@omote/babylon 0.2.0 → 0.3.2

package/dist/index.d.cts

@@ -1,5 +1,6 @@
 import * as _omote_core from '@omote/core';
-import { FaceCompositorConfig, CharacterControllerConfig, EmotionWeights, ConversationalState, A2EOrchestratorConfig } from '@omote/core';
+import { FaceCompositorConfig, CharacterControllerConfig, FrameSource, TTSBackend, TTSSpeakerConfig, SpeechListenerConfig, TranscriptResult, VoiceOrchestratorConfig, EmotionWeights, ConversationalState, CharacterProfile, TTSSpeaker, SpeechListener } from '@omote/core';
+export { FrameSource, TTSSpeakerConfig as TTSConfig } from '@omote/core';
 import { AbstractMesh, TransformNode, Scene, Camera } from '@babylonjs/core';
 
 /**
@@ -47,13 +48,6 @@ interface SceneDiscoveryResult {
  */
 declare function discoverScene(root: AbstractMesh): SceneDiscoveryResult;
 
-/** Generic frame source -- any object that emits 'frame' events */
-interface FrameSource {
-    on(event: 'frame', callback: (frame: {
-        blendshapes: Float32Array;
-    }) => void): void;
-    off?(event: 'frame', callback: (...args: any[]) => void): void;
-}
 interface OmoteAvatarOptions {
     /** Root mesh of the avatar (typically loaded via SceneLoader) */
     target: AbstractMesh;
@@ -82,6 +76,9 @@ declare class OmoteAvatar {
     private _camera;
     private frameSourceCallback;
     private connectedSource;
+    private ttsSpeaker;
+    private speechListener;
+    private voiceOrchestrator;
     private renderCallback;
     private lastTime;
     constructor(options: OmoteAvatarOptions);
@@ -107,6 +104,72 @@ declare class OmoteAvatar {
     connectFrameSource(source: FrameSource): void;
     /** Disconnect the current frame source (if any). */
     disconnectFrameSource(): void;
+    /**
+     * Connect a TTS backend for speak() / streamText() support.
+     * Loads LAM model and creates internal PlaybackPipeline.
+     *
+     * @param tts - TTS backend (e.g., KokoroTTSInference, ElevenLabs adapter)
+     * @param config - A2E, expression profile, and playback configuration
+     */
+    /** Warm up AudioContext for iOS/Safari autoplay policy. Call from user gesture. */
+    warmup(): Promise<void>;
+    connectSpeaker(tts: TTSBackend, config?: TTSSpeakerConfig): Promise<void>;
+    /**
+     * Synthesize text and play with lip sync.
+     * Auto-aborts previous speak if still in progress.
+     *
+     * @param text - Text to synthesize
+     * @param options - Optional voice override and abort signal
+     */
+    speak(text: string, options?: {
+        signal?: AbortSignal;
+        voice?: string;
+    }): Promise<void>;
+    /**
+     * Stream LLM tokens with sentence-buffered TTS + lip sync.
+     * Returns a sink: call push(token) for each token, end() when done.
+     */
+    streamText(options?: {
+        signal?: AbortSignal;
+        voice?: string;
+    }): Promise<{
+        push: (token: string) => void;
+        end: () => Promise<void>;
+    }>;
+    /** Stop current TTS playback. */
+    stopSpeaking(): void;
+    /** Disconnect speaker and dispose its resources. */
+    disconnectSpeaker(): Promise<void>;
+    /** @deprecated Use connectSpeaker(). Will be removed in v1.0. */
+    connectTTS(tts: TTSBackend, config?: TTSSpeakerConfig): Promise<void>;
+    /** @deprecated Use disconnectSpeaker(). Will be removed in v1.0. */
+    disconnectTTS(): Promise<void>;
+    /**
+     * Connect a speech listener for startListening() / onTranscript() support.
+     * Loads ASR + VAD models.
+     */
+    connectListener(config?: SpeechListenerConfig): Promise<void>;
+    /** Start listening for user speech. Requires connectListener() or connectVoice() first. */
+    startListening(): Promise<void>;
+    /** Stop listening. */
+    stopListening(): void;
+    /**
+     * Subscribe to transcript events. Returns an unsubscribe function.
+     * Requires connectListener() first.
+     */
+    onTranscript(callback: (result: TranscriptResult) => void): () => void;
+    /** Disconnect listener and dispose its resources. */
+    disconnectListener(): Promise<void>;
+    /**
+     * Connect voice with automatic speaker + listener + interruption wiring.
+     * Supports both local TTS (mode: 'local') and cloud TTS (mode: 'cloud').
+     * Does NOT auto-start listening — call startListening() when ready.
+     *
+     * Backward compatible: `mode` defaults to 'local' when not specified.
+     */
+    connectVoice(config: VoiceOrchestratorConfig): Promise<void>;
+    /** Disconnect voice (speaker + listener + interruption). */
+    disconnectVoice(): Promise<void>;
     /** Set blendshapes directly (alternative to connectFrameSource). */
     setFrame(blendshapes: Float32Array): void;
     /** Set emotion (string preset like 'happy' or EmotionWeights object). */
@@ -117,6 +180,8 @@ declare class OmoteAvatar {
     setState(state: ConversationalState): void;
     /** Set audio energy level (0-1, drives emphasis/gesture intensity). */
     setAudioEnergy(energy: number): void;
+    /** Update character expression profile at runtime. */
+    setProfile(profile: CharacterProfile): void;
     /**
      * Set the active camera for gaze tracking.
      * Required when using autoUpdate. Can also be passed directly to update().
@@ -130,10 +195,20 @@ declare class OmoteAvatar {
     get hasMorphTargets(): boolean;
     /** Number of successfully mapped ARKit blendshapes. */
     get mappedBlendshapeCount(): number;
+    /** Whether the avatar is currently speaking via TTS. */
+    get isSpeaking(): boolean;
+    /** Whether the avatar is currently listening for speech. */
+    get isListening(): boolean;
+    /** Current conversational state. */
+    get conversationalState(): ConversationalState;
+    /** Access the internal TTSSpeaker (null if not connected). */
+    get speaker(): TTSSpeaker | null;
+    /** Access the internal SpeechListener (null if not connected). */
+    get listener(): SpeechListener | null;
     /** Reset all state (smoothing, life layer, emotions). */
     reset(): void;
-    /** Clean up all resources: disconnect frame source, unregister render loop, dispose controller. */
-    dispose(): void;
+    /** Disconnect all voice resources, frame sources, unregister render loop, dispose controller. */
+    dispose(): Promise<void>;
     private registerAutoUpdate;
 }
 
