VoiSona Talk API TypeScript Library

English | 日本語

A TypeScript-based library for the VoiSona Talk API (v0.9.0), providing a type-safe and intuitive way to integrate speech synthesis and text analysis.

What is 'VoiSona Talk'?

VoiSona Talk is an AI-powered speech synthesis software that uses deep learning technology to generate high-quality, natural-sounding human voices. It allows users to create expressive narrations and dialogues with fine-grained control over emotions and speech parameters.

This library allows you to interact with the VoiSona Talk editor's local API, enabling automated speech generation from your TypeScript/JavaScript applications.

Important Notice on Usage

• Licenses: To use this library, you must possess a valid license for the VoiSona Talk editor and each voice library you intend to use.
• Character Guidelines: Each voice library character has its own set of usage guidelines (e.g., Koharu Rikka Guidelines). You are responsible for reviewing and adhering to the specific guidelines of each character when using their voices.

Features

• Type Safety: Full TypeScript support for all OpenAPI schemas.
• Easy Synthesis: synthesizeAndWait handles enqueuing and polling automatically.
• Default WAV Output: Synthesis defaults to .wav file output in the output/ directory.
• Emotion Control: getStyleWeights helper for intuitive style/emotion management by name.
• Bulk Processing: bulkSynthesize for handling multiple requests with concurrency control.
• Environment Support: Easy configuration via .env files.

Installation

pnpm add voisona-talk-api

Quick Start

Before using the library, ensure your VoiSona Talk editor is configured correctly. See the Connection Guide for step-by-step instructions.

import { VoisonaClient } from 'voisona-talk-api';
import 'dotenv/config';

const client = new VoisonaClient({
  email: process.env.VOISONA_EMAIL,
  password: process.env.VOISONA_PASSWORD,
});

// Simple synthesis (defaults to .wav file output in output/ directory)
const result = await client.synthesizeAndWait({
  text: 'こんにちは。VoiSona Talkのテストです。',
  language: 'ja_JP',
});

console.log(`Audio saved at: ${result.output_file_path}`);

Advanced Usage

Emotion & Style Control

Manage voice styles (emotions) using names instead of index arrays.

const voiceInfo = await client.getVoiceInformation('voice_name', '1.0.0');

const styleWeights = client.getStyleWeights(voiceInfo, {
  Happy: 0.8,
  Angry: 0.2,
});

await client.synthesizeAndWait({
  text: 'わーい！',
  language: 'ja_JP',
  global_parameters: { style_weights: styleWeights },
});

Global Parameters

Fine-tune the output with various parameters.

await client.synthesizeAndWait({
  text: 'もっと速く！',
  language: 'ja_JP',
  global_parameters: {
    speed: 1.5, // 0.2 to 5.0
    pitch: 100, // in cents (-600 to 600)
    intonation: 1.2, // expressive range
  },
});

Bulk Synthesis

Process multiple messages efficiently.

const results = await client.bulkSynthesize(
  [
    { text: 'Message one', language: 'ja_JP' },
    { text: 'Message two', language: 'ja_JP' },
  ],
  { concurrency: 2 },
);

Custom Pronunciation

If a word is read incorrectly, you can specify the correct Katakana pronunciation:

await client.synthesizeWithPronunciation(
  {
    language: 'ja_JP',
    text: '焦る必要はありません。',
  },
  {
    焦る: 'アセル', // Fix "Kogeru" to "Aseru"
  },
);

Development

Prerequisites

• Node.js v18+
• pnpm
• VoiSona Talk Editor (Running locally)

Scripts

• pnpm run build: Build for CJS and ESM.
• pnpm test: Run unit tests.
• pnpm run test:coverage: Generate coverage report.

License

MIT