Skip to content
Learni
View all tutorials
Intelligence Artificielle

How to Create AI Voices with Resemble AI in 2026

14 minBEGINNER
Intelligence ArtificielleAPIVoix
Lire en français

Introduction

Resemble AI lets you create realistic synthetic voices from just a few minutes of recordings. In 2026, the tool has become essential for podcasts, training materials, and voice assistants. This tutorial guides you from account creation to API integration in a Node.js application. You'll learn how to clone a voice, convert text to speech, and handle common errors. No advanced AI knowledge is required.

Prerequisites

  • Node.js 20 or higher
  • Resemble AI account (free to start)
  • Basic JavaScript/TypeScript knowledge
  • A microphone to record 5 minutes of voice

SDK Installation

terminal
npm init -y
npm install @resemble/node dotenv

We initialize the project and install the official Resemble SDK along with dotenv to securely manage API keys.

Environment Variables Configuration

.env
RESEMBLE_API_KEY=votre_cle_api_ici
RESEMBLE_PROJECT_UUID=votre_uuid_projet

Create a .env file at the root. Never commit this file. The key is obtained from the Resemble dashboard.

Initialize Resemble Client

resemble-client.ts
import { Resemble } from '@resemble/node';
import 'dotenv/config';

const client = Resemble.client({
  apiKey: process.env.RESEMBLE_API_KEY!,
});

This code initializes the client with your API key. The TypeScript exclamation mark indicates the variable is defined.

Clone a Voice

create-voice.ts
const voice = await client.voices.create({
  name: 'Ma Voix Test',
  description: 'Voix de démonstration 2026',
  consent: 'https://exemple.com/consent.mp3'
});
console.log(voice.uuid);

This request creates a new cloned voice. Provide a consent file and 3-5 minutes of clear audio for best results.

Generate Speech

generate-speech.ts
const audio = await client.voices.generate({
  voiceUuid: 'uuid-de-votre-voix',
  body: 'Bonjour, ceci est un test Resemble AI en 2026.',
  outputFormat: 'mp3'
});
console.log(audio.audio_src);

Generate an audio file from text. The result contains a temporary URL valid for 24 hours.

Best Practices

  • Always store the API key in environment variables
  • Limit text length to 3000 characters per request
  • Use high-quality voices for client projects
  • Save voice UUIDs in your database
  • Test latency with short phrases before deployment

Common Errors to Avoid

  • Forgetting to add the audio consent when cloning
  • Exceeding the free quota without upgrading to a paid plan
  • Ignoring 429 errors (rate limiting) without retry logic
  • Storing the API key directly in source code

Going Further

Explore advanced features like real-time streaming and emotions. Discover our voice AI resources for more in-depth learning.