correcting the Whisper Large Turbo download size, and @YuriNachos for correcting temperature handling, modifier-only shortcuts, and focus restoration in 1.6.8.