@@ -191,24 +266,4 @@ declare class BlendshapeController {
     dispose(): void;
 }
 
-interface OmoteA2EOptions extends A2EOrchestratorConfig {
-    target: AbstractMesh;
-    scene: Scene;
-    controllerOptions?: BlendshapeControllerOptions;
-}
-/** @deprecated Use {@link OmoteAvatar} instead. OmoteA2E will be removed in v0.8.0. */
-declare class OmoteA2E {
-    private orchestrator;
-    private controller;
-    constructor(options: OmoteA2EOptions);
-    load(): Promise<void>;
-    start(): Promise<void>;
-    stop(): void;
-    update(): void;
-    dispose(): Promise<void>;
-    get isReady(): boolean;
-    get isStreaming(): boolean;
-    get backend(): string | null;
-}
-
-export { BlendshapeController, type BlendshapeControllerOptions, type FrameSource, type MorphIndexEntry, OmoteA2E, type OmoteA2EOptions, OmoteAvatar, type OmoteAvatarOptions, type SceneDiscoveryResult, discoverScene, writeBlendshapes };
+export { BlendshapeController, type BlendshapeControllerOptions, type MorphIndexEntry, OmoteAvatar, type OmoteAvatarOptions, type SceneDiscoveryResult, discoverScene, writeBlendshapes };

