Getting started

Spellophone is espeak-ng compiled to WebAssembly, with a small TypeScript API on top. It synthesizes speech and transcribes text to phonemes, entirely on the device running it.

Install

pnpm add @spelling-creator/spellophone

The package is made of ES modules and needs Node.js 24 or later, or any browser that supports WebAssembly and DecompressionStream (every current one). Where WebAssembly is blocked, or in browsers as old as Internet Explorer 11, it can run as plain JavaScript instead, and the ES5 build also comes as a classic script; see Without WebAssembly.

Node.js

import { writeFile } from "node:fs/promises";
import { createEspeak, encodeWav } from "@spelling-creator/spellophone";

const espeak = await createEspeak();

espeak.setVoice("en-us");
console.log(espeak.phonemes("hello world")); // həlˈoʊ wˈɜːld

const result = espeak.synthesize("Hello world.");
await writeFile("hello.wav", encodeWav(result.samples, result.sampleRate));

Every language in espeak-ng works out of the box. Dictionaries are read from the package's data directory the first time a voice needs them.

More about Node.js

Browsers

The browser build fetches its data over HTTP. Tell it where the data is: either a copy of the package's espeak-ng-data directory that you host, or the jsDelivr CDN.

import { createEspeak, toFloat32 } from "@spelling-creator/spellophone";

const espeak = await createEspeak({ cdn: "jsdelivr" });

espeak.setVoice("en-us");
const result = espeak.synthesize("Hello from the browser.");

const context = new AudioContext();
const buffer = context.createBuffer(
  1,
  result.samples.length,
  result.sampleRate,
);
buffer.copyToChannel(toFloat32(result.samples), 0);
const source = context.createBufferSource();
source.buffer = buffer;
source.connect(context.destination);
source.start();

The @spelling-creator/spellophone import resolves to the Node build under Node.js and to the browser build everywhere else. You can also import @spelling-creator/spellophone/node or @spelling-creator/spellophone/browser explicitly.

More about browsers, bundlers and workers

What you get back

synthesize returns 16-bit mono PCM samples, the sample rate (22050 Hz) and the events espeak-ng raised while speaking: where each word starts, in the text and in the audio, SSML marks, sentence ends and optionally every phoneme.

phonemes returns a string: IPA by default, or espeak-ng's own ASCII phoneme names with { ipa: false }.

Events, SSML and phonemes

Examples

The repository has two runnable examples:

• examples/node: writes a WAV file and prints phonemes and word timings.
• examples/browser: a Vite page with voice selection, playback through the Web Audio API and a WAV download. It self-hosts the data in development and switches to jsDelivr with ?cdn in the URL.

pnpm example:node
pnpm example:browser