Voice-to-text dictation app with local (Nvidia Parakeet/Whisper) and cloud models (BYOK). Privacy-first and available cross-platform.
JavaScript aianthropiccross-platformgeminigroqlinuxmacosnvidiaopen-sourceopenaiparakeetspeech-to-texttranscribewhisperwindows
/**
* Single owner of the "may a new dictation start?" guard over AudioManager's
* getState() snapshot. Three call sites in useAudioRecording.js (the hotkey
* toggle, performStartRecording, and the hover-prepare handler) used to carry
* hand-copied flag lists that drifted: hover-prepare had lost the two
* streaming transition flags, so only AudioManager's internal stop promise
* kept a prepare from racing a streaming start/finalize. Pure module so the
* flag set is unit-testable (test/utils/dictationReadiness.test.js).
*/
export interface DictationReadinessState {
isRecording?: boolean;
isProcessing?: boolean;
isStreaming?: boolean;
isStreamingStartInProgress?: boolean;
isFinalizingStreaming?: boolean;
}
export function canStartDictation(state: DictationReadinessState | null | undefined): boolean {
if (!state) return false;
return !(
state.isRecording ||
state.isProcessing ||
state.isStreaming ||
state.isStreamingStartInProgress ||
state.isFinalizingStreaming
);
}