@elevenlabs/client 1.8.0 → 1.8.1

package/dist/lib.iife.js

@@ -92,7 +92,7 @@ var ElevenLabsClient = (function(exports) {
 	//#region src/sourceInfo.ts
 	let sourceInfo = Object.freeze({
 		name: "js_sdk",
-		version: "1.8.0"
+		version: "1.8.1"
 	});
 	//#endregion
 	//#region src/utils/events.ts
@@ -21955,14 +21955,39 @@ registerProcessor("audioConcatProcessor", AudioConcatProcessor);
 		};
 	}
 	//#endregion
+	//#region src/platform/web/compatibility.ts
+	function isIosDevice() {
+		return [
+			"iPad Simulator",
+			"iPhone Simulator",
+			"iPod Simulator",
+			"iPad",
+			"iPhone",
+			"iPod"
+		].includes(navigator.platform) || navigator.userAgent.includes("Mac") && "ontouchend" in document;
+	}
+	function isAndroidDevice() {
+		return /android/i.test(navigator.userAgent);
+	}
+	//#endregion
 	//#region src/platform/web/output.ts
+	function maybePrimeIosPlayback({ sampleRate, format, worklet, audioElement }) {
+		if (!isIosDevice()) return;
+		const primeFrameCount = Math.floor(sampleRate * 100 / 1e3);
+		const silentBuffer = format === "ulaw" ? new Uint8Array(primeFrameCount).fill(255) : new Int16Array(primeFrameCount);
+		worklet.port.postMessage({
+			type: "buffer",
+			buffer: silentBuffer.buffer
+		});
+		audioElement.play().catch(() => {});
+	}
 	var MediaDeviceOutput = class MediaDeviceOutput {
-		static async create({ sampleRate, format, outputDeviceId, workletPaths, libsampleratePath }) {
-			let context = null;
+		static async create({ sampleRate, format, outputDeviceId, workletPaths, libsampleratePath, audioContext }) {
+			let context = audioContext ?? null;
 			let audioElement = null;
 			try {
 				const supportsSampleRateConstraint = navigator.mediaDevices.getSupportedConstraints().sampleRate;
-				context = new AudioContext(supportsSampleRateConstraint ? { sampleRate } : {});
+				if (!context) context = new AudioContext(supportsSampleRateConstraint ? { sampleRate } : {});
 				const analyser = context.createAnalyser();
 				const gain = context.createGain();
 				audioElement = new Audio();
@@ -21988,6 +22013,12 @@ registerProcessor("audioConcatProcessor", AudioConcatProcessor);
 				});
 				worklet.connect(gain);
 				await context.resume();
+				maybePrimeIosPlayback({
+					sampleRate,
+					format,
+					worklet,
+					audioElement
+				});
 				if (outputDeviceId && audioElement.setSinkId) await audioElement.setSinkId(outputDeviceId);
 				return new MediaDeviceOutput(context, analyser, gain, worklet, audioElement);
 			} catch (error) {
@@ -22199,21 +22230,6 @@ class RawAudioProcessor extends AudioWorkletProcessor {
 }
 registerProcessor("rawAudioProcessor", RawAudioProcessor);
 `);
-	//#endregion
-	//#region src/platform/web/compatibility.ts
-	function isIosDevice() {
-		return [
-			"iPad Simulator",
-			"iPhone Simulator",
-			"iPod Simulator",
-			"iPad",
-			"iPhone",
-			"iPod"
-		].includes(navigator.platform) || navigator.userAgent.includes("Mac") && "ontouchend" in document;
-	}
-	function isAndroidDevice() {
-		return /android/i.test(navigator.userAgent);
-	}
 	//#endregion
 	//#region src/platform/web/input.ts
 	const defaultConstraints = {
@@ -22372,6 +22388,7 @@ registerProcessor("rawAudioProcessor", RawAudioProcessor);
 		inputFormat;
 		outputFormat;
 		outputListeners = /* @__PURE__ */ new Set();
+		pendingAudioEvents = [];
 		constructor(socket, conversationId, inputFormat, outputFormat) {
 			super();
 			this.socket = socket;
@@ -22472,13 +22489,19 @@ registerProcessor("rawAudioProcessor", RawAudioProcessor);
 			}
 		}
 		close() {
+			this.pendingAudioEvents = [];
 			this.socket.close(1e3, "User ended conversation");
 		}
 		sendMessage(message) {
 			this.socket.send(JSON.stringify(message));
 		}
 		addListener(listener) {
+			const hadListeners = this.outputListeners.size > 0;
 			this.outputListeners.add(listener);
+			if (hadListeners || this.pendingAudioEvents.length === 0) return;
+			const pending = this.pendingAudioEvents;
+			this.pendingAudioEvents = [];
+			for (const event of pending) listener(event);
 		}
 		removeListener(listener) {
 			this.outputListeners.delete(listener);
@@ -22487,6 +22510,10 @@ registerProcessor("rawAudioProcessor", RawAudioProcessor);
 			super.handleMessage(parsedEvent);
 			if (parsedEvent.type === "audio" && parsedEvent.audio_event.audio_base_64) {
 				const audioEvent = { audio_base_64: parsedEvent.audio_event.audio_base_64 };
+				if (this.outputListeners.size === 0) {
+					this.pendingAudioEvents.push(audioEvent);
+					return;
+				}
 				this.outputListeners.forEach((listener) => listener(audioEvent));
 			}
 		}
@@ -22553,6 +22580,70 @@ registerProcessor("rawAudioProcessor", RawAudioProcessor);
 		if (delayMs > 0) await new Promise((resolve) => setTimeout(resolve, delayMs));
 	}
 	//#endregion
+	//#region src/platform/web/audioUnlock.ts
+	const STASH_TTL_MS = 3e4;
+	const UNLOCK_EVENTS = [
+		"touchstart",
+		"touchend",
+		"click"
+	];
+	let stashedAudioContext = null;
+	let discardTimer = null;
+	let unlockListenerInstalled = false;
+	function unlockAudioContext(ctx) {
+		const buffer = ctx.createBuffer(1, 1, 22050);
+		const source = ctx.createBufferSource();
+		source.buffer = buffer;
+		source.connect(ctx.destination);
+		source.start(0);
+		ctx.resume().catch(() => {});
+	}
+	function scheduleStashDiscard() {
+		if (discardTimer) clearTimeout(discardTimer);
+		discardTimer = setTimeout(() => {
+			discardTimer = null;
+			discardStashedAudioContext();
+		}, STASH_TTL_MS);
+	}
+	function clearStashDiscardTimer() {
+		if (!discardTimer) return;
+		clearTimeout(discardTimer);
+		discardTimer = null;
+	}
+	/** Unlock iOS audio during a user gesture. No-op on non-iOS. */
+	function unlockIosAudioForSession() {
+		if (!isIosDevice() || stashedAudioContext) return;
+		const ctx = new AudioContext();
+		unlockAudioContext(ctx);
+		stashedAudioContext = ctx;
+		scheduleStashDiscard();
+	}
+	/**
+	* Listen for the first user interaction on the page and unlock iOS audio in
+	* the capture phase, before application click handlers run. Covers callers
+	* (such as the convai widget) that `await` before `Conversation.startSession`.
