@ericc/edge-tts

Generate speech audio from text using Microsoft Edge's text-to-speech API.

Heavily inspired by rany2/edge-tts and SchneeHertz/node-edge-tts

Features

• Zero dependencies
• Using standard web APIs. Should work in all (modern) JS environment
• Provides subtitle/caption data

Installation

# Deno
deno add @ericc/edge-tts

# NPM
npx jsr add @ericc/edge-tts

# Yarn
yarn dlx jsr add @ericc/edge-tts

# PNPM
pnpm dlx jsr add @ericc/edge-tts

# Bun
bunx jsr add @ericc/edge-tts

Usage

Check out the example usage and docs at the JSR page.

API Reference

Basic Usage

// Web
const { audio, subtitle } = await generate({
  text: "Hello, world!",
  voice: "en-US-JennyNeural",
  language: "en-US",
});

// Create an audio element and play the generated audio
const audioElement = new Audio(URL.createObjectURL(audio));
audioElement.play();

// Access subtitle data
console.log(subtitle);

Options

GenerateOptions

Options that will be sent alongside the websocket request:

• text (required): The text that will be generated as audio
• voice (optional): Voice persona used to read the message. Defaults to "en-US-AvaNeural"
  • Please refer to Language and voice support for the Speech service
• language (optional): Language of the message. Defaults to "en-US"
  • Please refer to Language and voice support for the Speech service
• outputFormat (optional): Format of the audio output. Defaults to "audio-24khz-96kbitrate-mono-mp3"
  • Please refer to SpeechSynthesisOutputFormat Enum
• rate (optional): Indicates the speaking rate of the text. Defaults to "default"
  • Please refer to Customize voice and sound with SSML
• pitch (optional): Indicates the baseline pitch for the text. Defaults to "default"
  • Please refer to Customize voice and sound with SSML

ParseSubtitleOptions

Options for parsing the subtitle:

• splitBy (required): The function will split the cues based on this option
  • "sentence": will split the text using Intl.Segmenter
  • "word": will split the text to X count of words for each cue
  • "duration": will split the text to X duration of milliseconds for each cue
• count (optional): Used when splitting by "words" or "duration"
  • When splitting by "words", count means the amount of words for each cue
  • When splitting by "duration", count means the duration in milliseconds for each cue
• metadata (required): Array of metadata received throughout the websocket connection

Credits

• rany2/edge-tts
• SchneeHertz/node-edge-tts