Joshua

You can find two examples here:

• https://github.com/xenova/kokoro-web
• https://github.com/hexgrad/kokoro/tree/main/kokoro.js/demo

The model itself has a maximum context length, so you can't feed everything through the model at once, unfortunately. To solve this, I implemented streaming in v1.2.0, which you can use as follows:

import { KokoroTTS } from "kokoro-js";

const model_id = "onnx-community/Kokoro-82M-v1.0-ONNX";
const tts = await KokoroTTS.from_pretrained(model_id, {
  dtype: "fp32", // Options: "fp32", "fp16", "q8", "q4", "q4f16"
  // device: "webgpu", // Options: "wasm", "webgpu" (web) or "cpu" (node).
});

const text = "Kokoro is an open-weight TTS model with 82 million parameters. Despite its lightweight architecture, it delivers comparable quality to larger models while being significantly faster and more cost-efficient. With Apache-licensed weights, Kokoro can be deployed anywhere from production environments to personal projects. It can even run 100% locally in your browser, powered by Transformers.js!";
const stream = tts.stream(text);
let i = 0;
for await (const { text, phonemes, audio } of stream) {
  console.log({ text, phonemes });
  audio.save(`audio-${i++}.wav`);
}

This is great! Does it work for nested cases too? For example,

Last week she said, “Hi there. How are you?”

should remain a single chunk.

Hi there - we recently fixed this issue and will release a new version for it soon!

Hey! Oh that's awesome - great work! Feel free to adapt any code/logic of mine as you'd like!

For this demo, ~150MB if using WebGPU and ~120MB if using WASM.