Knowledge Hub

Nova: A Multilingual AI Voice Assistant in the Browser

Engineering articleBy Naveed Sohail Gung2025-05-20ReactTypeScriptAIVoice

How I built Nova — a browser-based AI voice assistant supporting English, Arabic, and French with real-time speech recognition, natural language understanding, and text-to-speech.

Article summary

How I built Nova — a browser-based AI voice assistant supporting English, Arabic, and French with real-time speech recognition, natural language understanding, and text-to-speech.

What this article covers

The Challenge Most voice assistants are platform-locked. I wanted a browser-native voice assistant that works on any device with a microphone — no app install required — and supports three languages out of the box.

What this article covers (2)

Speech Recognition Nova uses the Web Speech API's SpeechRecognition interface with continuous mode and interim results. Language switching is handled by dynamically setting the lang attribute: recognition.lang = currentLanguage; // 'en-US', 'ar-SA', or 'fr-FR' recognition.interimResults = true; recognition.continuous = true;

What this article covers (3)

Natural Language Processing Transcribed text is sent to an AI backend that classifies intent, extracts entities, and generates a natural-language response. The system prompt is language-aware, so Nova responds in the same language the user speaks.

Topics and technologies

  • React
    Browse related articles in the knowledge hub.
  • TypeScript
    Browse related articles in the knowledge hub.
  • AI
    Browse related articles in the knowledge hub.
  • Voice
    Browse related articles in the knowledge hub.
  • Tailwind CSS
    Browse related articles in the knowledge hub.

Related navigation