Text to Speech

Our AI team focuses on cutting-edge artificial intelligence research and applications.

Introduction

Text-to-Speech (TTS) is a fundamental task in speech processing and human–computer interaction that aims to convert written text into natural and intelligible speech. Over the past decades, TTS technology has evolved significantly, especially with the emergence of deep learning–based approaches that enable the generation of speech with high naturalness, expressiveness, and speaker similarity.

Modern TTS systems play a crucial role in a wide range of applications, including voice assistants, accessibility tools for visually impaired users, audiobook narration, language learning, and interactive dialogue systems. Recent advances such as end-to-end neural architectures and neural vocoders have greatly improved speech quality, prosody modeling, and the ability to generate expressive and emotionally rich speech.

Our research group focuses on developing advanced TTS systems by leveraging machine learning and deep learning techniques, in combination with linguistic knowledge and acoustic modeling. We investigate methods for improving speech naturalness, prosody control, and speaker adaptation, as well as building scalable and high-quality speech datasets. In addition, we explore robust TTS solutions for multilingual and low-resource languages, aiming to enhance the accessibility and adaptability of speech synthesis technologies across diverse real-world scenarios.

Contact: Dr. Nguyen Thi Thu Trang | ✉️ trangntt@soict.hust.edu.vn

Research Direction

Voice Cloning: Create an artificial simulation of a person’s voice from a small amount of voice data.
Multilingual TTS: Generate speech in multiple languages while preserving the same speaker’s voice.
Expressive TTS: Produce human-like speech with emotional expressions such as angry, sad, surprise, and happy.
Controllable TTS: Allow users to adjust speech characteristics at inference time, including pitch, speaking rate, energy, and emotion.