@elevenlabs/client 0.12.1 → 0.13.0

package/dist/scribe/scribe.d.ts

@@ -73,7 +73,7 @@ export interface AudioOptions extends BaseOptions {
  */
 export interface MicrophoneOptions extends BaseOptions {
     microphone?: {
-        deviceId?: string;
+        deviceId?: ConstrainDOMString;
         echoCancellation?: boolean;
         noiseSuppression?: boolean;
         autoGainControl?: boolean;

package/dist/utils/BaseConnection.d.ts

@@ -29,6 +29,9 @@ export type BaseSessionConfig = {
         };
         tts?: {
             voiceId?: string;
+            speed?: number;
+            stability?: number;
+            similarityBoost?: number;
         };
         conversation?: {
             textOnly?: boolean;

package/dist/utils/errors.d.ts

@@ -0,0 +1,8 @@
+export declare class SessionConnectionError extends Error {
+    readonly closeCode?: number;
+    readonly closeReason?: string;
+    constructor(message: string, options?: {
+        closeCode?: number;
+        closeReason?: string;
+    });
+}

package/dist/utils/events.d.ts

@@ -1,4 +1,5 @@
 import { Outgoing } from "@elevenlabs/types";
+import type { AudioAlignmentEvent } from "@elevenlabs/types";
 import { AgentChatResponsePartClientEvent, AgentResponse, AgentResponseCorrection, AgentToolResponseClientEvent, AsrInitiationMetadataEvent as AsrMetadataEvent, Audio, AgentToolRequestClientEvent, ClientToolCallMessage, ConversationMetadata, ErrorMessage, Interruption, McpConnectionStatusClientEvent, McpToolCall, Ping, InternalTentativeAgentResponse as TentativeAgentResponseInternal, UserTranscript, VadScore } from "@elevenlabs/types/generated/types/asyncapi-types";
 export type UserTranscriptionEvent = UserTranscript;
 export type AgentResponseEvent = AgentResponse;
@@ -18,6 +19,7 @@ export type AsrInitiationMetadataEvent = AsrMetadataEvent;
 export type MCPConnectionStatusEvent = McpConnectionStatusClientEvent;
 export type AgentChatResponsePartEvent = AgentChatResponsePartClientEvent;
 export type ErrorMessageEvent = ErrorMessage;
+export type { AudioAlignmentEvent };
 export type IncomingSocketEvent = UserTranscriptionEvent | AgentResponseEvent | AgentResponseCorrectionEvent | AgentAudioEvent | InterruptionEvent | InternalTentativeAgentResponseEvent | ConfigEvent | PingEvent | ClientToolCallEvent | VadScoreEvent | MCPToolCallClientEvent | AgentToolRequestEvent | AgentToolResponseEvent | ConversationMetadataEvent | AsrInitiationMetadataEvent | MCPConnectionStatusEvent | AgentChatResponsePartEvent | ErrorMessageEvent;
 export type PongEvent = Outgoing.PongClientToOrchestratorEvent;
 export type UserAudioEvent = Outgoing.UserAudio;

package/dist/version.d.ts

@@ -1 +1 @@
-export declare const PACKAGE_VERSION = "0.12.1";
+export declare const PACKAGE_VERSION = "0.13.0";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@elevenlabs/client",
-  "version": "0.12.1",
+  "version": "0.13.0",
   "description": "ElevenLabs JavaScript Client Library",
   "main": "./dist/lib.umd.js",
   "module": "./dist/lib.module.js",

package/worklets/scribeAudioProcessor.js

@@ -1,5 +1,7 @@
 /*
  * Scribe Audio Processor for converting microphone audio to PCM16 format
+ * Supports resampling for browsers like Firefox that don't support
+ * AudioContext sample rate constraints.
  * USED BY @elevenlabs/client
  */
 
@@ -8,15 +10,67 @@ class ScribeAudioProcessor extends AudioWorkletProcessor {
     super();
     this.buffer = [];
     this.bufferSize = 4096; // Buffer size for optimal chunk transmission
+
+    // Resampling state
+    this.inputSampleRate = null;
+    this.outputSampleRate = null;
+    this.resampleRatio = 1;
+    this.lastSample = 0;
+    this.resampleAccumulator = 0;
+
+    this.port.onmessage = ({ data }) => {
+      if (data.type === "configure") {
+        this.inputSampleRate = data.inputSampleRate;
+        this.outputSampleRate = data.outputSampleRate;
+        if (this.inputSampleRate && this.outputSampleRate) {
+          this.resampleRatio = this.inputSampleRate / this.outputSampleRate;
+        }
+      }
+    };
+  }
+
+  // Linear interpolation resampling
+  resample(inputData) {
+    if (this.resampleRatio === 1 || !this.inputSampleRate) {
+      return inputData;
+    }
+
+    const outputSamples = [];
+
+    for (let i = 0; i < inputData.length; i++) {
+      const currentSample = inputData[i];
+
+      // Generate output samples using linear interpolation
+      while (this.resampleAccumulator < 1) {
+        const interpolated =
+          this.lastSample +
+          (currentSample - this.lastSample) * this.resampleAccumulator;
+        outputSamples.push(interpolated);
+        this.resampleAccumulator += this.resampleRatio;
+      }
+
+      this.resampleAccumulator -= 1;
+      this.lastSample = currentSample;
+    }
+
+    return new Float32Array(outputSamples);
   }
 
   process(inputs) {
     const input = inputs[0];
     if (input.length > 0) {
-      const channelData = input[0]; // Get first channel (mono)
+      let channelData = input[0]; // Get first channel (mono)
+
+      // Resample if needed (for Firefox and other browsers that don't
+      // support AudioContext sample rate constraints)
+      if (this.resampleRatio !== 1) {
+        channelData = this.resample(channelData);
+      }
 
       // Add incoming audio to buffer
-      this.buffer.push(...channelData);
+      for (let i = 0; i < channelData.length; i++) {
+        this.buffer.push(channelData[i]);
+      }
 
       // When buffer reaches threshold, convert and send
       if (this.buffer.length >= this.bufferSize) {