refactor(voice): defer speech confirmation and add diagnostics
This commit is contained in:
@@ -22,14 +22,13 @@ export function RealtimeVoiceOverlay({
|
||||
onStop,
|
||||
}: RealtimeVoiceOverlayProps) {
|
||||
const { theme } = useUnistyles();
|
||||
const { volume, isDetecting, isSpeaking } = useVoiceTelemetry();
|
||||
const { volume, isSpeaking } = useVoiceTelemetry();
|
||||
return (
|
||||
<View style={styles.container}>
|
||||
<View style={styles.meterContainer}>
|
||||
<VolumeMeter
|
||||
volume={volume}
|
||||
isMuted={isMuted}
|
||||
isDetecting={isDetecting}
|
||||
isSpeaking={isSpeaking}
|
||||
orientation="horizontal"
|
||||
/>
|
||||
|
||||
Reference in New Issue
Block a user