Year: 2025
Type: article
Abstract: The rapid advancements in Artificial Intelligence and Deep Learning have significantly improved Text-To-Speech (TTS) technology, enabling more accurate and natural voice conversion. This project presents a Voice Cloning System that leverages a Transformer-based encoder and a GAN-based vocoder to generate high-quality, natural-sounding speech from text. The system supports both Text-to-Speech (TTS), where textual input is converted into a default synthesized voice, and Voice Cloning, which allows the replication of a new voice using a short audio sample. By employing a one-shot learning approach, the system enables speaker adaptation with minimal training data, making it efficient and scalable for real-world applications. The Transformer-based encoder effectively captures linguistic and prosodic features, while the GAN-based vocoder enhances the realism of the generated speech by refining spectral details. The model's ability to generalize across different speakers ensures robustness, even when trained on limited datasets. This project highlights the potential of deep generative models in speech synthesis and their impact on various domains, including assistive technology, where it can help individuals with speech impairments communicate more naturally, personalized virtual assistants that adapt to user preferences, and entertainment industries for voiceovers and character dubbing.
Authors: Md. Sadik, P. Vijaya, Y. Revathi, V. Siva Naga Tanuja, B. Soudhamini, R. Vaishnavi
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4409868619