+	*/
+	function installIosAudioUnlockListener() {
+		if (!isIosDevice() || unlockListenerInstalled || typeof document === "undefined") return;
+		unlockListenerInstalled = true;
+		const onUserGesture = () => {
+			unlockIosAudioForSession();
+		};
+		for (const event of UNLOCK_EVENTS) document.addEventListener(event, onUserGesture, true);
+	}
+	function takeUnlockedAudioContext() {
+		const ctx = stashedAudioContext;
+		stashedAudioContext = null;
+		clearStashDiscardTimer();
+		return ctx;
+	}
+	function discardStashedAudioContext() {
+		if (!stashedAudioContext) return;
+		stashedAudioContext.close().catch(() => {});
+		stashedAudioContext = null;
+		clearStashDiscardTimer();
+	}
+	//#endregion
 	//#region src/platform/web/VoiceSessionSetup.ts
 	function detectPlatform() {
 		if (isAndroidDevice()) return "android";
@@ -22569,7 +22660,7 @@ registerProcessor("rawAudioProcessor", RawAudioProcessor);
 	* Sets up WebSocket-specific input and output controllers using
 	* web MediaDevice APIs (AudioContext, AudioWorklet, etc.).
 	*/
-	async function setupWebSocketIO(options, connection) {
+	async function setupWebSocketIO(options, connection, audioContext) {
 		const [input, output] = await Promise.all([MediaDeviceInput.create({
 			...connection.inputFormat,
 			preferHeadphonesForIosDevices: options.preferHeadphonesForIosDevices,
@@ -22579,7 +22670,8 @@ registerProcessor("rawAudioProcessor", RawAudioProcessor);
 		}), MediaDeviceOutput.create({
 			...connection.outputFormat,
 			outputDeviceId: options.outputDeviceId,
-			workletPaths: options.workletPaths
+			workletPaths: options.workletPaths,
+			audioContext: audioContext ?? void 0
 		})]);
 		const detachInput = attachInputToConnection(input, connection);
 		const detachOutput = attachConnectionToOutput(connection, output);
@@ -22602,6 +22694,7 @@ registerProcessor("rawAudioProcessor", RawAudioProcessor);
 		const useWakeLock = options.useWakeLock ?? true;
 		let wakeLock = null;
 		let preliminaryInputStream = null;
+		let unlockedAudioContext = null;
 		try {
 			if (useWakeLock) wakeLock = await requestWakeLock();
 			preliminaryInputStream = await navigator.mediaDevices.getUserMedia({ audio: true });
@@ -22610,12 +22703,20 @@ registerProcessor("rawAudioProcessor", RawAudioProcessor);
 			const connection = await createConnection(options);
 			let result;
 			try {
-				if (connection instanceof WebSocketConnection) result = {
-					connection,
-					...await setupWebSocketIO(options, connection)
-				};
-				else result = setupWebRTCSession(connection);
+				if (connection instanceof WebSocketConnection) {
+					unlockedAudioContext = takeUnlockedAudioContext();
+					result = {
+						connection,
+						...await setupWebSocketIO(options, connection, unlockedAudioContext)
+					};
+					unlockedAudioContext = null;
+				} else {
+					discardStashedAudioContext();
+					result = setupWebRTCSession(connection);
+				}
 			} catch (ioError) {
+				await unlockedAudioContext?.close().catch(() => {});
+				unlockedAudioContext = null;
 				connection.close();
 				throw ioError;
 			}
@@ -22650,6 +22751,7 @@ registerProcessor("rawAudioProcessor", RawAudioProcessor);
 				await wakeLock?.release();
 				wakeLock = null;
 			} catch (_e) {}
+			discardStashedAudioContext();
 			throw error;
 		}
 	}
@@ -23993,6 +24095,7 @@ registerProcessor("scribeAudioProcessor", ScribeAudioProcessor);
 	} };
 	//#endregion
 	//#region src/platform/web/index.ts
+	installIosAudioUnlockListener();
 	setWebRTCAudioAdapterFactory(() => new WebAudioAdapter());
 	setScribeMicrophoneSetup(webScribeMicrophoneSetup);
 	//#endregion