Create an AI Voice Assistant with ElevenLabs & Node.js
Learn how to build a command-line AI voice assistant using Node.js and ElevenLabs API for realistic text-to-speech and custom voice cloning.
Stock photo for illustration only, not from the actual event
- Build an AI voice assistant using ElevenLabs API and Node.js.
- Keep code lightweight with native fetch instead of extra HTTP libraries.
- Clone custom voices using just a 10-second target speaker audio sample.
- Integrate OpenAI to give your assistant intelligent Q&A capabilities.
If you want to give your application a natural, expressive voice that can even mimic specific speakers, ElevenLabs is the go-to service. Their API provides powerful text-to-speech (TTS) and voice cloning capabilities via simple HTTP calls, offering generous developer tiers suitable for hobbyist projects and prototypes.
In this tutorial, we explore how to wire up a simple AI voice assistant using Node.js and the ElevenLabs API, resulting in a functional command-line tool. Start by verifying your Node installation with node -v and installing ffmpeg using your system package manager, such as brew install ffmpeg or apt-get install ffmpeg.
Stock photo for illustration only, not from the actual event
The project directory structure for voice-assistant/ contains the following files:
.envfor storing API keysindex.jsas the main program entry pointtts.jsas a tiny wrapper around ElevenLabspackage.jsonfor dependency management
Create the project folder and initialize npm using mkdir voice-assistant && cd voice-assistant, npm init -y, and npm install dotenv openai. The implementation remains lightweight by avoiding extra HTTP dependencies, leveraging native fetch instead.
Utilizing native fetch in modern Node.js environments reduces unnecessary external dependencies like axios, streamlining small-scale prototypes such as voice assistants for faster deployment and easier maintenance.
Next, create a .env file to store your credentials:
ELEVENLABS_API_KEY=your_elevenlabs_key_hereOPENAI_API_KEY=your_openai_key_here(optional for LLM integration)
Load these runtime values, and if you want your assistant to answer questions instead of just echoing input, hook up a tiny OpenAI helper. Run node index.js to see the prompt, type a query, watch the LLM response in the console, and hear the spoken audio through your speakers.
ElevenLabs also allows you to upload a brief 10-second sample of a target speaker to generate a custom voice ID. Using a curl request to https://api.elevenlabs.io/v1/voices/add with your xi-api-key, multipart form data containing name=MyClone, and the sample file files=@/path/to/sample.wav, you can create customized voices.
The API response returns a voice_id, which you can plug into the voiceId argument of your synthesize() function to make your custom voice speak the LLM output.
Source: Dev.to
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment