Skip to content
Learni
View all tutorials
Intelligence Artificielle

How to Master the ElevenLabs API with TypeScript in 2026

15 minEXPERT
TypeScriptAPIElevenLabsIA Vocale
Lire en français

Introduction

ElevenLabs offers one of the highest-performing text-to-speech APIs on the market. In 2026, expert integrations require fine-grained management of audio streams, rate limits, and instant voice cloning. This tutorial guides you step by step through building a robust, production-ready TypeScript layer.

Prerequisites

  • Node.js 20+
  • ElevenLabs account with Pro API key
  • Advanced TypeScript knowledge and stream management
  • Axios or native fetch

Project Initialization

terminal
npm init -y
npm install axios dotenv
npm install --save-dev @types/node typescript
npx tsc --init

Initializes a TypeScript project and installs the necessary dependencies to consume the ElevenLabs API in a typed manner.

API Key Configuration

src/config/elevenlabs.ts
import dotenv from 'dotenv';
dotenv.config();

export const ELEVENLABS_API_KEY = process.env.ELEVENLABS_API_KEY!;
export const BASE_URL = 'https://api.elevenlabs.io/v1';

Loads the API key from environment variables. Always store secrets outside the source code.

Typed API Client

src/lib/elevenlabsClient.ts
import axios from 'axios';
import { ELEVENLABS_API_KEY, BASE_URL } from '../config/elevenlabs';

const client = axios.create({
  baseURL: BASE_URL,
  headers: {
    'xi-api-key': ELEVENLABS_API_KEY,
    'Content-Type': 'application/json',
  },
});
export default client;

Creates a reusable and typed Axios client for all requests to ElevenLabs.

Basic TTS Generation

src/services/ttsService.ts
import client from '../lib/elevenlabsClient';

export async function generateSpeech(text: string, voiceId: string): Promise<Buffer> {
  const response = await client.post(
    `/text-to-speech/${voiceId}`,
    { text, model_id: 'eleven_multilingual_v2' },
    { responseType: 'arraybuffer' }
  );
  return Buffer.from(response.data);
}

Complete function that returns an MP3 audio Buffer. Uses the multilingual v2 model for optimal quality.

Real-Time Audio Streaming

src/services/streamingService.ts
import client from '../lib/elevenlabsClient';

export async function* streamSpeech(text: string, voiceId: string) {
  const response = await client.post(
    `/text-to-speech/${voiceId}/stream`,
    { text, model_id: 'eleven_multilingual_v2' },
    { responseType: 'stream' }
  );
  for await (const chunk of response.data) {
    yield chunk;
  }
}

Implements chunk-by-chunk streaming to reduce perceived latency in real-time applications.

Best Practices

  • Always validate character limits before making calls
  • Implement a retry system with exponential backoff
  • Store voice_ids in the database rather than in code
  • Use webhooks for long-running tasks
  • Monitor credit consumption via the /user endpoint

Common Errors to Avoid

  • Forgetting to handle 429 status code (rate limit) from ElevenLabs
  • Not properly closing audio streams during streaming
  • Using cloned voices without explicit consent
  • Ignoring latency differences between regions

To Go Further

Discover our advanced training on voice AI integration: https://learni-group.com/formations