package/dist/index.d.ts

@@ -1,5 +1,6 @@
 import * as _omote_core from '@omote/core';
-import { FaceCompositorConfig, CharacterControllerConfig, EmotionWeights, ConversationalState, A2EOrchestratorConfig } from '@omote/core';
+import { FaceCompositorConfig, CharacterControllerConfig, FrameSource, TTSBackend, TTSSpeakerConfig, SpeechListenerConfig, TranscriptResult, VoiceOrchestratorConfig, EmotionWeights, ConversationalState, CharacterProfile, TTSSpeaker, SpeechListener } from '@omote/core';
+export { FrameSource, TTSSpeakerConfig as TTSConfig } from '@omote/core';
 import { AbstractMesh, TransformNode, Scene, Camera } from '@babylonjs/core';
 
 /**
@@ -47,13 +48,6 @@ interface SceneDiscoveryResult {
  */
 declare function discoverScene(root: AbstractMesh): SceneDiscoveryResult;
 
-/** Generic frame source -- any object that emits 'frame' events */
-interface FrameSource {
-    on(event: 'frame', callback: (frame: {
-        blendshapes: Float32Array;
-    }) => void): void;
-    off?(event: 'frame', callback: (...args: any[]) => void): void;
-}
 interface OmoteAvatarOptions {
     /** Root mesh of the avatar (typically loaded via SceneLoader) */
     target: AbstractMesh;
@@ -82,6 +76,9 @@ declare class OmoteAvatar {
     private _camera;
     private frameSourceCallback;
     private connectedSource;
+    private ttsSpeaker;
+    private speechListener;
+    private voiceOrchestrator;
     private renderCallback;
     private lastTime;
     constructor(options: OmoteAvatarOptions);
@@ -107,6 +104,72 @@ declare class OmoteAvatar {
     connectFrameSource(source: FrameSource): void;
     /** Disconnect the current frame source (if any). */
     disconnectFrameSource(): void;
+    /**
+     * Connect a TTS backend for speak() / streamText() support.
+     * Loads LAM model and creates internal PlaybackPipeline.
+     *
+     * @param tts - TTS backend (e.g., KokoroTTSInference, ElevenLabs adapter)
+     * @param config - A2E, expression profile, and playback configuration
+     */
+    /** Warm up AudioContext for iOS/Safari autoplay policy. Call from user gesture. */
+    warmup(): Promise<void>;
+    connectSpeaker(tts: TTSBackend, config?: TTSSpeakerConfig): Promise<void>;
+    /**
+     * Synthesize text and play with lip sync.
+     * Auto-aborts previous speak if still in progress.
+     *
+     * @param text - Text to synthesize
+     * @param options - Optional voice override and abort signal
+     */
+    speak(text: string, options?: {
+        signal?: AbortSignal;
+        voice?: string;
+    }): Promise<void>;
+    /**
+     * Stream LLM tokens with sentence-buffered TTS + lip sync.
+     * Returns a sink: call push(token) for each token, end() when done.
+     */
+    streamText(options?: {
+        signal?: AbortSignal;
+        voice?: string;
+    }): Promise<{
+        push: (token: string) => void;
+        end: () => Promise<void>;
+    }>;
+    /** Stop current TTS playback. */
+    stopSpeaking(): void;
+    /** Disconnect speaker and dispose its resources. */
+    disconnectSpeaker(): Promise<void>;
+    /** @deprecated Use connectSpeaker(). Will be removed in v1.0. */
+    connectTTS(tts: TTSBackend, config?: TTSSpeakerConfig): Promise<void>;
+    /** @deprecated Use disconnectSpeaker(). Will be removed in v1.0. */
+    disconnectTTS(): Promise<void>;
+    /**
+     * Connect a speech listener for startListening() / onTranscript() support.
+     * Loads ASR + VAD models.
+     */
+    connectListener(config?: SpeechListenerConfig): Promise<void>;
+    /** Start listening for user speech. Requires connectListener() or connectVoice() first. */
+    startListening(): Promise<void>;
+    /** Stop listening. */
+    stopListening(): void;
+    /**
+     * Subscribe to transcript events. Returns an unsubscribe function.
+     * Requires connectListener() first.
+     */
+    onTranscript(callback: (result: TranscriptResult) => void): () => void;
+    /** Disconnect listener and dispose its resources. */
+    disconnectListener(): Promise<void>;
+    /**
+     * Connect voice with automatic speaker + listener + interruption wiring.
+     * Supports both local TTS (mode: 'local') and cloud TTS (mode: 'cloud').
+     * Does NOT auto-start listening — call startListening() when ready.
+     *
+     * Backward compatible: `mode` defaults to 'local' when not specified.
+     */
+    connectVoice(config: VoiceOrchestratorConfig): Promise<void>;
+    /** Disconnect voice (speaker + listener + interruption). */
+    disconnectVoice(): Promise<void>;
     /** Set blendshapes directly (alternative to connectFrameSource). */
     setFrame(blendshapes: Float32Array): void;
     /** Set emotion (string preset like 'happy' or EmotionWeights object). */
@@ -117,6 +180,8 @@ declare class OmoteAvatar {
     setState(state: ConversationalState): void;
     /** Set audio energy level (0-1, drives emphasis/gesture intensity). */
     setAudioEnergy(energy: number): void;
+    /** Update character expression profile at runtime. */
+    setProfile(profile: CharacterProfile): void;
     /**
      * Set the active camera for gaze tracking.
      * Required when using autoUpdate. Can also be passed directly to update().
@@ -130,10 +195,20 @@ declare class OmoteAvatar {
     get hasMorphTargets(): boolean;
     /** Number of successfully mapped ARKit blendshapes. */
     get mappedBlendshapeCount(): number;
+    /** Whether the avatar is currently speaking via TTS. */
+    get isSpeaking(): boolean;
+    /** Whether the avatar is currently listening for speech. */
+    get isListening(): boolean;
+    /** Current conversational state. */
+    get conversationalState(): ConversationalState;
+    /** Access the internal TTSSpeaker (null if not connected). */
+    get speaker(): TTSSpeaker | null;
+    /** Access the internal SpeechListener (null if not connected). */
+    get listener(): SpeechListener | null;
     /** Reset all state (smoothing, life layer, emotions). */
     reset(): void;
-    /** Clean up all resources: disconnect frame source, unregister render loop, dispose controller. */
-    dispose(): void;
+    /** Disconnect all voice resources, frame sources, unregister render loop, dispose controller. */
+    dispose(): Promise<void>;
     private registerAutoUpdate;
 }
 
