🗣️ tts-synth

A lightweight, premium, formant-based text-to-speech engine written entirely in TypeScript.

tts-synth models the human vocal tract by simulating resonant frequencies (formants) to generate speech. Unlike modern neural-network-based TTS, it requires zero machine learning weights, has a near-zero memory footprint, and generates audio in real-time with extreme efficiency.

---

✨ Features

• 🧩 Zero Dependencies: Core synthesis logic is pure TypeScript/JavaScript.
• 🎙️ Formant Synthesis: Models human speech using 5-formant cascades (F1–F5).
• 📏 Compact & Fast: Generates WAV audio in milliseconds; ideal for low-resource environments.
• 🎚️ Deep Customization: Adjust pitch (f0), speaking rate, and vowel/consonant characteristics.
• 🌐 Browser & Node.js Ready: Works anywhere Node.js or a modern browser is available.
• 📄 Advanced Phonetic Engine: Robust Letter-to-Sound (LTS) rules for English pronunciation.

---

🚀 Quick Start

Installation

# Clone the repository
git clone https://github.com/raghav/tts-synth.git
cd tts-synth

# Install dependencies (only for development/demos)
npm install

# Build the project
npm run build

Basic Usage (Node.js)

import { toWav } from 'tts-synth';
import { writeFileSync } from 'node:fs';

// Simple one-liner to generate a WAV buffer
const wav = toWav('Hello, this is a test of the formant synthesizer.');
writeFileSync('output.wav', wav);

Advanced Usage

Customize your voice by passing an options object to the synthesizer:

import { synthesize, toWav } from 'tts-synth';

// 1. Adjusting Pitch & Rate
const slowHighPitch = toWav('I am speaking slowly with a high voice.', {
  pitch: 200,   // Hz (default: 120)
  rate: 0.6     // multiplier (default: 1.0)
});

// 2. Getting raw PCM data (Float32Array)
const pcm = synthesize('Extracting raw audio data.', {
  pitch: 110,
  rate: 1.2
});
console.log(`Generated ${pcm.length} samples at 22050Hz.`);

---

🛠️ Development & Testing

The project includes several built-in scripts to help you get started:

Script	Description
npm run build	Transpile TypeScript to the dist directory.
npm run dev	Watch mode for development.
npm run demo	Runs the Node.js example to generate several WAV files in the samples/ directory.
npm run typecheck	Run the TypeScript compiler in noEmit mode.

Examples

• Node.js Example: Run npm run demo to see the engine in action. It will save demo-output.wav, demo-high-pitch.wav, and demo-slow.wav in the samples/ root.
• Browser Example: Open examples/browser.html in any modern web browser to use the interactive synthesis UI.

---

🧠 How it Works: Formant Synthesis

Formant synthesis models the human vocal tract as a series of resonators.

1. Excitation: A buzz (glottal pulse) or hiss (noise) is generated.
2. Filtering: The signal passes through several band-pass filters (resonators) representing the oral and nasal cavities.
3. Formants: By shifting the frequencies ($F_1, F_2, F_3$) of these filters, the engine transitions between different vowels and consonants.

This project implements a simplified version of the Klatt synthesis model, using rules to map textual phonemes to specific formant trajectories over time.

---

📜 License

This project is licensed under the MIT License - feel free to use it for whatever purpose you like!