RunPod Serverless Endpoint for Voice Cloning

Overview

• REST API
• Call Endpoint API with YouTube link and prompt
• Downloads first 60 seconds of provided YouTube audio
• Invokes Chatterbox TTS model
• Returns base64 encoded WAV audio

Deploy

• Get a RunPod account, maybe use my referral link

Deploy endpoint (docs):

• Go to https://console.runpod.io/serverless
• -> New Endpoint
• -> GitHub Repo, choose https://github.com/geronimi73/runpod_chatterbox
• Check Endpoint ID of your deployed endpoint at https://console.runpod.io/serverless
• Create an API key at https://console.runpod.io/user/settings

Usage

JavaScript

const requestConfig = {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "Authorization": "Bearer " + RP_API_KEY
  },
  body: JSON.stringify({
    "input": {
      "prompt": "Hello world",
      "yt_url": "https://www.youtube.com/shorts/jcNzoONhrmE",
    }
  })
};
const url = "https://api.runpod.ai/v2/" + RP_ENDPOINT + "/runsync";
const response = await fetch(url, requestConfig);
let data = await response.json();
data = data.output

// audio data in data.audio_base64

Issues

• Takes 3-4 Minutes to init worker