@@ -191,24 +266,4 @@ declare class BlendshapeController {
     dispose(): void;
 }
 
-interface OmoteA2EOptions extends A2EOrchestratorConfig {
-    target: AbstractMesh;
-    scene: Scene;
-    controllerOptions?: BlendshapeControllerOptions;
-}
-/** @deprecated Use {@link OmoteAvatar} instead. OmoteA2E will be removed in v0.8.0. */
-declare class OmoteA2E {
-    private orchestrator;
-    private controller;
-    constructor(options: OmoteA2EOptions);
-    load(): Promise<void>;
-    start(): Promise<void>;
-    stop(): void;
-    update(): void;
-    dispose(): Promise<void>;
-    get isReady(): boolean;
-    get isStreaming(): boolean;
-    get backend(): string | null;
-}
-
-export { BlendshapeController, type BlendshapeControllerOptions, type FrameSource, type MorphIndexEntry, OmoteA2E, type OmoteA2EOptions, OmoteAvatar, type OmoteAvatarOptions, type SceneDiscoveryResult, discoverScene, writeBlendshapes };
+export { BlendshapeController, type BlendshapeControllerOptions, type MorphIndexEntry, OmoteAvatar, type OmoteAvatarOptions, type SceneDiscoveryResult, discoverScene, writeBlendshapes };

package/dist/index.js

@@ -1,5 +1,12 @@
 // src/OmoteAvatar.ts
-import { CharacterController, createLogger as createLogger2 } from "@omote/core";
+import {
+  CharacterController,
+  TTSSpeaker,
+  SpeechListener,
+  VoiceOrchestrator,
+  createLogger as createLogger2,
+  getClock
+} from "@omote/core";
 
 // src/SceneDiscovery.ts
 import { LAM_BLENDSHAPES, createLogger } from "@omote/core";
@@ -116,7 +123,7 @@ function writeBlendshapes(blendshapes, morphEntries) {
 }
 
 // src/OmoteAvatar.ts
-var logger2 = createLogger2("OmoteAvatar");
+var logger2 = createLogger2("OmoteAvatar.Babylon");
 var OmoteAvatar = class {
   constructor(options) {
     // State
@@ -129,6 +136,12 @@ var OmoteAvatar = class {
     // Frame source connection
     this.frameSourceCallback = null;
     this.connectedSource = null;
+    // TTS integration
+    this.ttsSpeaker = null;
+    // Speech listener
+    this.speechListener = null;
+    // Voice orchestrator
+    this.voiceOrchestrator = null;
     // Auto-update
     this.renderCallback = null;
     this.lastTime = 0;
@@ -207,9 +220,15 @@ var OmoteAvatar = class {
    * Only one source can be connected at a time; calling again disconnects the previous.
    */
   connectFrameSource(source) {
+    if (this.ttsSpeaker && source !== this.ttsSpeaker.frameSource) {
+      this.ttsSpeaker.stop();
+    }
     this.disconnectFrameSource();
     this.frameSourceCallback = (frame) => {
       this.currentBlendshapes = frame.blendshapes;
+      if (frame.emotion !== void 0) {
+        this._emotion = frame.emotion;
+      }
     };
     source.on("frame", this.frameSourceCallback);
     this.connectedSource = source;
@@ -225,6 +244,186 @@ var OmoteAvatar = class {
     this.frameSourceCallback = null;
   }
   // ---------------------------------------------------------------------------
+  // Speaker (TTS → lip sync)
+  // ---------------------------------------------------------------------------
+  /**
+   * Connect a TTS backend for speak() / streamText() support.
+   * Loads LAM model and creates internal PlaybackPipeline.
+   *
+   * @param tts - TTS backend (e.g., KokoroTTSInference, ElevenLabs adapter)
+   * @param config - A2E, expression profile, and playback configuration
+   */
+  /** Warm up AudioContext for iOS/Safari autoplay policy. Call from user gesture. */
+  async warmup() {
+    if (this.ttsSpeaker) await this.ttsSpeaker.warmup();
+  }
+  async connectSpeaker(tts, config) {
+    await this.disconnectSpeaker();
+    this.ttsSpeaker = new TTSSpeaker();
+    await this.ttsSpeaker.connect(tts, config);
+    this.connectFrameSource(this.ttsSpeaker.frameSource);
+  }
+  /**
+   * Synthesize text and play with lip sync.
+   * Auto-aborts previous speak if still in progress.
+   *
+   * @param text - Text to synthesize
+   * @param options - Optional voice override and abort signal
+   */
+  async speak(text, options) {
+    if (this.voiceOrchestrator) {
+      await this.voiceOrchestrator.speak(text, options);
+      return;
+    }
+    if (!this.ttsSpeaker) {
+      throw new Error("No speaker connected. Call connectSpeaker() first.");
+    }
+    this._isSpeaking = true;
+    this._state = "speaking";
+    try {
+      await this.ttsSpeaker.speak(text, options);
+    } finally {
+      this._isSpeaking = false;
+      if (this._state === "speaking") {
+        this._state = "idle";
+      }
+    }
+  }
+  /**
+   * Stream LLM tokens with sentence-buffered TTS + lip sync.
+   * Returns a sink: call push(token) for each token, end() when done.
+   */
+  async streamText(options) {
+    if (this.voiceOrchestrator) {
+      return this.voiceOrchestrator.streamText(options);
+    }
+    if (!this.ttsSpeaker) {
+      throw new Error("No speaker connected. Call connectSpeaker() first.");
+    }
+    this._isSpeaking = true;
+    this._state = "speaking";
+    const stream = await this.ttsSpeaker.streamText(options ?? {});
+    return {
+      push: stream.push,
+      end: async () => {
+        try {
+          await stream.end();
+        } finally {
+          this._isSpeaking = false;
+          if (this._state === "speaking") this._state = "idle";
