FluidVoice v1.5.13

What's New

Added a compact notch presentation mode with cleaner top-overlay controls.
Added a compact pill bottom overlay with the app icon and responsive visualizer only.
Added hover details to activity stats bars so each day shows its word count.
Added optional notifications when AI cleanup falls back to raw transcription, opening History when clicked.

Smoothed stop-state overlay animations during final transcription.
Improved compact notch presentation and command overlay behavior.
Refined bottom overlay visualizer motion and removed unused preview space when preview is off.
Fixed recording overlays appearing on the wrong display in multi-monitor setups.
Fixed recording notch dismissal getting stuck behind stale hover state.
Fixed prompt-test runs leaving the bottom overlay in its release animation state.
Removed the duplicate Live Preview toggle from Settings.

Fixed very short utterances being dropped by padding them with trailing silence before transcription.
Fixed repeated middle phrases in long Cohere transcripts by delegating long-audio merging to FluidAudio.
Fixed AirPods route changes during recording and polished bottom overlay fade/visualizer behavior.

Fixed dictation cleanup prompts so question-like transcripts are cleaned instead of answered, with ${transcript} template support.
Fixed AI post-processing errors being typed into dictation output.
Added history warnings when AI cleanup falls back to raw transcription.
Fixed custom AI providers resetting after reopening Settings.
Fixed rollback picking the wrong saved app version.
Fixed startup focus retries bringing FluidVoice back over other apps after clicking away.

Updated the LM Studio setup guide link to the OpenAI-compatible chat completions documentation.
Updated README install links and Homebrew cask instructions.

Thanks @domdomegg for the short-utterance padding, prompt-template, Anthropic, Ollama, and AI cleanup contributions.
Thanks @shhac for the hotkey display fixes.