Add voice processing settings and a speak-time bypass

poc/src/audio-service.ts

@@ -25,6 +25,8 @@ export interface AudioServiceOptions {
     getVoiceContext?: () => Promise<{ guildId?: string | null; channelId?: string | null } | undefined>;
     getSpeakerName?: (userId: string) => Promise<string | undefined>;
     onTranscript?: (event: Record<string, unknown>) => void;
+    /** Called true when a TTS burst starts and false when the queue drains. */
+    onSpeakActivity?: (active: boolean) => void;
 }
 
 export interface SpeakResult {
@@ -51,6 +53,8 @@ export class AudioService {
     private readonly speakerIntervals: SpeakerInterval[] = [];
     private readonly speakerNames = new Map<string, string>();
     private playing = false;
+    /** True while a TTS burst holds the voice-processing bypass. */
+    private bypassed = false;
     private duckEnabled = true;
     private nextId = 1;
     private ducked = false;
@@ -381,6 +385,14 @@ export class AudioService {
         }
 
         this.playing = true;
+        if (!this.bypassed) {
+            this.bypassed = true;
+            try {
+                this.options.onSpeakActivity?.(true);
+            } catch {
+                // Best effort: speech still plays.
+            }
+        }
         try {
             // Streaming word mode: the wrapper writes the WAV while it synthesizes, so
             // playback starts on the first word instead of after the whole line.
@@ -408,6 +420,14 @@ export class AudioService {
         } finally {
             this.playing = false;
             await this.applyDuck(false);
+            if (this.bypassed && this.queue.length === 0) {
+                this.bypassed = false;
+                try {
+                    this.options.onSpeakActivity?.(false);
+                } catch {
+                    // Best effort.
+                }
+            }
             void this.pump();
         }
     }