+        }
+      }
+    };
+  }
+  /** Stop current TTS playback. */
+  stopSpeaking() {
+    if (this.voiceOrchestrator) {
+      this.voiceOrchestrator.stopSpeaking();
+      return;
+    }
+    this.ttsSpeaker?.stop();
+  }
+  /** Disconnect speaker and dispose its resources. */
+  async disconnectSpeaker() {
+    if (this.ttsSpeaker) {
+      this.disconnectFrameSource();
+      await this.ttsSpeaker.dispose();
+      this.ttsSpeaker = null;
+    }
+  }
+  /** @deprecated Use connectSpeaker(). Will be removed in v1.0. */
+  async connectTTS(tts, config) {
+    return this.connectSpeaker(tts, config);
+  }
+  /** @deprecated Use disconnectSpeaker(). Will be removed in v1.0. */
+  async disconnectTTS() {
+    return this.disconnectSpeaker();
+  }
+  // ---------------------------------------------------------------------------
+  // Listener (mic → VAD → ASR → transcript)
+  // ---------------------------------------------------------------------------
+  /**
+   * Connect a speech listener for startListening() / onTranscript() support.
+   * Loads ASR + VAD models.
+   */
+  async connectListener(config) {
+    await this.disconnectListener();
+    this.speechListener = new SpeechListener(config);
+    await this.speechListener.loadModels();
+  }
+  /** Start listening for user speech. Requires connectListener() or connectVoice() first. */
+  async startListening() {
+    if (this.voiceOrchestrator) {
+      await this.voiceOrchestrator.startListening();
+      return;
+    }
+    if (!this.speechListener) {
+      throw new Error("No listener connected. Call connectListener() first.");
+    }
+    this._state = "listening";
+    await this.speechListener.start();
+  }
+  /** Stop listening. */
+  stopListening() {
+    if (this.voiceOrchestrator) {
+      this.voiceOrchestrator.stopListening();
+      return;
+    }
+    this.speechListener?.stop();
+    if (this._state === "listening") this._state = "idle";
+  }
+  /**
+   * Subscribe to transcript events. Returns an unsubscribe function.
+   * Requires connectListener() first.
+   */
+  onTranscript(callback) {
+    const listener = this.speechListener ?? this.voiceOrchestrator?.listener;
+    if (!listener) {
+      throw new Error("No listener connected. Call connectListener() or connectVoice() first.");
+    }
+    listener.on("transcript", callback);
+    return () => {
+      listener.off?.("transcript", callback);
+    };
+  }
+  /** Disconnect listener and dispose its resources. */
+  async disconnectListener() {
+    if (this.speechListener) {
+      await this.speechListener.dispose();
+      this.speechListener = null;
+    }
+  }
+  // ---------------------------------------------------------------------------
+  // Voice (combined speaker + listener + interruption)
+  // ---------------------------------------------------------------------------
+  /**
+   * Connect voice with automatic speaker + listener + interruption wiring.
+   * Supports both local TTS (mode: 'local') and cloud TTS (mode: 'cloud').
+   * Does NOT auto-start listening — call startListening() when ready.
+   *
+   * Backward compatible: `mode` defaults to 'local' when not specified.
+   */
+  async connectVoice(config) {
+    await this.disconnectVoice();
+    this.voiceOrchestrator = new VoiceOrchestrator();
+    await this.voiceOrchestrator.connect(config);
+    if (this.voiceOrchestrator.frameSource) {
+      this.connectFrameSource(this.voiceOrchestrator.frameSource);
+    }
+    this.voiceOrchestrator.on("state", (state) => {
+      this._state = state;
+      this._isSpeaking = state === "speaking";
+    });
+  }
+  /** Disconnect voice (speaker + listener + interruption). */
+  async disconnectVoice() {
+    if (this.voiceOrchestrator) {
+      this.disconnectFrameSource();
+      await this.voiceOrchestrator.disconnect();
