Voice I/O

Browser-based Speech Recognition & Speech Synthesis with support for multiple languages and voices.

Quickstart

npm install @swankylegg/voice-io

import { VoiceIO } from "@swankylegg/voice-io";

// Initialize
const voice = new VoiceIO();

// Listen for speech
voice.setState(voice.states.LISTENING);

// Handle speech recognition
voice.config.onRecognitionResult = (results, bestTranscript) => {
  console.log("User said:", bestTranscript);

  // Respond with speech
  voice.setState(voice.states.RESPONDING, "I heard you!");
};

Demo

Try it out: Live Demo

Browser Support

This library uses the Web Speech API and requires browser support for:

• SpeechRecognition (or webkitSpeechRecognition)
• speechSynthesis

Voice Support

Voice I/O automatically detects available system voices for speech synthesis. Different operating systems and browsers provide different voices:

• macOS/iOS: High-quality system voices in multiple languages
• Windows: Microsoft system voices + installed language packs
• Android: System voices vary by device/manufacturer
• Chrome: Additional cloud-based voices when online

To get available voices:

voice.getVoices(); // Get all available voices
voice.getVoicesForCurrentLanguage(); // Get voices matching current language

To set a specific voice:

voice.setVoice("Samantha"); // Use a specific voice by name

Language Support

Voice I/O supports multiple languages for both recognition and synthesis. Available languages include:

• English (en-US)
• Spanish (es-ES)
• French (fr-FR)
• German (de-DE)
• Italian (it-IT)
• Japanese (ja-JP)
• Korean (ko-KR)
• Chinese (zh-CN)

Note: Actual language availability depends on browser support for both recognition and synthesis in the selected language. The library will automatically filter voices to match the selected language.

States

Voice I/O has four possible states:

• IDLE: Default state, not listening or speaking

  🎤🚫 | 🔈🚫

• LISTENING: Actively listening for speech input

  🎤🟢 | 🔈🚫

• THINKING: Transitional state between listening and responding (use for loading states)

  🎤🚫 | 🔈🚫

• RESPONDING: Currently speaking using speech synthesis

  🎤🚫 | 🔊🟢

License

This project is licensed under the MIT License - see the LICENSE file for details.