Stream dictation via OpenAI Realtime and simplify retry UI
This commit is contained in:
@@ -5,6 +5,7 @@ export interface SpeechmaticsAudioConfig {
|
||||
onSpeechStart?: () => void;
|
||||
onSpeechEnd?: () => void;
|
||||
onError?: (error: Error) => void;
|
||||
enableContinuousStreaming?: boolean;
|
||||
volumeThreshold: number;
|
||||
silenceDuration: number;
|
||||
speechConfirmationDuration: number;
|
||||
|
||||
Reference in New Issue
Block a user