+      this.voiceOrchestrator = null;
+    }
+  }
+  // ---------------------------------------------------------------------------
   // State setters
   // ---------------------------------------------------------------------------
   /** Set blendshapes directly (alternative to connectFrameSource). */
@@ -247,6 +446,10 @@ var OmoteAvatar = class {
   setAudioEnergy(energy) {
     this._audioEnergy = energy;
   }
+  /** Update character expression profile at runtime. */
+  setProfile(profile) {
+    this.controller.setProfile(profile);
+  }
   /**
    * Set the active camera for gaze tracking.
    * Required when using autoUpdate. Can also be passed directly to update().
@@ -273,6 +476,26 @@ var OmoteAvatar = class {
   get mappedBlendshapeCount() {
     return this.discovery.mappedBlendshapeCount;
   }
+  /** Whether the avatar is currently speaking via TTS. */
+  get isSpeaking() {
+    return this._isSpeaking;
+  }
+  /** Whether the avatar is currently listening for speech. */
+  get isListening() {
+    return this._state === "listening";
+  }
+  /** Current conversational state. */
+  get conversationalState() {
+    return this._state;
+  }
+  /** Access the internal TTSSpeaker (null if not connected). */
+  get speaker() {
+    return this.ttsSpeaker ?? this.voiceOrchestrator?.speaker ?? null;
+  }
+  /** Access the internal SpeechListener (null if not connected). */
+  get listener() {
+    return this.speechListener ?? this.voiceOrchestrator?.listener ?? null;
+  }
   // ---------------------------------------------------------------------------
   // Lifecycle
   // ---------------------------------------------------------------------------
@@ -285,8 +508,11 @@ var OmoteAvatar = class {
     this._audioEnergy = 0;
     this.controller.reset();
   }
-  /** Clean up all resources: disconnect frame source, unregister render loop, dispose controller. */
-  dispose() {
+  /** Disconnect all voice resources, frame sources, unregister render loop, dispose controller. */
+  async dispose() {
+    await this.disconnectVoice();
+    await this.disconnectSpeaker();
+    await this.disconnectListener();
     this.disconnectFrameSource();
     if (this.renderCallback) {
       this.scene.unregisterBeforeRender(this.renderCallback);
@@ -299,9 +525,9 @@ var OmoteAvatar = class {
   // Internal
   // ---------------------------------------------------------------------------
   registerAutoUpdate() {
-    this.lastTime = performance.now();
+    this.lastTime = getClock().now();
     this.renderCallback = () => {
-      const now = performance.now();
+      const now = getClock().now();
       const delta = (now - this.lastTime) / 1e3;
       this.lastTime = now;
       if (this._camera) {
@@ -419,45 +645,8 @@ var BlendshapeController = class {
     this.scene = null;
   }
 };
-
-// src/OmoteA2E.ts
-import { A2EOrchestrator } from "@omote/core";
-var OmoteA2E = class {
-  constructor(options) {
-    const { target, scene, controllerOptions, ...orchestratorConfig } = options;
-    this.controller = new BlendshapeController(target, scene, controllerOptions);
-    this.orchestrator = new A2EOrchestrator(orchestratorConfig);
-  }
-  async load() {
-    return this.orchestrator.load();
-  }
-  async start() {
-    return this.orchestrator.start();
-  }
-  stop() {
-    this.orchestrator.stop();
-  }
-  update() {
-    const w = this.orchestrator.latestWeights;
-    if (w) this.controller.update(w);
-  }
-  async dispose() {
-    await this.orchestrator.dispose();
-    this.controller.dispose();
-  }
-  get isReady() {
-    return this.orchestrator.isReady;
-  }
-  get isStreaming() {
-    return this.orchestrator.isStreaming;
-  }
-  get backend() {
-    return this.orchestrator.backend;
-  }
-};
 export {
   BlendshapeController,
-  OmoteA2E,
   OmoteAvatar,
   discoverScene,
   writeBlendshapes